Acquaintance and Description Again
Wilfrid Sellars
University of Minnesota
Published in The Journal of Philosophy 46 (1949): 496-504.
I
My aim in this brief essay is to analyze a nexus of ideas which has played a most significant role in the development of the climate of contemporary epistemological opinion. Springing from Lord Russell's work in the early years of this century, and finding its clearest formulation in his lectures on The Philosophy of Logical Atomism,1 this cluster of ideas soon gave convincing evidence of its power to define a point of view, and today there are few empirically-minded philosophers who do not show its influence. At the center of this conceptual structure is the thesis that has come to be known as the Principle of Acquaintance, and even, on occasion, as the Principle of Empiricism. This thesis has been given many formulations. To the historically minded it is but the most recent and most sophisticated form of a tradition which can be traced through John Locke to Aristotle; though only in the last fifty years has it been applied to particulars as well as to universals. Since, however, my purpose is systematic rather than historical, I shall state it in my own words, and in the thoroughgoing form in which, at one time, it seemed to be almost a self-evident truth. Thus conceived, the principle consists of a "preamble" which formulates certain of its more immediate presuppositions, and two "articles" which together make up the principle itself.
Preamble: The world consists of atomic facts which are constituted by basic particulars exemplifying basic qualities and standing in basic relations. With few, if any, exceptions, the "things" of common sense are logical constructions and would not appear in a list of the basic constituents of the world. Even more obvious is the distinction between simple (basic) and complex universals (qualities and relations). All particulars and all universals can be described; basic particulars and simple universals can not be analyzed. The principle we are formulating is concerned with basic particulars and simple universals. Whatever may be the proper analysis of thought, there are two ways in which a thought can be about a particular universal: (1) directly; (2) descriptively. These two modes of aboutness correspond to the logical distinction between names and descriptive phrases. Article I: Particulars. A thought can be directly about only such particulars as the thinker is acquainted with at the time of the thought. From the standpoint of the epistemologist, a language is always someone's language at a time, for it contains names, in the strict sense, of such particulars as the user of the language is acquainted with at the time the language is used. Article II: Universals. A thought can be directly about only such (simple) universals as the thinker is acquainted with at the time of the thought. A thinker can be acquainted with only such (simple) universals as have been exemplified by particulars with which he has been (or is then) acquainted. From the standpoint of the epistemologist, a language is always someone's language at a time, for it contains names, in the strict sense, of such (simple) universals only as the user of the language is acquainted with at the time the language is used. I am now convinced that this "principle" is a mistake, and that, together with all "principles" of similar import, it crystallizes and perpetuates a confusion between two distinguishable aspects of the relation between sign-behavior and the world. Of this, more later. Let me first lend a friendly ear to a familiar argument. II
In order to understand the influence of "principles" such as the above on recent philosophical thought, we must see how they fitted in with developments in logical theory. We have taken account of some of these developments in the very formulation of our principle, but it is by virtue of a close association with that specific logical achievement which is Russell's theory of descriptions that the Principle of Acquaintance made its full impact on contemporary epistemology. Our next step, therefore, must be a brief statement of the fundamental contention of this theory. We shall begin by formulating it as a purely logical thesis, ignoring those broader epistemological issues which peer over the logician's shoulder. We shall write as a Russell might have written who, seeking to avoid the logical paradoxes involved in the traditional account of descriptive phrases, was unlike the historical Russell - - completely unaware of the problems of philosophy, and who did not worry about aboutness and the nature of the name relationship. Such a Russell might well have summed up his theory of descriptions as follows.
"Let 'x1' 'x2' 'x3' . . . be the names of basic particulars, and 'f,' 'g,' 'h,' . . . 'Q,' 'R,' 'S,' . . . the names, respectively, of simple qualities and relations. Now it is clear that a name isn't a name unless it names something. But what of descriptive phrases such as "the f"? Do they belong to the category of names? If so, then either "the f" is nonsense or else the f exists. Since descriptive phrases which do not have application are nevertheless meaningful, they clearly are not names. "The f exists" is a meaningful statement which may well be false. Its sense is given by a conjunction of two sentences neither of which contains a definite description. These sentences are:
(1) (Ex) fx.
(2) ~(Ey) fy & y =/= x.Together these sentences say that of all the particulars in the world, at least one and at most one exemplifies f. That these sentences are about all the particulars in the world can be seen by remembering that the first sentence, "(Ex} fx," is the defined2 equivalent of
(1') fx1 V fx2 V fx3 V . . .We thus see that descriptions rest on names, and that we can refer to one particular by means of a description only by virtue of referring to all particulars by means of names. To be sure, it doesn't look as though in using the phrase "the f" we are, even if only indirectly, using the names of all particulars, but from the standpoint of the logician that is exactly what we are doing, and if that is a paradox, I leave it to the philosophers to puzzle out."
