Logic, Ontology, and Language, Essays on Truth and Reality 3884050303

Citation preview

Herbert Hochberg Logic, Ontology, and Language

Analytica Investigations in Logic, Ontology and the Philosophy of Language Editors: Ignacio Angelelli · Austin (Texas/USA) Joseph M. Bochenski · Fribourg (CH) Christian Thiel · Erlangen Editor-in-chief: Hans Burkhardt · Erlangen

Philosophia Verlag Miinchen Wien

Herbert Hochberg

Logic, Ontology, and Language Essays on Truth and Reality

Philosophia Verlag Miinchen Wien

CIP-Kurztitelaufnahme der Deutschen Bibliothek Hochberg, Herbert:

Logic, Ontology, and Language. Essays on Truth and Reality/ Herbert Hochberg. Miinchen •Wien: Philosophia Verlag, 1984. ( Analytica) ISBN 3-88405-030-3

ISBN 3-88405-030-3 © 1984 by Philosophia Verlag GmbH., Miinchen All rights reserved. No part of this book may be reproduced in any manner, by print, photoprint, microfilm. or any other means without written permission except in the case of short quotations in the context of reviews. Typesetting: FotoSatz Pfeifer, Germering Manufactured by Pera Druck, Matthias KG., Grafelfing Printed in Germany 1984

For Jon-Erik

Table of Contents

Preface Introduction: Ontological Analysis and the Linguistic Turn Frege on Concepts as Functions: � Fundamental Ambiguity Russell's Attack on Frege's Theory of Meaning Professor Quine, Pegasus, and Dr. Cartwright On Pegasizing

9 11 48 60 86

Nominalism, General Terms, and Predication

101 105 133

Nominalism, Platonism and Being True of

150

Mapping, Meaning, and Metaphysics Sellars and Goodman on Predicates, Properties, and Truth

157 185

Russell's Proof of Realism Reproved

196

Logical Form, Existence, and Relational Predication

204

Elementarism, Independence, and Ontology

231

Ontology and Acquaintance

238

Things and Descriptions Universals, Particulars, and Predication

244

Strawson and Russell on Reference and Description

263 279

Facts and Truth Negation and Generality

296

Arithmetic and Propositional Form in Wittgenstein's Tractatus

313

Russell's Reduction of Arithmetic to Logic

321

Properties, Abstracts, and the Axiom of Infinity

339

Of Mind and Myth

353

Physicalism, Behaviorism, and Phenomena Belief and Intention

374 388

Bibliography

419

Index of Names

445

Author's Note

447

Preface

Though it is an essay collection, the book, as a whole, is concerned with ontological questions that focus on various aspects of the problems of universals, reference, predication, intentionality, and mental entities. As a group the essays offer a defense of realism against nominalism, of facts and a correspondence theory of truth against "pragmatic-idealism", and of the existence of mental entities as against physicalistic reductionism. The positions set forth derive much from the writings of Russell and Moore at the turn of the century and the work of Gustav Bergmann, though some of the latter's basic views are the targets of critical analyses. Together, the essays attempt to set forth an adequate ontological basis for the resolution of a number of fundamental problems within the traditon of "analytic" philosophy. While many issues are dealt with, the basic problem that is the focus of the book is the question of the connection between thought and language, on the one hand, and what they are "about", on the other hand. Thus the questions "What gives a thought the content it has?" and "What is the role of language in thought?" are central. Descartes set a question that gave rise to a sceptical concern: "How do we know that thoughts correspond to reality?" But, there is an even more basic question, which the Cartesian query presupposes: "How do we know what the content of a thought is?" In dealing with the connection between thought and the objects of thought, the essays lead to a proposed resolution of that question. The bibliography includes works which, while not cited in the essays, are relevant to the issues discussed.

Introduction: Ontological Analysis and the Linguistic Turn

Gustav Bergmann aptly called the turh philosophy took· under the influence of the Logical Atomism of Russell, Moore and Wittgenstein and the Logical Positivism of Schlick, Carnap and the Vienna Circle 'the linguistic turn'. In the writings of Quine, Goodman, Sellars and Bergmann that turn was to reintroduce some classical and fundamental questions of ontology in a linguistic guise, as questions about the syntax and semantics of clarified formal 'languages'. While Quine, Sellars and Goodman were to revive classical nominalism in a contemporary 'analytic' form, Bergmann would be, for a time, virtually alone among the positivistically oriented philosophers in retaining the realism of Russell and Moore within the analytic tradition. The classical issue, in its revised form, was raised as a question about the meaning and role of predicates in perspicuous or clarified or 'ideal' languages. In part, the revival of the nominalism-realism controversy reflected a more general issue that divided "ideal language' philosophers: an issue about the nature and role of formal, and purportedly perspicuous, language schemata in the analysis and resolution of philosophical problems. (This general issue is discussed in the essays 'Mapping, Meaning, and Metaphysics', 'Of Mind and Myth', and 'Professor Quine, Pegasus, and Dr. Cartwright'.) Ironically the basic theme of the nominalist's treatment of predicates and of Russell's earlier realistic treatment, from which Bergmann's is derived, stemmed from Frege 's analysis of predication and the difference between subject and predicate terms that Frege focused on. Frege had ingeniously attempted to resolve a number of fundamental philosophical problems in terms of a few basic theses. Like F. H. Bradley, Frege saw a problem in the realist's analysis of predicative judgments in terms of a predicative relation or connection (exemplification, participation, standing in) between one or more particulars and a property or relation. He believed, as did Bradley, that such an analysis faced a vicious infinite regress. For, if one introduced a relation, say exemplification, to correspond to the predicative copula and held that a particular, a, exemplifying a property, F, was the truth condition for the statement or thought that a is F, then one would require, as the truth condition for 'a exemplifies F', a further three-term relation connecting a, the relation of exemplification, and the property F-- and so on ad infinitum. (Frege's concern with this problem is discussed in 'Frege on

11

Concepts as Functions'. Other aspects of the problem are discussed in 'Ontology and Acquaintance', 'Universals, Particulars, and Predication', and 'Facts and Truth'. Sellars' revival of the issue as an argument against realism is analyzed in 'Logical Form, Existence, and Relational Predication'.) An infinite regress of truth conditions means that one could not even set down the truth conditions for a simple subject­ predicate sentence like'a is F'. (In'Russell's Proof of Realism Reproved' I propose an argument, derived from Russell's classic but invalid refutation of nominalism, that nominalism, understood in a certain way, falls victim to such an infinite regress of truth conditions.) Bradley had held that this purported'paradox' of predication, along with a similar argument against a correspondence 'definition' of truth, forced philosophers to recognize a holistic Absolute as the one subject of and truth ground for all judgments. Frege, who also employed a similar argument against any purported analysis of truth, took the'paradox' to show that concepts or functions, replacing the realist's universal properties, which were abstract'objects', could not be taken to be objects like particulars (or classes, or senses, or propositions, or truth values). Rather, the solution of the problem lay in the recognition of the incomplete nature of F, as a concept or function, as opposed to a complete or'saturated' object like a.Objects, then, would be said to 'fall under' concepts, but this was, purportedly, simply another way of expressing the matter and did not imply the recognition of falling under as a connection forming a complex (a fact) from objects and concepts. Recognizing such an incomplete constituent of propositions or thoughts, expressed by sentences, one need not also recognize any connection between concepts and other constituents of propositions. The complex thought or proposition need only contain an incomplete constituent, a concept, to provide the connection that held the proposition together. The link was supplied by the concept F itself.Such an incomplete entity was thus necessary as a constituent of every proposition or thought to provide the' connecting link'. Frege thus blocked the apparent infinite regress that has come to be known as 'Bradley's paradox' by taking concepts to be incomplete entities that did not require a connecting relation to join them to other entities (senses) to form propositions. At times Russell was to follow Frege's solution, and Wittgenstein's well known metaphor in the Tractatus of the 'links in a chain' may well indicate that Frege's pattern influenced his conception of a fact. At other times, Russell took exemplification to be a basic asymmetrical connection that required no further connection, a theme that was to be picked up and advocated by Strawson and Bergmann. For Frege, names and definite descriptions referred to objects (when they referred) and expressed senses, which were also objects. Sentences 12

