It is perhaps fitting that a book that frets about the limits and possibilities of interpretation should be hard to pin down. The title itself is a riddle crying out to be deciphered. Eco anticipates this on page one:
What has Kant got to do with the platypus? Nothing. As we shall see from the dates, he couldn’t have anything to do with it. And this should suffice to justify the title and its use of an incongruous set that sounds like a tribute to Borges’s ancient Chinese encyclopedia.
He goes on:
So what is this book about? Apart from the platypus, it’s about cats, dogs, mice, and horses, but also chairs, plates, trees, mountains, and other things we see every day, and it’s about the reasons why we can tell an elephant from an armadillo (as well as why we don’t normally mistake our wife for a hat). This is a formidable philosophical problem that has obsessed human thought from Plato to present-day cognitivists, and it is one that even Kant (as we shall see) not only failed to solve but didn’t even manage to express in satisfactory terms. So you can imagine how much chance I’ve got.
Playful, paradoxical, exasperating, oblique, disarming: this is the Eco style, take it or leave it. Not that Eco pretends that the dazed reader’s problems are his alone; he admits to a certain unclarity in the ruminations that make up the six essays of Kant and the Platypus:
Written therefore in a spirit of indecision and beset by numerous doubts, these essays spring from my feeling of not having honored certain debts incurred when I published A Theory of Semiotics in 1976…. These debts concerned the problems of reference, iconism, truth, perception, and what in those days I used to call the lower threshold of semiotics.
Nothing, characteristically, is said to explain that last opaque phrase, which seems to capture the predicament of the struggling reader. (Semiotics, for the uninitiated, is the general study of signs—whether they take the form, to name only a few examples, of letters, words, pictures, sounds, musical notes, or physical gestures or characteristics. Semiotics is concerned with the syntax, semantics, and pragmatic function of signs; it seeks to show how they serve as symbols, convey meaning, can be interpreted.)
Eco’s six chapters take up each of the subjects in the paragraph I have just quoted, with many examples aimed at showing how we use language to refer to one object rather than another and how we classify many different kinds of objects from oysters to planets. How, then, to proceed? I am going to try to distill some clear theses from Eco’s shifting text, and evaluate their cogency. This may not do justice to Eco’s historical and linguistic erudition, which ranges from cartoons to medieval history, from structuralism to neurophysiology, from obscure Latin texts to an imaginary town called Vanville (named after Willard van Orman Quine), but it should focus the issues raised by his work a bit more sharply.
The first chapter, “On Being,” starts with the sensible thought that no theory of signs can avoid addressing reality—that which signs signify. The whole point of signs, particularly human language, is to represent the world beyond signs—tables, cats, mountains, etc. Eco prefers to formulate this elementary truth by asking what induces us to produce signs: “What makes us talk?” he inquires. This is an odd way to put the point, since presumably what causes us to produce speech acts are the motives we have in saying things. Why did I just say “It’s snowing”? Because I wanted to convey to my audience the information that it is snowing. The snow alone will not induce me to say anything unless I have a reason to speak of it. So it is not quite accurate to say that “the Dynamical Object is what drives us to produce semiosis”; it is the desire to communicate. (Semiosis is the process of interpreting or producing a meaningful sign, as when I say “It’s snowing” or make a drawing of falling snow or try to act like snow: it is symbolic activity of any sort.) Still, Eco’s underlying point is surely correct: there is no point in signs unless there is an extralinguistic reality to signify. In the beginning was the world, not the word.
This leads Eco into a lengthy discussion of what he insists on calling Being, following Heidegger, where we encounter such sentences as: “Without speech there is no more entity: as the entity flees, there arises the nonentity, in other words, nothingness” and “If being-there is the entity that fully recognizes the semiosical nature of its relation with the entities, it is not necessary to duplicate Seiende and Sein.” Disconcertingly, Eco confuses two questions: Leibniz’s question of why there is something rather than nothing, and Descartes’s question of whether we can establish that anything exists. He thinks he has answered Leibniz’s question by agreeing with Descartes that our own existence as thinkers is indubitable. But that is not Leibniz’s problem: Leibniz is accepting that a great many things exist—he is not raising an epistemological doubt—but is then asking why the universe is replete with Being instead of just a vast emptiness. The question for Leibniz is not whether things exist, but why they exist: Why is there something rather than nothing? Cartesian certainty about selves and their thoughts has no bearing on that question, since we can ask also why these things exist instead of never having come into existence.