III
We are now in a position to lay bare the skeleton of a train of reasoning which takes this brief statement of the theory of descriptions as its point of departure, and leads via the Principle of Acquaintance, to an epistemology characteristic of much recent empiricism, if rarely held in a completely explicit form. It will be convenient to develop this reasoning in the form of a dialogue between our "Russell" -- whom we have supposed innocent of all puzzles relating to our knowledge of the external world -- and a philosophical alter ego who, of empirical bent, is making his first acquaintance with modern logic, Cantabrigian by name. As the dialogue begins, Russell has just finished the above statement of his theory of descriptions.
Cantab. Your theory of descriptions is, indeed, a most valuable contribution to logical theory, and an exemplary specimen of philosophical analysis. But how can you say that the description of a particular (or of a universal, for that matter) is, from the standpoint of a logician, an expression which contain the names of all particulars (or universals) or is the defined equivalent of such an expression? Surely it is clear not only that we can not name all particulars (or universals), our language doesn't even begin to contain such a list of names! Your account of the relation of descriptions to names must be mistaken. Russell. I don't see how it can be mistaken. How can a language have a greater scope than the names it contains? I will grant that we can not utter the names of all particulars in one breath (or in one life-time, for that matter), but the name "x1" is not itself an utterance. The name "red" is no more an individual utterance than redness is an individual red patch. In some sense which, as logician, I can not clarify, a language has an existence or reality over and above the utterances which are its tokens. It is humanly necessary to use tokens of "(Ex) fx" rather than tokens of the definitionally equivalent expression, "fx1 V fx2 V fx3 V . . . ." But the language contains the latter expression in its entirety, even though we are unable to token more than an infinitesimal segment of it. I will admit that there is much here that I can not explicate, but as logician I must stand firm. A language can describe no further than it names. I hasten to add, however, in accordance with the distinctions I have drawn above, that a user of language can describe further than he can utter names. Cantab. You logicians are all alike! Adept at manipulating abstractions, you can not see what is before your eyes! Languages are human facts, and like other human facts, they are clumsy, imperfect, and incomplete. French, German, Bantu . . . these are languages. They are historical facts. They are born, they grow, they die. I agree that the English word "red" is not a single utterance. It is, however, a class of utterances, a class which can only be defined in sociological, historical, psychological -- in short, anthropological -- terms. Can you seriously say that English, or German, or Bantu contains names for all particulars and (simple) universals? Furthermore, a language isn't a mere set of utterance-classes. It finds its existence in occasions on which utterances are used as linguistic utterances. Words, to be sure, have meaning, but it is more accurate to say that words have meaning for users of words. Indeed, it is primarily utterances which have meaning, and to say that a word (as utterance-class) has meaning, is to say that each event which is properly an utterance or token of the word (and not a mere parroting) has meaning. Russell. I must admit that I have been thinking of a language as an ideal system, tidy and complete to an immeasurably greater extent than any historical language. Yet I would have said that my analysis applies to French, German, and Bantu, so that in some sense I must have been talking about sociological facts of the kind you have been mentioning. You are a philosopher. Can you resolve my perplexity? Cantab. To the epistemologist, the key to the understanding of the human cognitive enterprise is the concept of the given. In whatever manner knowledge may be said to extend in scope beyond the given, givenness is the fundamental mode, if not of knowledge itself, then of being within the domain or purview of knowledge. Again, in whatever manner meaning may be said to extend in scope beyond the given, givenness is the fundamental mode, if not of meaning itself, then of being within the domain or purview of meaning. Whatever the heights that can be scaled by piling language habit on language habit, whatever, indeed the internal structure of language habits themselves -- a problem we gladly leave to the psychologist -- this much is clear to the epistemologist who is worth his salt:3 the towering edifice of language rests on the confrontation of sign-event with datum in a field of acquaintance. lt is not enough to say that a name isn't a name unless it names something (has a nominatum). This axiom can itself be understood only in terms of the more fundamental principle that only a sign-event whose referent, be it a universal or a particular, is given to the user of the sign-event can be a name. Indeed, name and nominatum must fall together in one field of direct awareness. To the epistemologist who penetrated below the anthropological standpoint I was presenting a moment ago to shake you out of your naive linguistic realism, language is by its nature datum-centric. It is in the given that he finds the cash value of the vast structure of linguistic habits and events which is studied in so many ways by the various sciences of Man. Russell. I find your argument which moves from language as studied by empirical linguistics to its roots and purchase in the given quite persuasive. Unfortunately, however, much your interpretation of language and meaning may appeal to me as a budding epistemologist, as a logician I am puzzled and unhappy. In my unenlightened