1
How to understand language? This is a preoccupation going back to the very beginning of our intellectual tradition. What is the relation of language to other signs? To signs in general? Are linguistic signs arbitrary or motivated? What is it that signs and words have when they have meaning? These are very old questions. Language is an old topic in Western philosophy, but its importance has grown. It is not a major issue among the ancients. It begins to take on greater importance in the seventeenth century, with Hobbes and Locke. And then in the twentieth century it becomes close to obsessional. All major philosophers have their theories of language: Heidegger, Wittgenstein, Davidson, Derrida, and all manner of âdeconstructionistsâ have made language central to their philosophical reflection.
In what we can call the modern period, from the seventeenth century, there has been a continual debate, with philosophers reacting to and feeding off each other, about the nature of language. I think we can cast light on this debate if we identify two grand types of theory. I will call the first an âenframingâ theory. By this I mean that the attempt is made to understand language within the framework of a picture of human life, behavior, purposes, or mental functioning, which is itself described and defined without reference to language. Language is seen as arising in this framework, which can be variously conceived as we shall see, and fulfilling some function within it, but the framework itself precedes, or at least can be characterized independently of, language.
The other type of theory I want to call âconstitutiveâ. As this word suggests, it is the antitype of the enframing sort. It gives us a picture of language as making possible new purposes, new levels of behavior, new meanings, and hence as not explicable within a framework picture of human life conceived without language.
These terms mark a major issue at stake between the two theories. But as it turns out, they are divided on a number of other major questions, and the two approaches can be contrasted on a number of other dimensions as well, and so they are sometimes referred to as the âdesignative-instrumentalâ and the âconstitutive expressiveâ theories respectively. And besides this, they even end up differing on the contours and limits of what they are trying to explain, viz., language; as well as on the validity of atomistic versus holistic modes of explanation. They belong, in fact, to very different understandings of human life. But we have to enter the labyrinth at some point, and I will do so at first through this contrasting of enframing versus constitutive, and gradually connect up with the other dimensions of controversy later.
2
The classical case, and most influential first form of an enframing theory, was the set of ideas developed from Locke through Hobbes to Condillac. I have discussed this in âLanguage and Human Nature.â Briefly, the Hobbes-Locke-Condillac (HLC) form of theory tried to understand language within the confines of the modern representational epistemology made dominant by Descartes. In the mind, there are âideasâ. These are bits of putative representation of reality, much of it âexternalâ. Knowledge consists in having the representation actually square with the reality. This we can only hope to achieve if we put together our ideas according to a responsible procedure. Our beliefs about things are constructed; they result from a synthesis. The issue is whether the construction will be reliable and responsible or indulgent, slapdash, and delusory.
Language plays an important role in this construction. Words are given meaning by being attached to the things represented via the âideasâ which represent them. The introduction of words greatly facilitates the combination of ideas into a responsible picture. This facilitation is understood in different ways. For Hobbes and Locke, they allow us to grasp things in classes, and hence make possible synthesis wholesale where nonlinguistic intuition would be confined to the painstaking association of particulars. Condillac thinks that the introduction of language gives us for the first time control over the whole process of association; it affords us âdominion over our imaginationâ [empire sur notre imagination].
The constitutive theory finds its most energetic early expression in Herder, precisely in a criticism of Condillac. In a famous passage of the treatise on the Ursprung der Sprache, Herder repeats Condillacâs fableâone might say âjust soâ storyâof how language might have arisen between two children in a desert. He professes to find something missing in this account. It seems to him to presuppose what itâs meant to explain. What itâs meant to explain is language, the passage from a condition in which the children emit just animal cries to the stage where they use words with meaning. The association between sign and some mental content is already there with the animal cry (what Condillac calls the ânatural signâ); the prelinguistic infants, like other animals, will cry out in fear when they are faced with danger, for instance. What is new with the âinstituted signâ is that the children can now use it to focus on and manipulate the associated idea, and hence direct the whole play of their imagination. The transition just amounts to their merely tumbling to the notion that the association can be used in this way.
This is the classic case of an enframing theory. Language is understood in terms of certain elements: ideas, signs, and their association, which precede its arising. Before and after, the imagination is at work and association takes place. Whatâs new is that now the mind is in control. Thus the cry of fear can be used to communicate the presence of danger to another, as a voluntary and not just a reflex action; as a way of designating danger, it can be used in reasonings about the antecedents and consequences of certain forms of threat.
This control itself is, of course, something that didnât exist before. But the theory establishes the maximal possible continuity between before and after. The elements are the same, combination continues, only the direction changes. We can surmise that it is precisely this continuity which gives the theory its seeming clarity and explanatory power: language is robbed of its mysterious character and is related to elements that seem unproblematic.
Herder starts from the intuition that language makes possible a different kind of consciousness, which he calls âreflectiveâ [besonnen]. That is why he finds a continuity explanation like Condillacâs so frustrating and unsatisfying. The issue of what this new consciousness consists in and how it arises is not addressed, as far as Herder is concerned, by an account in terms of preexisting elements. Thatâs why he accuses Condillac of begging the question. âThe Abbot Condillac ⌠had already presupposed the whole of language as invented before the first page of this bookâ [Der Abt Condillac ⌠hat das ganze Ding Sprache schon vor der ersten Seite seines Buchs erfunden vorausgesetzt].