But Eco also sometimes espouses a view of reality that has become only too familiar in this “postmodern” era, namely that reality is somehow a product of human linguistic practices. He writes: “for being to exist, it is necessary to say as well as to think”; “being can be nothing other than what is said in many ways”; “being…is radically the effect of language and nothing else but the effect of language”; “the world as we represent it to ourselves is an effect of interpretation.” The thought here is that we categorize the world by means of language, therefore the world is “an effect of language.” But this conclusion is absurd, taken literally, and the inference patently invalid. There was obviously a universe before language came along, so the universe cannot be the product of language. On the contrary, language is a product of the world of brains, vocal organs, sound waves, and so forth. There could be no language without a language-independent world to give rise to it, phylogenetically as well as ontogenetically. And the inference is a resounding non sequitur, because from the fact that we categorize reality by means of language it does not follow that there are no categories in the world independent of language. For we might simply be responding to, and recording, the categories that exist independently of our linguistic acts.
We categorize objects by means of their shapes and sizes, for example, but objects have their shapes and sizes independently of the words we use: the words serve merely to record properties that things have independently of language, not to create these properties of things—we cannot make a round thing square just by calling it so. A tree will be taller than a blade of grass whether we are there to observe it or not. Eco is flirting here with a kind of linguistic idealism—the thesis that there is nothing more to reality than our symbolic interpretations. But interpretations presuppose reality, not the other way around.1
This linguistic idealism is encour-aged by the curious idea of reality as an amorphous continuum: “We use signs,” Eco says,
to express a content, and this content is carved out and organized in different forms by different cultures (and languages). What is it made from? From an amorphous stuff, amorphous before language has carried out its vivisection of it, which we will call the continuum of the content, all that may be experienced, said, and thought: the infinite horizon, if you will, of that which is, has been, and will be, out of both necessity and contingency. It would seem that before a culture has organized it linguistically in the form of content, this continuum is everything and nothing and therefore eludes all determination.
But this idea of the formless continuum is pure myth, a piece of metaphysical nonsense, an unredeemable metaphor. It can hardly be the physical world of atoms in the void, since they have “determinations” and laws that govern them. And it makes no sense to conceive of a reality that has no properties whatsoever; even the most formless fluid has properties of one kind or another. The idea of propertyless substance is conceptually incoherent; even Kant’s thing-in-itself had properties, whether we could know them or not.
Equally incoherent is the idea that experience is an indeterminate continuum before language has got to work in “segmenting” it. Experience inherently presents the perceived world to us as containing spatial objects with determinate observable properties; there is no such thing as some kind of featureless flow of sensation that has no structure in advance of linguistic intervention. Reality and experience have to be, in some objective way or another, possessed of some properties and not others; they cannot be a mere nothing—and yet a something—that awaits our linguistic attention before assuming a specific shape. Of course, some classifications are humanly imposed and have no antecedent existence—as with the difference between money and mere paper—but the world as a whole cannot be conceived as intrinsically without form and content until human linguistic practices confer categories upon it. On the contrary, we generally discover the properties of things; we don’t invent them.
The idea that reality itself is brought into being by acts of interpretation is clearly wrong, but there is another thesis—often confused with it—which is not similarly absurd, and to which Eco also assents. This is the thesis that all awareness of the world is mediated by interpretation—that there is no such thing as direct, interpretation-free cognition of reality. We impose categories on the world as we interpret the signs all around us; it is not that the world simply reveals its categories to us. We are always taking one thing to be a sign of another, performing an act of symbolic inference, decoding something.