days I thought of a language as an ideal structure so correlated with the world that (in addition to its specifically logical devices) it contains a name for each basic particular and for each simple universal ingredient in the world. Now I am being driven to hold that a language can contain names for only what is surely a vanishingly small portion of the ingredients of the world, namely, the elements of someone's field of acquaintance -- including universals as well as particulars -- at a time; the language itself being a set of sign-events within that same field of acquaintance. But if the names a language contains are as limited as this, how can the user of the language refer to items (universals as well as particulars) which fall beyond the scope of this momentary field of acquaintance? Does not your theory of names commit you to a semantic solipsism of the present moment? Indeed, how could one entertain such a meaning as now? for surely this meaning intrinsically involves the contrast meaning then! Cantab. (Startled) But surely it is you who gave us the answer to this question! It was your distinction between names and descriptions which enabled the epistemologist to draw the distinction between knowledge by acquaintance and knowledge by description which disposes of the pseudo-problem into which you have talked yourself! We refer to objects which are not given by means of descriptive phrases, thus: "the object which I shall see a moment later than this." It is by the use of descriptions that we refer to the world of which our data constitute such a vanishingly small part. Russell. No! That won't do at all. We are going around in a circle. The theory of descriptions as I formulated it above rested on the very conception of language of which you have so vigorously tried to disabuse me. You will remember that I claimed that a sentence involving a description is logical shorthand for a set of sentences involving the names of all items of the type to which the descriptum belongs. Thus, we saw that, of the two sentences which explicate the meaning of "the f exists," the first, namely,
(Ex) fxis the defined equivalent of
fx1 V fx2 V fx3 V . . .which is a sentence of which I have tokened only the first three alternants. Indeed, if we suppose that the world consists of a finite set of particulars whose number is t, then the sentence, "The f exists," is equivalent to the following expression:
It surely follows from this analysis that if all nominata are data-here-now, then descriptive phrases can not possibly enable one to refer to what is not a datum-here-now. That would indeed be to pull from a top-hat a rabbit that wasn't there! That is the business of magicians, not logicians, and, alas! even magicians only seem to do it.
fx1 & ~fx2 & ~fx3 & . . . ~fxt or ~fx1 & fx2 & ~fx3 & . . . ~ fxt or ~fx1 & ~fx2 & fx3 & . . . ~fxt .................................................. or ~fx1 & ~fx2 & ~fx3 & . . . fxt Cantab. Can it really be that your theory of descriptions is infected with this naive conception of a language as an ideal structure which covers the world like a metaphysical blanket? But wait! Surely the theory of descriptions says only that "The f exists" is equivalent to "(Ex) fx & (y) fy --> y = x." It is your interpretation of the latter as the defined equivalent of an expression which involves all the names of the language which is at fault. In short, the trouble lies in your account of the existential operator. It may be helpful nonsense to say that to an angel whose world was present in one field of acquaintance, "(Ex) fx" would be logically equivalent to an alternation of atomic sentences whose constituents were names. But for us there can be no such equivalence. In our language, "(Ex) fx" is not an abbreviation. It -- or "(x) fx," whichever one takes as primitive -- is a basic logical expression. More accurately, &(E_) --_" represents a primitive logical operation, for, to put it bluntly, variables are not logical constructions out of names, as you suggest. They are blanks. The logical meaning of general operators lies in the role they play in rules relating to the filling of blanks and in rules relating expressions involving blanks to one another. Russell. This is an interesting suggestion, though I am not quite sure that I grasp what you are driving at. Perhaps if you were to explain how it enables you to dissolve my puzzle as to how we can mean further than we can name. . . . Cantab. Yes, I think that I can exorcise your puzzle now, though I must confess that before I thought of this new approach to the existential operator, you had me worried. I now see that we must distinguish between two modes of empirical meaning. (1) The mode characteristic of names. Here there is a direct tie between the linguistic and the non-linguistic. Both names (sign-events) and nominata fall within one field of empirical givenness. (2) A mode of meaning built on the former, but not reducible to it. This mode of meaning consists in the employment of general operators and variables. The empirical significance of expressions belonging at this level of language lies in the rules relating it to the first mode of meaning. Thus, we can go from "(x) fx" to "fa" where "a" is a name and thus has empirical meaning of the first mode. On the other hand, the unique contribution of this second mode of meaning lies in the fact that the use of expressions on this level constitutes, subject to an anchoring in the first mode of meaning, "a reference to the world as including but extending beyond the given." Empirical meaning proper is a matter of givenness. Reference to the world is a purely formal matter of the logical structure of our language. It has empirical significance or "cash value" only by virtue of the rules which tie general operators and variables up with names, and hence with empirical meaning proper. We leave this unhistorical dialogue uncompleted. By isolating two of the many souls which vie in Lord Russell's