expressed propositions. In the case of 'a is F' the sense of the name 'a' combined with the incomplete concept F to form the proposition , expressed by· a is F'. In the case of the sentence'a is in R to b , the senses of the names •a' and 'b' combined with the incomplete relational concept to form the proposition expressed. On Frege's pattern signs thus stood in two relations to things. Names expressed senses and referred to objects, while sentences expressed propositions and referred to truth values, which were also objects. It is interesting to note that just as, for Bradley, all true judgments were ascriptions or modifications of one subject so, for Frege, all true propositions denote or refer to one object: the True. In fact, just as Bradley explicitly held that all judgments had one subject, Reality, Frege once held that all judgments had one predicate, is true. The connection between Bradley's holistic Reality and Frege's the True is more than coincidental. In both it stems from a rejection of facts and a correspondence theory of truth. (In 'Facts and Truth' I discuss a recent attack on the correspondence theory by Davidson that makes use of a Fregean pattern: that all true sentences denote the same thing. Many problems of interpretation and of substance have arisen regarding the connection of senses to objects and of predicates to concepts, as well as problems about the various roles concepts play for Frege. These are addressed in 'Frege on Concepts as Functions' and 'Russell's Attack on Frege's Theory of Meaning'. ) Frege 's pattern helped resolve a second problem that preoccupied him in dealing with mathematical statements: the need to specify the difference between significant identity statements, like 'a is the ', '2 is the only even prime number', and 'the morning star is the evening star', and trivial identity statements, like 'a is a'. It would seem that the type of truth condition is the same in all these cases, that signs to the left and right of an identity sign refer to the same object. The fact that in some cases the signs are tokens of the same type, while in other cases they are not, would not seem to be relevant to the truth conditions for sentences. To hold that it is relevant leads to the problem of having the signs somehow enter into the truth conditions or forces one to hold that the difference between significant and trivial identity statements is not a matter of what is said, but of how it is said. Thus, neither the referent of the sign(s) nor the fact that something is identical with something furnishes the truth condition. Frege appealed to the relation between a sign and its sense (or meaning) to resolve the problem. In the case of trivial true identity statements, only one meaning or sense is involved, since the sense of the two tokens of '2' in '2=2' is the same. In the case of non-trivial, but true, identity statements, the senses of the subject terms are different, while the referent is the same. Frege's pattern also enabled him to resolve the perennial question 13

about the difference between relational judgments and judgments ascribing a monadic property to an object. He took relations and properties to be concepts (or functions) that differed in the number of complete terms they combined with to form a proposition. Bradley had held that relational propositions must be construed in terms of relational properties, like being in R to b. Thus, in place of a relation R being jointly exemplified by two terms, he took the monadic relational property to be attributed to one term. This way of construing relations provided an argument for his holistic idealism. Arguing that particulars, as simple or 'bare' bearers of properties did not exist, he construed objects in the style of Berkeley, as bundles of qualities. By including monadic relational properties as qualities of particulars, he concluded that, as every particular was related to every other particular, it followed that every particular entered into the complete concept of every other particular. Hence, no particular was independent of any other particular in two ways. First, the complete or adequate concept of any particular would involve every other particular. Second, as particulars were bundles of universal concepts, the existence of any particular involved the existence of every other particular. For Bradley, this showed, as it had in its way for Spinoza, that there was only one real or independent particular, the Absolute or Reality as a whole. Frege's account of relations, and Russell's consequent recognition of relations and of the different logical forms of propositions and facts, provided crucial support for fundamental themes in the realism of Moore and Russell and the atomism of Russell and Wittgenstein. One theme was the insistence on the independence of atomic facts from each other. A related thesis was the claim that particulars are unanalyzable, and hence independent, constituents of atomic facts along with simple and unanalyzable universal properties. (Aspects of these themes are considered in 'Things and Descriptions', 'Universals, Particulars, and Predication', and 'Elementarism, Independence, and Ontology'.) We can see how such theses came to characterize the realistic revolt of Moore and Russell against the idealism of McTaggart and Bradley that dominated British philosophy just prior to the turn of the century. Moore initiated the attack on idealism in an early paper that still showed the influence of Bradley and McTaggart. In The Nature of Judgment' he followed Bradley's view that an 'ordinary' particular must be understood to be a combination or bundle of universal concepts affirmed as existing. Bradley, in turn, had followed Berkeley's classic theme and maintained that to think of an object was to think in terms of properties of it. One could not conceive of or have an idea of an object simply as a bare or pure particular--Locke 's "that I know not what" that served as a 'substratum' for properties. When we judged of an object, say a, that it had a property,

14

say F, we judged that F cohered with or combined with other properties (concepts )--those in terms of which we conceived of a in that judgment. Let "P' and "Q' stand for such concepts and assume, for simplicity, that there are only two, so that being P and Q is the concept had when we think of the object a. In judging that a is F we are really judging, on this view, that being P and Q and F exist together or in reality , even though we may express that judgment by using the sentence 'a is F'. In proposing this analysis, Bradley set down a theme that was to become central in Russell's theory of descriptions: the idea that an apparent subject-predi­ cate judgment (proposition) is really (or logically) an existential judgment (proposition). And this theme, of course, was to be crucial in the development of analytical philosophy with its distinction between the apparent grammatical form and the real logical form of a proposition or thought. Eventually the basic theme would emerge in contemporary linguistics in talk about 'deep' and ·surface' structure. The construal of nominally subject-predicate judgments as existential judgments was also to be crucial in the development of Moore's and Russell's realism, regarding physical objects, within the empiricist tradition. For it would give them a way of indirectly referring to physical objects, by definite descriptions, while acknowledging that one could only refer directly to objects of direct acquaintance, which they held, in their early writings, to be phenomenal objects, like sensa. Bradley's way of putting his claim was both dramatic and confused. Since both being P and Q and F were properties, or concepts, or 'predicates', he held that the content of judgment was a property, being P and Q and F, which we ascribed , in judging, to Reality as the subject. Reality , as a whole, thus became the one subject of all subject-predicate judgments. In so putting matters, Bradley thought he was escaping from his paradox of predication, since he replaced a purported connection in thought (predication) with an activity (ascription or judgment). Just as a concept of a pure particular was unintelligible, because empty, so no concept of predication was intelligible, since such a concept was paradoxical. One escaped the problem, Bradley thought, by recognizing complex concepts as the contents of judgment. Thus, the content of a judgment could not be that a is F, since this would involve the problems about the particular a and the predicative connection. The content of a judgment could be the complex concept being P and Q and F, since that involved us in neither problem. In judgment we then attributed what was in thought, the complex concept, to what was not in thought, the subject, Reality .Implicit in Bradley' s line of argument is the idea that we construe judgments expressed by sentences like ' a is F' as existential judgments Hence, Bradley has the seeds of two ideas that will become very influential in the development of analytic philosophy: the distinction