This process is to be clearly distinguished from merely bringing things under some general category, as when you perceive an apple as red or think of an acquaintance as untrustworthy. It is a virtual truism that all mental representation is “aspectual,” in the sense that it cannot represent everything about an object but only some aspects of it; but this is quite different from saying that all mental representation involves interpreting one thing as a sign of another. Eco writes: “The moment it appears before us, being arouses interpretation; the moment we can speak of it, it is already interpreted.” This goes with the idea that all awareness of reality is inferential or “abductive,” to use C.S. Peirce’s term. (Eco places himself in the tradition of Peirce, generally acknowledged to be the founder of semiotics.) Inference requires premises to infer from, and a premise is precisely a symbol of some sort; if cognition is essentially interpretational, then it must also involve inferring a conclusion from premises, where a premise functions as a “sign” of the conclusion. Hence the idea that perception and cognition are semiotic or interpretative processes, species of inference from signs to signified.
This thesis, often taken to be a truism, is at the heart of the semiotic picture of the knowing mind, and it is clearly dear to Eco’s heart: if all knowledge involves interpretational inference, then the key to understanding the human mind is to study the workings of semiosis—of symbol-manipulating activity. It might be said that Eco’s main concern in this book is to reconcile the semiotic conception of cognition with the fact that our awareness is somehow constrained by the nature of the external things of which we are aware. I shall therefore spend a little time showing that the semiotic conception of cognition in general is not only mistaken but also logically incoherent: cognition cannot in general consist in the interpretation of signs. (This is not to deny that it may always be carried out in signs: thoughts can be implemented in signs without being about signs; that is, not all words refer to other words, but to worldly things like cats and tables. The semiotic conception of cognition is not at all the same as the idea that there is a “language of thought” that serves as the vehicle of our cognitive activities.)
Semiosis, as has been said, occurs when a mind takes one thing x as a sign of another thing y; it involves a relation between a mind and two objects of awareness, x and y. Thus we can take clouds to be a sign of rain or a portrait to be a depiction of Winston Churchill or the word “London” to refer to the capital of England. The sign x is the primary object of awareness that enables the mind to take the second, signified object y as its object of thought: in other words, we make an inference from the sign to its referent. We hear a sentence uttered or look at a picture or notice clouds and we perform an act of interpretation on these perceived entities, taking them to be signs of something beyond themselves. So we are aware of the second signified object y by means of semiosis practiced on the first sign object x, which is our primary object of awareness.
But what of that first object, the sign itself: How are we aware of it? Are we aware of it by means of some other act of semiosis? If we are, then there must be a further sign x that we are taking as a sign of the initial sign x, because it is in the structure of the act of semiosis that there must be a double awareness—something must be interpreted as standing for something else. For any object that we are aware of there must be a sign that we are also aware of and from which we infer the first object. But that just raises the question of how we become aware of x: Is it by semiosis or not? If it is, then we need a further sign x to be the sign of x*—which obviously raises the same question about x. And now we are off on an infinite regress, which means that nothing can become an object of awareness unless some other thing already is, which means that awareness of anything is impossible. The semiotic conception of cognition, if taken to be universally true, undermines itself.
But if we are not aware of the initial sign x by means of semiosis, then not all awareness is semiotic in nature; we are aware of some things without interpretation, directly, noninferentially—these things being signs themselves. These signs are indeed merely objects of ordinary perception, physical entities like acoustic signals or marks on paper or pigmented pieces of canvas or masses of water vapor in the sky. The upshot is that if all awareness involved interpreting one object of awareness as standing for another then no awareness of anything would be possible, since to be aware of one thing would require an infinite series of acts of awareness corresponding to the series of signs x, x*, x**, etc. But awareness of things is possible, so at least some awareness is not semiotically generated.
To make this abstract argument concrete consider the interpretation of speech. I hear you utter the sound “London” and I begin to think about London: first I am aware of the sound you produce, then I interpret this as standing for London, so that my thought about London is semiotically mediated. But what about my awareness of the sound? Did I interpret anything else in order to become aware of it? It certainly seems not, but suppose for the sake of argument that we allow that I did. Then I had to be aware of something S, distinct from the sound, that operated as a sign of the sound, and this something S I was aware of without inferring it from some other sign—on pain of never being able to be aware of anything. The plain fact is that we are aware of some things directly, without undertaking semiosis on some prior sign, and these things can function as the basis for an awareness of something else, by means of an interpretational inference. So it is not true that all cognition of reality is mediated by interpretation; semiosis is not a universal feature of cognition.