breast, and causing them to argue, we have been able to reconstruct the dialectic which has manoeuvered many logically-minded empiricists, though never -- completely -- Lord Russell, himself, into the position which has come to be called "syntactical positivism."4 We have not given the "Russell" of our dialogue a crack at this position, as formulated by Cantabrigian, for the obvious reason that the historical Russell has been far more adept at getting people into this quicksand than at getting himself out of it. But what shall we say in reply to Cantabrigian? Here we must be brief, for our main purpose has been to present the above dialectic, rather than to dissolve it, a task which would take far more space, not to say insight, than we have at our disposal. Is it enough to point out that the psychology of human cognition is most certainly not built on the concept of a meaning relation which holds between sign-events and other items, particulars or universals, in one field of acquaintance? No, for the mistake which Cantabrian makes is a more basic one which could be allied with sound psychological doctrine. It is that of taking the "designation relation" of semantic theory to be a reconstruction of being present to an experience. This mistake is the same whether combined with an adequate psychology of "experience" or associated with the pseudo-psychology of the "given." Semantic designation reconstructs neither "phenomenal givenness" nor "behavioral response to an environmental stimulus." In so far as semantic designation is a reconstruction of an aspect of man's adjustment to his environment by means of sign-behavior, it concerns rather the relation of sign habits to features of the environment in abstraction from particular acts of experiencing these features. It is the pragmatic concept of verification which reconstructs the meeting of language and world in a cognitive situation. The student of pure semantics will recall how minor a role the type-token distinction plays in treatises on the subject. It is mentioned in the early pages, but does not enter into the definitions and theorems themselves. On the other hand, pragmatics, from which, after all, semantics is an abstraction, is concerned with the contact of a linguistic structure with the world, and this contact essentially involves linguistic tokens or sign-events. It is in pragmatics that we find the theory of demonstratives, words such as "this." "here," "now." Words of this type involve an intrinsic relation to a particular cognitive situation. Here is the crux of the matter. Cantabrigian is confusing between names and demonstratives. Names as designators are in the object language. They can be tokened many times with each token having the same meaning. Demonstratives are in a meta-language, and not only are essentially sign-events, having a different sense on each occasion of their use, but also refer to the confrontation of an object-language sign-event with an experienced situation. It is demonstratives and not names which are limited in their reference to items belonging to the same experiential situation in which they occur. We are thus in a position to distinguish the question, "How can a language mean further than it names?" from the question which has so often been confused with it, "How can a language mean further than a this-here-now?" This, in turn, enables us to call attention to an important truth in the Cantabrigian position. It is indeed the case that our use of variables and general operators or their counterparts in common usage, is not a mere shorthand for the use of the names our language contains. It is a sound insight which leads him to suggest that the reach of our language beyond the items for which it contains names consists in the use of variables and general operators. These are undoubtedly essential features of the structure of linguistic habits by means of which we adjust to our environment. But it is still more important for the philosopher to see that these assertions belong to the empirical anthropology of sign-behavior. Is there, then, no truth in the contention made by our unhistorical "Russell" that a language can not mean further than it names? This contention, if we lay aside semantic complications, is exemplified by the assertion that
(Ex) fx = Df fx1 V fx2 V fx2 V . . . I want to suggest that in the logical frame of reference, where it obviously belongs, this is a perfectly legitimate statement. We must not confuse between an "understanding of the logician's utterance" which consists in an empirical description and explanation of the utterance in terms of the linguistic habits and limitations of the logician as a source of utterances, and that "understanding of the logician's utterance" which consists in a re-enactment of the utterance. To understand in the latter sense is to do logic. What, then, do logicians do? In what does their vision of the infinite horizon of the possible consist? What is the relation of their activity to the empirical psychology of sign-behavior? In what sense do they talk about this behavior? These are exciting questions. I am happy that I have not committed myself to answer them on this occasion.
Notes
1 Published in the Monist, 1915-1919.
2 This account of the existential operator will come under scrutiny at a later stage of our argument. Even the reader who is suspicious of it must grant that it has played an important role in the development of modern logical theory.
3 I hasten to remind the reader that Cantabrigian speaks for himself, though I find his remarks interesting.
4 The term "syntactical positivism" does not have the meaning of "positivism in its syntactical phase" as opposed to "semantic phase of positivism." Syntactical positivism is a position which is very much alive, and is defended by several positivists who have found a place in their theories for the pure semantics of Tarski and Carnap.
Transcribed into hypertext by Andrew Chrucky, Sept. 19, 1997.