15

between the real and apparent form of a judgment and the construal of subject-predicate sentences as existential sentences. This is not to say that Bradley explicitly took '(3x)(Px & Ox & Fx)' to be the perspicuous form of ' a is F'. Doing that would force him to recognize that the problematic subject-predicate distinction was still involved in the content of the judgment.It would also force him to acknowledge that a concept of existence, represented by the existential quantifier, was a part of the content.This, too, he found problematic. He sought to avoid both problems by holding that the complex concept being P and O and F was ascribed to Reality . The act of ascribing , or judging, was not part of the content of the judgment, and could unproblematically replace the predicative connection, while Reality , which was also not part of the content of the judgment, replaced a problematic concept of existence. Nevertheless, the basic idea of transforming subject-predicate judgments into existential judgments is involved in Bradley's analysis. From the claim that subject-predicate judgments are to be analyzed as ascriptions of predicates to reality, it is a short jump to the related theme that objects are to be analyzed as collections of properties.In combining these themes, Bradley does not clearly distinguish a particular object, as a complex of properties, from a particular object as a complex property. This failure helps him argue that there is only one subject or particular, since other particulars, as complex concepts (or complexes of concepts), are attributes of it .To avoid the criticism that, on his view, the Absolute itself was nothing but a bare substratum or particular, Bradley held that it, too, was a complex of all concepts coherently attributed to it. The Absolute , or Reality , was as it were the ultimate complex concept. Moore's early attack on idealism centered on providing an alternative analysis of judgment and truth. He held that universal concepts (properties, sometimes called "predicates') were the ultimate constituents of the world.Among simple concepts Moore included such properties as red , now , this , and existence . Propositions , distinguished from sentences, were objective entities, like Frege's thoughts, and were composed of such concepts. In fact, all existents were held to be composites of concepts. 1 An object, say a , was a composite of concepts that included existence. We might represent the object a by an expression like (1) C(F, G, .. . , existence), where 'C' stands for a relation obtaining among the constituent concepts to form the object. There is also, on Moore's analysis, a concept of a . Such a complex concept would be composed of all the concepts in a , 16

except for the concept existence , combined by a relation, say C, and represented by:

(2)

C'(F, G , . . . ) .

While (2) represents a concept that is of an obj ect and that is a complex concept , it does not represent a concept that is a universal property, like F, red , and existence. A proposit ion, li ke that expressed by the sentence ·a exists' , is, for Moore, also a combination of concepts; the concepts that are menti oned in (2) and the concept of existence. But what then distingui shes the obj ect a from the existential proposi tion expressed by 'a , exists ? Moore' s remarkable answer is that nothing does and he identifies obj ects with true existential propositions declari ng that: Even the description of an existent as a proposition (a true existential proposition) seems to lose its strangeness . when it is remembered that a proposition is here to be unde rstood . not as anything subj ective --an assertion or affirm ation of somethi ng-­ but as the combination of concepts which is affirmed . 2

Such existential proposit ions ( or complex objects) are the prototypes of the facts Moore and Russell were shortly to defend. 3 But to arrive at his later notion of a fact, Moore first had to alter his analysis of particulars and their relation to universal properties. On his early view, which shows the obvious influence of Bradley, a proposition expressed by a sentence like · a is F' is also a complex of universal concepts, but such a proposition, unlike existential propositions, is not identified with an obj ect. Rather , such a proposition, whether true or false, is an entity that is about its obj ect since it contains a concept that is a concept ofthe obj ect. But , what is it for a concept to be of an obj ect? I n the case of a concept like that represented by (2), there is an obvious answer. Except for existence, the obj ect and the concept of it are composed of the same constit uent universal properties. If propositions about an object are held to contain the complete concept of an object the problem of picking out an object is thus resolved. But , Moore, like Bradley, holds that when we j udge about an object we need n ot have such a complete concept 'in min d'. Thus, a question arises about how a partial concept of an object picks out that object. Russell' s theory of descriptions will provide one kin d of answer, as Frege' s senses provide another. (Problems with Frege' s pattern and Russell's attempt to resolve such problems are di scussed in ' Russell' s Attack on Frege's Theory of M eaning' . ) Moore' s concepts are very si milar to Frege's senses, just as Moore's account of thought and judgment is li ke Frege's. A particular act of thought or j udgment has the content it does owing to its being related to a proposit ion, which is a 17

complex of concepts for Moore, as it was a compound of a concept and a (or several) sense(s), for Frege :Moore's account of propositional truth marked his break with Bradley. Bradley, as we noted, had held that in judgment we ascribe the content of thought to Reality . This led him to reject a correspondence account of truth, since we could not know any correspondent that would make a proposition true. We could not know any such correspondent since what can be known is what can be a content of thought. But only ideational contents, complex concepts, can be contents of thought. In short and in other terms, we can only know propositions and not their purported correspondents or grounds of truth. Hence, the mind cannot compare a proposition to, say, a fact and determine the former to be true because it corresponds to the latter. All one can do is compare propositions. Consequently , truth must be understood in terms of the coherence of propositional contents. Coherence is thus not only the criterion for ascribing a content to reality : the very meaning of 'is true' is to be understood in terms of coherence. Moore, while adopting Bradley's analysis of particulars, seeks to reject Bradley's idealism by rejecting a coherence theory of truth. Yet, he is disturbed by the idealist's attack on the correspondence theory, which, as we shall soon note, closely parallels Frege's attack on the correspondence theory. It is interesting that both Frege's and Bradley"s attacks on the correspondence theory are structurally like their attacks on a purported relation connecting subjects to properties in propositions. In fact, in an unpublished 1 906 paper entitled 'On Truth', Russell, too, was bothered by a purported infinite regress involved in the correspondence theory of truth. In his early paper, Moore avoids the problem by holding that thoughts (propositions) are not true in virtue of corresponding to facts but in that true propositions are facts. We have already seen that in the case of true existential propositions Moore identified them with the objects they were 'about'. In such cases he can be thought to identify three things: an object, the fact that the object exists, the true existential proposition. In the case of a subject-predicate proposition expressed by "a is F' , what grounds the truth of the proposition is the relation between the constituent universal concepts in the proposition . Thus, he writes : A proposition is constituted by any numher of concepts . together with a specific relation between them ; and according to the nature of this re lation the proposition may be either true or false . What kind of relation . . . cannot be further defined . but must be immediately recognize d . 4