Bertrand Russell had this right years ago in his distinction between knowledge by acquaintance and knowledge by description.2 Russell thought that there are certain privileged entities that we are aware of by direct acquaintance—such as our current sense-data—and that we can extend our knowledge by describing these privileged items in certain ways, in effect taking them to be signs of other things. When I am acquainted with a sense-datum of red I use this as the basis of the description “the object that causes this sense-datum,” and hence I come to think, by means of semiosis, about the external object that caused my object of acquaintance. What is right about this is not the sense-datum theory of perception but the idea that some awareness is not inferential but direct, pre-semiotic. The canonical form of semantic description is “the referent of S,” where S is a sign; but this description refers to the sign S, which is the basis for the interpretational act—it is what we interpret when we assign a particular referent to it. Signs enable us to be aware of other things by means of semantic interpretation, but the awareness of signs itself is not in general an interpretational phenomenon.
Of course, to be aware of signs, as of anything else, involves classifying the sign in some way, perceiving it as having certain properties: for example, one hears the sound of the word “London” as having a certain auditory quality. But it is a mistake to refer to this mental act as interpretation, since interpretation must always involve interpreting one thing as a sign of another, and not merely seeing an object under a certain aspect. Conflating the latter with the former converts a truism into a highly controversial and (as I have argued) self-defeating theory about how cognition of objects is achieved. The culprit in all this is a loose and ambiguous use of the word “interpretation,” probably one of the most misused words in the contemporary humanities.
It is important to keep this clear when considering perception and semiosis, one of Eco’s main interests in these essays. Eco expresses sympathy for Peirce’s thesis that perception is inferential and semiotic in nature: that is, it involves the interpretation of sensory signs. Certainly some of our perceptual judgments involve interpretation: when on a dark night I see a shadowy something as a cat I do so by means of an interpretation of some more primitive component of my perceptual experience, that of an animal-like shape moving stealthily across the room. The question has always been how far back we can press the inferences that go into the perception of objects. Some say that we only directly see colors and surfaces and two-dimensional shapes; others (like Russell) go further and say that even these are inferred from the character of the experience we have, viewed purely psychologically (“sense-datum theories”). This is the old question of what is given in perception as opposed to what is brought to perception, what is directly present in the experience itself and what we infer from it given our background assumptions—though Eco, for all his erudition, seems unaware of the enormous and sophisticated literature on these topics within analytical philosophy (which he seems to take a somewhat dim view of).
Still Eco is clearly impressed with the point that something about perception is forced upon us by the object, so that we do not see the world purely by means of freely projected interpretations (how this is meant to be consistent with his earlier linguistic idealism he does not explain: presumably it is that “amorphous continuum” making itself felt). He remarks, very plausibly, that the sun looks the same both to geocentric and heliocentric theorists of the solar system. The stimulus imposes its own imprint on the nature of the perceptual encounter. So all is not unconstrained semiosis, after all. Here I would also suggest considering introspection, the awareness we have of our own mental states, as when I judge that I am having a sensation of red. It is highly implausible to suppose that this is a matter of mere interpretation, as if I could freely decide what to judge about my own mental states according to my semiotic whim. I do not infer that I have a sensation of red from some sign or other; rather, this mental state directly forces itself on my attention. Again, semiosis has its limits as an account of cognition in general, though it clearly plays a role in some cognitive acts—those that really do involve the interpretation of genuine signs, as with the understanding of speech.
By far the longest essay in Kant and the Platypus, nearly 100 pages, is entitled “Cognitive Types and Nuclear Content,” and is concerned with recognition, classification, and empirical concepts. Eco invites us to consider the Aztecs when first they encountered the horses of the invading Spaniards. These were strange animals to them, and they had no word for horses, yet their perceptual experience furnished them with a means for recognizing new instances of the type: they formed a mental representation that applied to horses in general. This representation, which Eco labels a Cognitive Type (CT), proceeds from the particular to the general; it takes individual token horses and derives the type “horse,” which is then used by the mind in classifying new tokens. It is the basis of perceptual recognition, a cognitive structure inside the minds of the Aztecs.