What did Herder mean by âreflectionâ [Besonnenheit]? This is harder to explain. I have tried a reconstruction in âThe Importance of Herder.â We might try to formulate it this way: prelinguistic beings can react to the things which surround them. But language enables us to grasp something as what it is. This explanation is hardly transparent, but it puts us on the right track. To get a clearer idea we need to reflect on what is involved in using language.
You ask me what kind of shape this is, and I say âa triangleâ. Letâs say it is a triangle. So I get it right. But whatâs involved in getting it right in this sort of case? Well, it involves something like knowing that âtriangleâ is the right descriptive term for this sort of thing. Perhaps I can even tell you why: âsee, the thing is bounded by three straight sidesâ. But sometimes I recognize something and I canât say very much if anything about why. I just know that thatâs a classical symphony weâre hearing. Even in this case, however, I acknowledge that the question âwhy?â is quite in order; I can imagine working further on it and coming up with something, articulating what underlies my confidence that Iâve got it right.
What this brings out is that a certain understanding of the issue involved is inseparable from descriptive language, viz., that the word can be right or wrong, and that this turns on whether the described entity has certain characteristics. A being who uses descriptive language does so out of a sensitivity to issues of this range. This is a necessary proposition. We would never say that a being like a parrot, to whom we can attribute no such sensitivity, was describing anything, no matter how unerringly it squawked out the âright wordâ. Of course, as we prattle on, we are rarely focusing on the issue of rightness; we only do so when we get uncertain and are plumbing unexplored depths of vocabulary. But we are continuously responsive to rightness, and that is why we always recognize the relevance of a challenge that we have misspoken. Itâs this nonfocal responsiveness which Iâm trying to capture with the word âsensitivityâ.
So language involves sensitivity to the issue of rightness. The rightness in the descriptive case turns on the characteristics of the described. We might call this âintrinsic rightnessâ. To see what this amounts to, letâs look at a contrast case. There are other kinds of situations in which something we can roughly call a sign can be rightly or wrongly used. Suppose I train some rats to go through the door with the triangle when this is offered as an alternative to a door with a circle. The rats get to do the right thing. The right signal behavior here is responding to the triangle positively. The rat responds to the triangle door by going through it, we might say, as I respond to the triangle by saying the word.
But now the disanalogy springs to light. What makes going through the door the right response to the triangle is that itâs what brings the rats to the cheese in the end-chamber of the maze. The kind of rightness involved here is one which we can define by success in some task, here getting the cheese. Responding to the signal plays a role in completing the task, and thatâs why thereâs a âcorrect useâ of the signal. But this is a different kind of rightness from the one involved in aligning a word with the characteristics of some described referent.
But, one might object, doesnât the rat do something analogous? Doesnât it recognize that the triangle indicates âcheeseâ? It is after all responding to a characteristic of the triangle door, even if an instrumental one. The rat, we might say, aligns its action with a characteristic of this door, viz., that itâs the one behind which the cheese always is. So perhaps we might better âtranslateâ his understanding by saying that the triangle indicates ârush through hereâ. But this shift in translation alerts us to what is wrong with this assimilation. There are certainly characteristics of the situation in virtue of which ârush through hereâ is the right response to a triangle on a door. But getting the response right has nothing to do with identifying these characteristics or any others. Thatâs why the question, under what precise description the rat gets it rightââthatâs where the cheese isâ, or âwhere reward isâ, or âwhere to jumpâ, or whateverâis pointless and inapplicable.
What this example brings out is the difference between responding appropriately in other ways to features of the situation, on one hand, and actually identifying what these features are, on the other. The latter involves giving some definition, some explicit shape, to these features. This takes us beyond merely responding to them; or, otherwise put, it is a further response of its own special kind. This is the response we carry out in words. We characteristically define the feature in applying the word, which is why this application must be sensitive to issues of intrinsic rightness, to the fact that the word applies because of the defined features, else it is not properly a word.
By contrast, letâs call what the rat responds to a âsignalâ, marking by this term that the response involves no definition of features, but rather rushing through to reward. Otherwise put, where responding to a signal plays a role in some task, correct signal behavior is defined by success in that task. Unless this success is itself defined in terms of getting something intrinsically rightâwhich is not the case for winning through to cheeseâcorrect response to the signal need involve no definition of any particular characteristics; it just involves reacting rightly, and this is compatible with recognizing a whole host of such characteristics, or none at all: the rat just knows to rush through here; it knows from nothing about descriptions and qua what it should rush it.
The rightness involved in description is crucially different. We canât just define it in terms of success in some taskâunless we define this task itself in terms of what I called above intrinsic rightness. In other words, intrinsic rightness is irreducible to what we might call task rightness simpliciter: the account in terms of some task only works for language if we have already incorporated intrinsic rightness in our success criteria.
We might make this distinction in another way, in terms of notions of âawarenessâ. For a nonlinguistic animal A, being aware of X consists of Xâs counting in shaping Aâs response. A characteristically responds to X in a certain way: if X is food, and A is hungry, A goes for it, unless deterred; if X is a predator, A flees; if X is an obstacle, A goes around it, and so on. By contrast, linguistic awareness of X canât be reduced to or equated with its triggering a particular response, or range of responses, in certain circumstances. We could think of this as an awareness which is independent from, or can sit alongside of...