Thus , the difference between true and false propositions is an internal one in the sense that the combining relation differs in the two cases. But this is a way of treating propositions as facts and recognizing two kinds of 18

facts: existent facts and me re possibilitie s or possible facts. This theme would recur in Moore ' s late r work Some Main Problems of Philosophy , as well as in the writings of Russell an d Wittgenstein. (Aspects of these issues about truth are discusse d in ·Facts and Truth' "Elementarism ' Independence, and Ontology' . 'Mapping, Meaning, and Metaphysics' ,' and ·Logical Form, Existence, and Relational Pre dication' . ) Taking propositions to be their own grounds of truth enables Moore to rej ect the idealist' s coherence theory without accepting a correspondence account. Yet, like Fre ge's taking true propositions to be propositions that denote the True , there is a clear sense in which either no account is given or the account is extre mely Pl atonistic. It is the forme r, if one takes the view to be a way of stating that the concept of truth is empty in that to say that a proposition is true is to do no more than assert the proposition. It is the latter if one takes the condition of truth as a relation obtaining among universal concepts or between a proposition an d an obj ect, the True . In Moore ·s case , it is clear that he was a Platonist in his recognition of combinations of universals as truth conditions. For Frege, and certainly for contemporary philosophers like Sellars, Dummet, and Davidson, the former interpretation is viable, since Frege doe s not recognize a fact consisting of a relation of denotation obtaining between a proposition and the True . Ironically, because of this, Frege's account faces an infinite series of the kind he attribute d to a correspondence theory. (We shall examine his argument shortly. ) Where · Fa ' is true, the proposition expressed by • Fa ' denotes the True . B ut, on Frege ' s view, this means that the proposition expre ssed by The proposition that -F a denotes the True denotes the True , an d so on. For Frege' s view, the series forms a problematic regress, since the explanation of what it is for a proposition to be true is given, in one sense, by the statement that the proposition denotes the True . In another sense, no explanation is supposedly offered since the True is a basic object. As Plato had rej ected scepticism by grounding the truths of ethics, mathematics, and natural law in relations among ' forms', Moore rej ected idealism by recognizing obj ective propositional entities. In so doing, he paid a price that was to later plague him, Russell, and Wittgenstein, since, on his pattern, one also recognizes 'objective falsehoods', as Russell was to call them. Just as unive rsal concepts combined in one relation to form a true proposition, so they combined in another basic relation to form false propositions. The two kinds of propositions, as e ntities, of course mirror Frege' s recognition of the False, as well as the True, as an obj ect den oted by propositions. In 1 901 Moore introduced a further and historically important argument against idealism. A s we noted, one ide alistic theme common to 19

Bradley and Berkeley is the construal of an object as a complex of qualities. Berkeley's attack on the idea of a material substratum and Bradley's similar criticism of ' bare particulars' are initial steps in the analysis of objects as composites of qualities. The next step is the identification of qualities with concepts or 'sensations' or 'ideas'. While Berkeley, in nominal istic fashion, had problems with qualities as common characteristics, Bradley's qualities are clearly universals. Moore seeks to refute the idealistic analysis by challenging the first step in it: the analysis of an object as a composite of universal qualities. He does this by arguing that such an analysis denies the existence of ·numerical , difference' as opposed to 'conceptual difference and that such a denial is mistaken. I t is mistaken since numerical difference must be recognized and yet cannot be analyzed in terms of conceptual difference, or difference in qual ity. Moore, thus, in 1 901 , argues against what some call the Russell-Leibniz analysis of identity, whereby identity of particulars is taken as having all abstract qualities in common. In his important 1 91 1 paper, 'On the Relations of Universals and Particulars', Russell adopted, as he acknowledged, the arguments of Moore. The insistence on the unanalyzabil ity of particulars (as collections of universals) and on the existence of universals fitted with Moore's and Russell's development of a correspondence theory of truth that recognized facts, as combinations of particulars and universals, to be the truth conditions for thoughts and statements. Russell' s analysis of relations led to the recognition of the various kinds of atomic facts, differing in the kinds of universals and the number of particul ars that were constituents. But the realism of Moore and Russell, and the Logical Atomism it spawned , faced a number of problems regarding the n ature and variety of facts, the order in facts, and the logical form of facts. (These issues are taken up in " N egation and General ity', ' Logical Form, Existence, and Relational Predication' , 'Belief and Intention', and "Elementarism, I ndependence, and Ontology' . ) Russell was eventual ly to abandon his views on particulars. I n one of those intriguin g ironies that spice the history of philosophy, in his 1 940 book An Inquiry into Meaning and Truth , Russel l reverted to the earl y view of Moore and Bradley, that particulars were analyzable in terms of universal s. He did this for one of the very reasons that motivated Bradley: the view that bare particulars , or pure subjects of qualities, were not objects of acq uaintance and, hence, unintelligible on empiricist principl es. For Russell, a principle of acquaintance, derived from the classical empiricists' discussions of simple ideas, had always been a guide in ontological analysis. Thus, he hel d that the ultimate constituents of the facts which we judged about and of the propositions we understood must be objects of acquaintance. Pure particulars, not being such objects.

20

could not be constituents of ultimate or atomic facts. Bergmann, who adopted much from Russell and Moore in his gradual movement away from the logical positivism of Carnap, was eventually to hold that the existence of basic or bare particulars was the key to the failure of idealism. Holding that realism about universals required the acknowledgment of basic particulars to exemplify universals, Bergmann held that particulars were fundamental to the rej ection of both nominalism and idealism. In part , his position stemmed from the desire to avoid Platonism, the view that uninstantiated universals exist. Yet, guided by a Russellian principle of acquaintance , Bergmann, followed by a number of his former students , including Grossmann, Allaire, Wilson, and Addis, rej ected a standard theme of classical empiricism and claimed to be acquainted with basic or bare particulars. In the attempt to make their claim plausible, Bergmann and other members of the Iowa group were forced to make a variety of desperate moves. Thus , Bergmann identified the bare particular of a visual object with the area of the object, while Grossmann identified the substratum of an ordinary particular with the particular and Allaire argued that all that was meant by the claim of acquaintance with bare particulars was that ordinary obj ects were obj ects of experience. (Several facets c,f these issues are explored in "Elementarism, Independence, and Ontology' , where a principle of exemplification is introduced, " Universals, Particulars, and Predication', 'Onto logy and Acquaintance', and 'Things and Descriptions'.) The correspondence theory of truth that Moore and Russell advocated naturally led to 1 questions about the relation between thought and language, on the one hand, and facts, on the other. These , in turn, focused attention on analyses of j udgment that would offer realistic alternatives to the idealistic analysis of Bradley. Frege's views on meaning, reference, and predication also provided a basis for an analysis of intentional contexts and of the relation between thoughts and what thoughts were about. Believing that and thinking that were construed as relations between persons and propositions. Propositions, in turn, denoted truth values, the True and the False , just as senses denoted other objects. A person's thought was about an object since the proposition involved contained a constituent that denoted the obj ect. Thus , where 'TH ' stands for a relation of thinking that , 'p' for a person , and 'Fa' is a sentence expressing the proposition, (S)