What kind of thing is this CT? What features does it contain? How does it relate to the meaning of the word “horse”? The CT the Aztecs have, though useful for visual recognition, is not the same as a dictionary definition of “horse.” Nor is it an encyclopedia entry for the equine species, or the zoological essence of horsiness; rather, it is something that works at a perceptual level to divide the world up into animal types. It is perhaps the most primitive form of classification we have. But it is not clear what it involves: What principle of similarity were the Aztecs working with? Did they select a paradigm horse and then judge every new instance by its similarity to that paradigmatic (the “prototype” theory of classification), or did they have a general “stereotype” that applies to all horses? How did they abstract away from the particular features of instances of the type in order to derive the general idea of the type? What color does the typical horse have, if any?
As Eco remarks, this is the old problem of “general ideas”: What is the general idea of a triangle—is it isosceles or equilateral or what? Eco declines to venture into the “black box” that lies inside a person’s head, where the cognitive apparatus functions, preferring to stick to the common-sense level of what people say to each other about horses and what they do in respect of horses. He distinguishes Cognitive Type from what he calls Nuclear Content (NC), which consists of the publicly accepted criteria for calling something a horse; the NC is close to the communal meaning of the term “horse.” The CT is the basis for the NC, but they are not the same thing: the NC is a shared linguistic construct that gives public expression to the many idiosyncratic CTs—e.g. horses of many different colors and sizes—present in the community. We might compare the CT to the perceptual means by which we recognize faces, and the NC to the verbal descriptions we might offer of faces (the latter will typically be coarser grained than the former). Finally, he defines the idea of the Molar Content (MC) of “horse,” which consists of what a competent zoologist or other expert might know about horses.
These seem like sensible distinctions, but theoretically they do not take us very far, and I am not sure how Eco intends us to profit from them. They are descriptive rather than explanatory. In the case of the eponymous platypus, which Eco takes to be emblematic, people were confused about what type of animal they were looking at, though they could recognize a platypus when they saw one: Was it a fish, a bird, a mammal? (In fact, it is an egg-laying mammal.) They suffered from acute classificatory uncertainty. This is an interesting piece of zoological history, vividly related by Eco, but I am unclear how it is meant to shed light on the process of concept formation. Some items may indeed be difficult to classify because they are not similar to anything already classified or because they are similar to too many things. There can thus be rational arguments about which classificatory decision to make. What does that tell us about the nature of empirical concepts? At most it seems to show that zoological taxonomy is not a quick exercise in casual inspection, but requires complex theoretical and empirical considerations. Eco does not, to my mind, succeed in connecting the history of the confusion over platypuses to the cognitive science of their classification in any very illuminating way. One has the sense that he is simply collating two enthusiasms, for intellectual history and for cognitive psychology.
He does, however, make the important point that we classify things in very different ways. Sciences like chemistry classify according to the laws and internal structure of things; common sense tends to go by surface sensory appearance. Artifacts like chairs and bottle openers are classified by function; games are grouped together by Wittgensteinian “family resemblance”; numbers possess abstract principles of similarity. There is no one notion of likeness here, so our concepts are accordingly various in what they require of objects. Traditional empiricists neglected this point to their detriment.
I must disagree, however, with Eco’s assertion that “recourse to the universal is not strength of thought but weakness of discourse. The problem is that man always talks in general while things are singular.” Of course things are singular, but they also share characteristics in common; talk of the universal is simply the way we register these similarities. If we had no general terms, but only singular terms, we would be limited indeed; in fact, we would not be able to say anything, since all statements involve identifying some particular and then attributing some general characteristic to it. Sentences cannot be made wholly of names but also require predicates. The singular and the general are complementary partners in thought and speech, not rivals.
At certain points in his book Eco takes up, and not always with happy results, some technical ideas of contemporary analytical philosophy of language, for example how, when we use a proper name, such as, say, “Umberto Eco,” we can reliably connect the use of the name to the person who bears it.3 The view of reference Eco favors he calls the “contractual theory.” So far as I can see, this is simply the insistence that what we refer to by a name is determined partly by pragmatic considerations—for example, the experts in the community from whom I picked up the name—and that it is sometimes unclear what a speaker is referring to when he uses a term for which there are several competing candidates. The reference of a term can reflect a multitude of contextual factors.