TH(p, Fa)

represents the situation obtaining when p thinks that - Fa . On such a view, one does not preserve the truth value of (S) by substituting a sentence for "Fa' , in (S), simply because the substituted sentence is equivalent in truth

21

value to 'Fa' . That is, one cannot guarantee the preservation of a truth value if one substitutes a sentence that has the same denotation as ' Fa' , the True or the False . For, in (S), the sentence 'Fa' does not denote the truth value it normally denotes (in truth functional contexts) ; rather, it denotes the proposition that it normally expresses. This referential shift allows a Fregean to hold that one may substitute one term for another i n a sentence whenever they denote the same 'thing' and preserve the truth value. In (S) we would have to replace 'Fa' by another sentence or expression that denotes the proposition that - Fa. A consequence of this Fregean move is that the name ' a ' also shifts reference in the sentence (S) and denotes its sense. By appealing to referential shifts in certain contexts, Frege attempts to preserve the theme that one may replace expressions by expressions with the same denotation. (Some problematic consequences of Frege 's analysis are considered in 'Frege on Concepts as Functions' and ' Russell's Attack on Frege's Theory of Meaning'. ) One reason Frege held to the view that propositions with truth values denoted the True or the False was his belief that it was not possible to provide a viable analysis or theory of truth , since he thought that all such theories, and in particular the correspondence theory , were paradoxical. The paradox he raised is very similar to one that Bradley raised, and which caused Russel l to doubt, for a time , the viabil ity of a correspondence theory. It is also reminiscent of Bradley's and Frege's worries about a predicative connection. Frege's argument about truth is concise. Can it not be laid down that truth exists when there is correspondence in a certain respect? But in which? For what would we then have to do to decide whether something were true? We should have to inquire whether it were true that an idea and a reality , perhaps , corresponded in the laid-down respect . And then we should be confronted by a question of the same kind and the game could begin again . So the attempt to explain truth as correspondence collapses . And every other attempt to define truth collapses too . For in a definition certain characteristics would have to be stated . And in application to any particular case the question would always arise whether it were true that the characteristics were present . So one goes round in a circle . Consequently , it is probable that the content of the word 'true ' is unique and indefinable . 5

There are several arguments implicit in the passage. One is that no definition of 'is true' is feasible since such a defi nition will be of the form: . . . is true - . . . is X, for some condition X. But, if this is so, then to hold that something is true is to hold that it is true that something has X. Hence, in any given case, we

22

would be involved in a vicious circle . Even the kind of pattern associated with Tarski's Convention T will not help. For, take the case,

"Fa' is true = Fa . There is no characteristic X, as above. But, for Frege, we would have to know what "is true· means in order to know that the sentence

"Fa' is true = Fa is true. Thus, we do not explain the meaning of 'is true·, since we must know what it means in order to understand the purported definition and know that it is true . Frege has a point , but not a paradox. If you are going to give a definition of "is true' or a theory of truth, you have to know what it is to be true in order to know whether the theory is true . It does not follow from this that we cannot give such theories or definitions. What we must do is make certain distinctions. We all know what it is for an object to have a certain characteristic, say a color . This is not to say that we know what it is in the sense of having an adequate account of predication. Of course we must know, in one sense, what predication is in order to offer alternative philosophical analyses of it . But this neither prevents our proposing analyses nor produces a paradox. Likewise, in the case of truth , we know, in ordinary contexts, what it is for a statement to be true, rather than false. And, we also have ways of evaluating philosophical theories and arguments for such theories. But this does not mean that we must presuppose a theory in order to propose a theory. For, what we make use of, or presuppose, is an "ordinary' or pre-analytic notion of truth, when we purport to offer a "true' theory of truth. For that matter, it is not clear that one need hold that all true statements are so on the same kind of ground. Thus, a correspondence theorist need not hold that statements incorporating or reflecting philosophical analyses need correspond to facts . A second way of tak ing the argument is as claiming that the correspondence theory introduces a truth predicate, 'T', in sentences like: (T)

T'Fa' = d/3f) ( ' Fa' corresponds to f),

with the quantifier ranging over 'facts'. One may then argue that (T) itself will be true if and only if there is a fact to which it corresponds. But, then (T) will be true if and only if: (T')

(3f) ( (T) corresponds to f) 23

is true. As the process goes on without end, (T) will be true if and only if there are an infinite number of facts. Thus, we cannot set down all the truth conditions for the truth of (T) and, hence, for 'Fa' . This argument assumes that 'Fa' , (T), and (T') have different truth conditions ; for if the fact that is the ground of truth for 'Fa' is also such a ground for (T) and (T'), there is not an unending series of facts initiated by 'Fa' being true. There is only an unending series of sentences involving embedded uses of T' and 'corresponds to'. In short, one may hold that the fact that a is F grounds the truth of 'Fa' and each member of the series of sentences: T 'Fa', T " T 'Fa' ", . . . Moreover, one who holds that there is an infinite series of facts need not hold that the sentence 'Fa' corresponds to the series in the sense in which it corresponds to the fact that a exemplifies F. Assuming that a different fact is the truth condition for each member of the series, one knows that if there is a fact corresponding to 'Fa' there are an infinite number of facts. This is not any more paradoxical than holding that the series 'Fa', 'Fa & Fa' , 'Fa & (Fa & Fa)', . . . generates an infinite number of facts, but that 'Fa' corresponds to only one fact. This indicates an ambiguity in Frege's use of 'corresponds'. On a correspondence theory, we must distinguish between the possibility an atomic sentence corresponds to and the fact that makes a sentence true. Thus, one can hold that the fact that a exemplifies F grounds the truth of 'Fa & Fa' but that the conjunction does not correspond to that fact in the sense that 'Fa' does correspond to it. (This issue is discussed in 'Negation and Generality'.) This raises a question about the sense of correspondence that is involved in a correspondence theory. Consider the atomic sentence "Fa' and the following conjunction : (C)

'Fa' Den Fa & Fa,

with 'Den' read as denotes. The first conjunct makes use of one notion involved in the ambiguous term 'corresponds'. It expresses the fact that on the correspondence theory an atomic sentence is linked with or denotes a possibility or situation, as Wittgenstein put it in the Tractatus . The second conjunct expresses the idea that the denoted situation's existence is the condition of truth for the sentence. Thus, to say that a true atomic sentence corresponds to a fact is to say that the possibility denoted by the sentence exists or is a fact. In putting matters this way , we are forced to recognize that a sentence is used in two ways. In the expression " 'F a' Den Fa" the sentence 'Fa' stands for a possibility, irrespective of that sentence being true or false . As the second conjunct of (C), the sentence 'Fa' is used to state that the possibility exists or is a fact. As Wittgenstein once put it : 24

4 . 022 A proposition shows its sen se . A proposition shmvs how th ings stand if it is true . And it says that they do so stand .

(One might note a sim ilarity between this two-fold use of a sentential pattern on the correspondence theory and Frege's taking a sentence to shift reference, sometimes referring to a truth value and sometimes to a proposition . The twofold use also corresponds to Frege's taking a sentence to express a proposition while denoting a truth value.) The use of the sentence "Fa' in two ways in (C) may be seen as the upshot of Frege 's argument. For, it would seem hopeless to replace (C) by (C')

' Fa' Den Fa & Fa exists,

since one would apparently be forced to recognize the further possibility stated by 'Fa exists'. This , in turn, would give rise to yet another possibility, and so on. And this could be taken as a way of construing Frege's argument, though he does not put it in terms of possibilities which may exist or be realized. Thus, using (C) and taking a sentence to play two roles, or both show and state in Wittgenstein's manner, can be taken to implicitly involve the purported unending series that (C') involves . An adequate correspondence theory must resolve this problem. 6 If one modifies (T), by usi ng (C), so that we have

Frege's argument may be seen to raise the threat of another infinite series of facts. For, even if the conjunction ""Fa' Den Fa & Fa" is not taken to denote a fact, and aside from the problem raised by (C'), there would appear to be a further fact denoted by "" 'Fa' Den Fa . " And, if this is so, then (T2)

T"Fa' Den Fa" - (((''"Fa' Den Fa") Den ('Fa' Den Fa)) & ("Fa ' Den Fa))

would involve yet another fact, and so o n ad infinitum. This, too, is reminiscent of a feature of Frege's view: the infinite hierarchy of senses initiated by a term with a sense. But, the correspondence theorist need not take (C) to be true in vi rtue of any fact other than the fact that a exemplifies F. The first conjunct of (C) is true by the rules of the schema and the interpretation of the sentence 'Fa' . That conjunct, however, neither states such rules nor states , about itself, that it is interpreted the 25

way it is. To take it to do so would generate a regress and a paradox. For that matter , one should not take (T 1 ) to be a statement of the correspondence theory as applied to a specific case. Given the context of a correspondence theory, we understand how to take (C) and (T 1 ) , but neither is a statement of the correspondence theory. The statement of such a theory will involve the use of a background language making use of notions such as fact, property , denotation, etc. , with which we may discuss schemata containing expressions like (C) and (T 1 ) . The introduction of a predicate like 'T' into a schema does not constitute a theory of truth. (A number of the issues involved are discussed in 'Mapping, Meaning, and Metaphysics', 'Sellars and Goodman on Predicates, Properties, and Truth', 'Facts and Truth' , 'Logical Form, Existence , and Relational Predication', and 'Belief and Intention'. ) I said that "'Fa' Den Fa" does not state a fact. Yet, it is a fact that the signs 'a' and 'F' have been interpreted the way they have and that certain rules for forming sentential expressions are used. In a sense one may say that sentences like 'Fa' are, in part, true in virtue of such facts. The point is that they do not state such facts. Neither do sentences like ""Fa' Den Fa. '' If we had not interpreted the sign "a', then "" 'a' Den a" would be gibberish. And, if we had named a different object by using the sign 'a', then the second occurrence of 'a' in '"a' Den a" would denote a different object. But, given that 'a' has been interpreted, we may infer that ''"a' Den a" is true. In a perspicuous schema, all instances of the pattern .. .. . . ' Den . . . ", for primitive constants in place of the dots , are consequences of the interpretation rules. Thus, it will not do to hold that '" a' Den a" states that the sign 'a' has been coordinated to the individual a. For , had we interpreted the sign differently, we would still have an instance of the general pattern '" ' . . . ' Den . . . '' , and, hence , a true statement . It will not do to say that the original statement is no longer true , since the "new' true statement is a different statement. Given a different interpretation we have a different statement, but there is no "old' statement that is now false. The sign pattern "''a' Den a" will be a true sentence under any interpretation of the sign •a'. Hence , if it is a statement at all , i. e. a correctly interpreted sign pattern, it is true. In that sense , it cannot be taken to state a fact. The same holds for '" "Fa' Den Fa. " Thus , one cannot generate an infinite series of facts from (C) , since the only fact expressed or stated by (C) is the fact that a is F. If one does not realize this , then one might think that the problem Frege 's argument establishes for the correspondence theory of truth is that we cannot go from (T 1 ) , taken as stating the truth condition for "Fa', to

T'Fa' 26

= Fa.

For, while we could get

we could not get

Fa =:) TFa· , since that would involve deriving

Fa =:) ' Fa· D en Fa . B ut, th is latter is supposedly not derivable, since it might be true that a is F but not true that a was named •a' and that 'F denoted the property F. However, we must realize that (C) and (T 1 ) are interpreted sign patterns and, as such, given the schema, the context of interpretation, and the way we construe denotation, ""Fa' Den Fa" is derivable in th e schema. Hence , it trivially holds and so does "'Fa =:) 'Fa' Den Fa . " All of this highlights th e point that a philosophical position is not stated by sentences of the schema , but in a discussion of th e interpretation and use of the schema. We must understand the correspondence theory in terms of a general statement about atomic sentences of perspicuous schemata, and not in terms of sentences within such schemata. Perhaps Frege' s obj ection reduces to the point that in one sense to state that a sentence is true is to do no more than to assert the sentence. If one offers a th eory of truth , like the correspondence theory, he purports, in another sense, to say something more. But statements like (T) and (T 1 ) do not state such theories , though, taken in the proper context, they may be said to reflect such theories. Some of the misguided criticism of th e correspondence theory may stem from taking statements like (T 1 ) to be tantamount to a statement of the theory in a formal language. Freges' s distinction between concepts and obj ects, and his account of predication based upon that distinction, enabled him to take up another problem that was later to perplex Russell and to become a, if not the, fundamental question that occupied Wittgenstein in the Tractatus : the nature of logical concepts. Just as Frege had noted a structural similarity between predicates and function signs, he saw the same feature in truth functional connectives and quantified contex ts. Thus, one could take '--- ', '& ', etc. to indicate truth functions correlating truth values (or pairs of truth values) to truth values. One could also consider such signs to indicate propositional functions that correlated propositions to propositions: the proposition expressed by "Fa' is correlated to the proposition expressed by · --- Fa' by the function indicated by ' --- ', for