This is plausible enough and fairly orthodox in linguistics and philosophy of language; but it is quite unclear why Eco calls this a contractual theory, as opposed to a contextual theory, unless he means simply to be reminding us of the platitude that public linguistic meaning depends upon interpersonal agreement about what we mean by our words. However, here we see a welcome convergence between the conclusions of a prominent semiotician and the ideas of contemporary analytical philosophy of language. Eco is perhaps a little less original than he appears to think, but he does fasten onto some sound and sensible points about how the reference of our terms is determined.
The entire premise of semiotics as a discipline is that there can be a significant theory of everything we can call a sign—from symptoms of diseases to portraits of people, from traffic signals to natural languages, from the genetic code to clouds portending rain. This was the subject of Eco’s earlier book A Theory of Semiotics, and it figures intermittently in Kant and the Platypus, though here the focus is far more on basic cognitive processes and the relation between representation and reality. It is, as critics have pointed out, quite unclear whether it is possible to unify the different semiotic phenomena in a theoretically illuminating way. What unified nontrivial theory can be derived from treating human languages and measle spots as both instances of interpretable signs?
But there is a more fundamental question about the primacy of semiosis in understanding the mind, since the perceptual recognition of signs and other objects is not itself a process of semiotic inference, as I argued earlier. In this book Eco is beginning to see that perceptual processes are more basic than culture and systems of conventional signs, and indeed are what make these possible. But he is caught between this recognition and the habits of mind that shaped his earlier work on cultural semiotics, in which external conventional signs are the primary locus of cognitive activity.
The result is an uneasy exploration of basic cognitive processes such as perception, recognition, and conceptual categorization, in the light (or shadow) of a misplaced emphasis on the role of cultural signs in shaping our fundamental modes of knowing about the world. Maybe Eco should have considered the platypus’s own awareness of the world, and not just our efforts in trying to classify this taxonomically problematic little animal. Animal cognition, after all, does not rest on a system of cultural signs; still less is the world of the platypus “an effect of language.”
June 15, 2000
There is a good discussion of this point in John Searle’s Construction of Social Reality (Free Press, 1995). ↩
See his Problems of Philosophy (Oxford University Press, 1968, first published 1915), Chapter 5. ↩
He sets out in “Notes on Referring as Contract” to discuss the idea that certain terms in natural language, notably proper names, are “rigid designators,” which for present purposes we can take to mean that they are not synonymous with a set of descriptions but function as mere labels for objects. (For the cognoscenti: strictly speaking, a rigid designator is an expression, such as “Bertrand Russell” or “the number two,” that takes the same reference with respect to any possible world, unlike, say, “the inventor of the zipper”; but Eco is really, though he seems unaware of it, discussing the concept of a directly referential term, i.e., one whose sense is exhausted by its reference. And he appears to believe that this concept takes reference to be quite independent of the mental states of a speaker or community of speakers, as if terms referred by means of some kind of magical metaphysical glue.)
Eco interprets this theory (which is quite complex) as claiming that the reference of a term is “independent of all knowledge or intention or belief on the speaker’s part,” and he suggests that the theory requires that there be a Divine Mind that serves to pin terms down to objects, since human minds are not supposed to do it.
But this is a total misunderstanding of the theory in question. Speakers’ intentions do play a part in determining reference, both when the term is first introduced by means of an “initial baptism” and as it is passed from speaker to speaker. The introducer of the term must intend to be referring to some particular object with the term, as when a child is baptized “Umberto Eco,” and subsequent users of the term must intend to refer to the same object that earlier users intended. It is not that the term is supposed to be linked to its reference by some kind of divine intervention that occurs outside of the minds of speakers who use the term. The question is which kinds of intentions do the referential work, not whether intentions are involved at all. When I use the name “Umberto Eco,” do I intend to refer to whoever satisfies a certain set of descriptions (the so-called description theory), such as “the author of A Theory of Semiotics,” or do I intend to use the name to stand for the same person as the speaker I heard it from intended (the so-called causal chain theory)? Here Eco is attacking a straw man and revealing a startling lack of understanding of the material he is discussing. ↩