27

example. (This points to an ambiguity in Frege's notion of a function.The ambiguity and its implications for Frege's pattern are discussed in 'Frege on Concepts as Functions. ') Frege applies the pattern to quantified expressions so that the proposition expressed by '(x) (Fx � Gx)' could be taken to result from a function , indicated by the expression '(x) (x � lLJx)' , taking the concepts indicated by 'Fx' and 'Gx' as arguments in the appropriate order. Frege thus applied a few fundamental ideas to a variety of fundamental problems oflogic and ontology. The fundamental distinction on which the pattern is based , that between functions (concepts) and objects, as well as other features of the pattern, markedly influenced the development of Wittgenstein's and Russell's Logical Atomism . It has not been sufficiently appreciated just how much of a Fregean Wittgenstein was in the Tractatus. Frege's ideas also provided the basis for the resurgence of nominalism in analytic ontologies of the twentieth century. Quine, Sellars , Goodman and a number of their followers have adopted the pattern whereby predicate signs, used only as predicates, are held to be free of ontological commitment. Frege had held that predicate signs may not occur in subject place in a sentence , since such occurrence would contradict their role as predicative expressions . Thus , if we represent functions by signs , like 'Fx' and ' Rxy' (or 'xRy'), revealing the predicative role of the signs , rather than by 'F' and "R' , which appear to be too much like names of objects, we can see that such signs may not occupy subject place, since an 'incomplete' expression would result. Wittgenstein, recall , used this point to dismiss Russelrs paradox in Tractatus 3. 333. Thus , by combining 'Fx' and ' a' we get ' Fa', a complete sentential expression, but if we combine 'F2 (0) ' and 'Fx' we get 'F2 (Fx)', an incomplete expression. The same result is obtained in the case of the context '. . . = . . . ', since we obtain 'Fx = Fx' , an incomplete, and hence non-sentential , expression. Thus, functions (concepts) cannot be said to be self-identical. To be self-identical becomes , for Frege , a criterion or condition for being an obj ect. For Quine, guided by the slogan 'no entity without identity', identity contexts come to serve as a mark of ontological commitment. This leads to his famous ,, pronouncement that '"to be is to be the value of a variable . For, where one has identity contexts for terms, one allows their use in subject place, and subject place, for Quine , is the position natural to q uantification. This fits with his attempt to replace all names and singular expressions by definite descriptions, construed in Russellian fashion. (These issues are taken up in 'On Pegasizing', 'Professor Quine, Pegasus , and Dr. Cartwright', ' Nominalism, General Terms, and Predication', and 'Strawson and Russell on Reference and Description'. ) If one uses Russellian descriptions in place of proper names then subj ect-predicate sentences become existential statements. Hence, one makes a literal

28

statement of an ontological commitment. Moreover , using a name, the closest one can come to a statement that the named obj ect exists is by means of a sentence like " (3x) (x = a) '. Yet , such a statement is normally a consequence of ·a = a', which , in turn, is a consequence of '(x) (x = x)'. Thus , one must give up such standard sequences or not take " (3x) (x =a)' to state that a exists , assuming it is not acceptable to have 'a exists' follow from '(x) (x = x) '. But , giving up such standard sequences is in the pattern of those who speak of possible worlds, possible objects , and , loosely and absurdly, of ontically •free· existential quantification. Such moves are understandably not to Quine 's taste. A simple resolution of the problem is to replace names by Russellian descriptions. Such matters about names aside , Quine adopts Frege 's pattern and maintains that one may use primitive predicates in a perspicuous schema without ontological commitment to entities correlated with such predicates, so long as one does not allow for quantification 'over' predicates. Sellars, while rejecting Quine ·s restriction on predicate quantification, also believes that one may use primitive predicates without taki ng such predicates to have referents. Thus, he denies that properties, whether taken as universals or as quality instances , exist. He, too, thus incorporates a Fregean theme, for while he allows for predicate quantification , he does not allow for the use of predicates in subject place. Goodman goes even further. For whi le Quine and Sellars speak of predicates being 'true of' , obj ects, whereas names 'refer to obj ects , and face problems explicati ng such a use of ·true of' , Goodman suggests that we use predicates , like names , as classificatory devices that we impose upon obj ects. He seeks to support his move with classical arguments of the relativist and idealist and thus explicitly links contemporary nominalism with relativistic idealism. The link between nominalism and idealism is found on the contemporary scene in yet another way. Once properties are rejected, facts soon follow , since the traditional construal of facts, as in Moore, Russell, and , I believe, Wittgenstein, is as complexes of universals and particulars. With the rejection of facts, one also rejects the correspondence theory of truth that Moore and Russell made a cornerstone of the attack on idealism and which Wittgenstein also embraced. Thus , two Fregean themes join together in contemporary nominalism : the attack on the correspondence theory of truth and the claim that predicates do not refer to objects that are self-identical, while different from other things of the same kind , for, j ust as we cannot have 'Fx = Fx' as a sentence, so we cannot have 'Fx -=I=- Gx'. The rejection of the correspondence theory of truth , via the rejection of properties and facts, has led to, or at least supported , the contemporary revival of holistic­ ideal istic patterns associated with coherence accounts of truth and meaning. Goodman ' s extreme form of both nominalism and relativism,

29

whereby we supposedly classify objects by our use of predicates and 'make things white' by 'calling them white' , as well as 'make worlds' by our classificatory and representational frameworks, is an obvious example. Goodman emphasizes the ' arbitrariness' and 'conventionality' of ascriptions of predicates. This form of 'conventionalism' is reinforced by conventionalist themes stemming from issues in the philosophy of science, regarding the purported conventionalism in theory selection, and from questions about the conventionality of representational frameworks. Such mutual reinforcement plays a significant role in the idealistic-pragmatic revival led by Quine, Sellars, and Goodman and prophetically advocated in the work of E. A. Singer' s student C. W. Churchman. On the general cultural scene, Ernst Gombrich' s writings on the theory and history of art popularized such idealistic themes and supported them with the classical data from gestalt psychology and the psychology of perception. It is not an accident that Goodman's extreme relativism went hand in hand with his widely known work in the philosophy of art focusing on the topic of representation . (Various stran ds of the 'new nominalism' and the attack on the correspondence theory of truth are taken up in 'Mapping, Meaning and Metaphysics' , ' Sellars and Goodman on Predicates, Properties and Truth' , 'N ominalism, General Terms and Predication' , 'Russell's Proof of Realism Reproved' , 'Platonism, Nominalism and Being True of and "Logical Form, Existence and Relational Predication· . ) To many, idealism is naturally opposed to materialism. B ut, the structural pattern of holistic idealism, involving coherence accounts of truth and meaning, is found in the materialistic philosophies of Quine and Sellars. The connection is und erstandable. The atomism of Russell and Wittgenstein and the realism of Moore involved the recognition of independent facts that were the truth conditions for atomic sentences. The empiricist orientation of Moore and Russell also led them to hold that such truth conditions were presented in experience and , hence. known by experience. as were the constituent particulars and properties in such facts. A natural target of attack for the twentieth century nomina lists became the claim that independent facts were given in experience. This led to an assault on the empiricists· claims about the objects of experience and a critique of the idea that there are fundamental or basic atomic facts that are 'independent'. For the realistic atomists. such facts were independent in two important senses. First. the existence of any one of them wa s logically independent of the existence of any other atomic fact. Second, such facts were held to be presented or known in experience independently of the experience of other facts or things. Such fundamental facts became the basis for all empirical knowledge on the realistic empiricist's pattern. In this sense, atomic facts were the

30

"foundations' of factual knowledge. Not surprisingly , the nominalists attacked the idea that there were such ultimate foundations of truth and knowledge. They emphasized , in holistic fashion, the role of the conceptual scheme or framework which purportedly determines both truth and meaning. Thus , Sellars attacked the 'myth of the given' and Quine attacked the "dogmas of empiricism' . One dogma was the claim that there were ultimate statements whose truth and meaning was 'independent' of other statements . The other dogma was the corollary claim that there was an absolute difference between synthetic and analytic statements . For Quine, such a difference could only be a matter of degree, depending upon the role of a statement within its conceptual framework. Implicit in this dispute is the reference theory of meaning espoused by Russell and Wittgenstein, and implicitly by Moore, in opposition to Frege's insistence on the distinction between meaning and reference. The meaning of primitive terms was taken by Russell to derive from their reference to items of experience. Atomic statements were 'independent' precisely because their constituent terms meant what they referred to , and statements which contained such terms "non-vacuously' were synthetic, while such terms ocurred ·vacuously' in analytic statements. Thus, yet another Fregean theme was employed by the nominalistic holists. Since the ultimate or atomic facts had traditionally been taken by Russell and Moore, and later by Bergmann, to involve phenomenal constituents, such as sense data and mental acts, the nominalistic holists attacked the purported existence of mental entities . Thus, nominalism joined with materialism in the philosophies of Quine and Sellars. This merging of nominalism, holism and materialism was aided by the way in which the nominalists treated predicate terms as "being true of particular objects without standing for properties. So treating such predicates, one may hold that a predicate like "is angry' applies to a person in virtue of certain manifest and dispositional behavioral and physiological states without holding that the predicate "means' or 'refers to' a common characteristic of such states. Frege's separation of meaning and reference also played a role in the arguments for materialism. It allows the nominalistic materialist to hold that primitive predicates may be meaningful without being taken to represent anything , except in the sense of being true of particular things. One may then take the meaning of such terms to be supplied by the conceptual framework and hold that j ust as the expressions 'the morning star' and 'the evening star' do not mean the same thing, though they refer to the same thing, so particular states of anger may be physical states, even though 'being angry' does not mean the same thing as any description of a physical state. Moreover, since truth, like meaning, is a matter of the total conceptual scheme, due to the rej ection of fundamental atomic facts, another argument emerges. 31

Quine has held that the 'unit' of meaning and truth is the conceptual framework as a whole. What entities one then takes to exist are those entities whose postulation yields the ' simplest' and most'comprehensive' framework. Whether mental entities exist then becomes a question about whether such entities need be postulated and not a question about whether they are found in experience. If one then asks about what conceptual framework determines the entities to be postulated , the answer is the framework suitable for the accommodation of science, logic, and mathematics. These themes (discussed in 'Physicalism, B ehaviorism and Phenomena' , 'Of Mind and Myth' , 'Mapping, Meaning and Metaphysics' and 'Belief and Intention') show the influence of another historical force behind the merging of nominalism, holism, and materialism in recent analytic philosophy. The logical positivism of Schlick, Carnap and the Vienna Circle declared that traditional metaphysics and ontology offered meaningless answers to meaningless questions. They perceived the philosopher's task to be the analysis of the language and structure of scientific theories, of logical systems, and of mathematics. Declared to be the handmaid of theology in earlier times, philosophy was now seen as the means for obtaining conceptual clarification in the empirical and formal sciences. To many philosophers in that tradition, science was materialistic. In part, this attitude stemmed from the long struggle to free various sciences from teleological, theological, and u nfortunate philosophical influences. An awareness of and preoccupation with questions of methodology and concern with meaning criteria also played a role. In particular, introspection was rej ected in psychology, and scientific psychology was equated with behaviorism and physiological psychology. Such a view of psychology was taken to imply materialism. This development was reinforced in the United States by the pragmatists, who, under the influence of Dewey's "instru mentalism' , hailed the destruction of traditional metaphysics and ontology. It is worth recalling Dewey' s early 'functionalism' in psychology and the early behaviorism of other American pragmatists such as E. A. Singer. Preoccupation with introspection in psychology and, along with it, belief in mental entities were considered unscientific and out of date. Developments within the growing area of the philosophy of science also contributed to the widening influence of the holistic pattern: the emphasis on the theory laden nature of data; the insistence on the choice factors that, in principle and in practice, are involved in the selection of theories and hypotheses and the consequent importance of notions like simplicity and comprehensiveness , as opposed to truth; the highlighting of the roles of alternative geometries in physical theories; the concern with meaning criteria for scientific--- terms and the preoccupation with

32

" operational definitions· , which focused attention on the observer and the activity of the scientist; the anomalies of quantum physics and the apparently heightened role of the "observer· in both quantum and relativity physics. All of these developments contributed arguments and grist for the idealistic mill. Moreover, a long history of the construal of universals as concepts or objects for conternplation and apprehension by the mind, rather than of perception by means of the bodily senses, associated realism about universals with claims about the existence of mental entities. This not onl y helped nominal ism to fuse with materialism. but it encouraged those who advocated nominalism, holism, and materialism to see themselves as scientific anti-metaphysi­ cians. As some saw it. metaphysicians believed in minds and universals, while scientifically minded or, in Feigl' s memorable phrase, 'tough minded' philosophers were nominalists and materialists who were satisfied with tentative hypotheses warranted by their role in the ongoing and growing conceptual framework. B radl ey' s Hegelian Absolute was replaced by Dewey' s unobtainable, but approachabl e, culmination of 'inquiry' and Sellars' "Peirceian conceptual structure. ' Thus, holism, by its emphasis on the hypothetical and 'open' nature of factual and theoretical claims. supposedly fit, in yet another way, with the picture of science as dynamic, ongoing and open. Ironically, the arguments employed and the theses promulgated by the holism spawned by pragmatism and positivism were in structure, if not in terminology, akin to the early arguments of Bradley and his concern with an Absolute, rather than an ultimate or ideal conceptual framework. That the 'new idealists' · take reality to be material, rather than mental , is a relatively insignificant departure from classical idealism, since what exists is determined. rather than described, by our ultimate and ideal conceptual framework. Bergmann, whom Sellars has labeled the 'arch realist' on the contemporary scene, consistently opposed these major trends. He early opposed what he called 'philosophical behaviorism' as opposed to · methodological behaviorism', which he advocated . The former amounted to a materialist metaphysics, while the latter was a view, within the philosophy of science, about the nature of the concepts and laws of the science of psychology. While a behaviorist in regard to the science of psychology, B ergmann rejected materialism as a philosophy of mind. In his writings in the philosophy of science ( covering operationism, the place of values in social science, philosophy of psychology, psychoanalysis, quantum physics, scientific real ism, measurement, probability, and space and time) , Bergmann consistently attacked the idealistic and holistic trends. In this he was aided by May Brodbeck's writings on the social sciences and critical analyses of pragmatism. As

33

Bergmann's philosophy developed, it became apparent that his views about particulars and universals and his analysis of intentionality were the cornerstones of his realism, as opposed to idealism. In this respect he followed the path taken by Moore and Russell at the turn of the century and helped bring the issues that concerned them back onto the contemporary scene. At a time when many contemporary analytic philosophers were behaviorists and materialists it was not surprising that the problems of intentionality should be ignored. Yet, some philosophers tried to handle such problems in a behavioristic framework. There was a reason for their doing so that reveals yet another motive behind the rejection of universal properties by contemporary analytic philosophers. In Meaning and Necessity Carnap had taken properties to be intensions of predicates while taking classes to be extensions of such terms. Terms were no longer construed as having referents, but as having extensions and intensions, since, as Carnap saw it, the notion of 'reference' was problematic. ('Properties, Abstracts, and the Axiom of Infinity' is, in part, a critical analysis of Carnap's attack on referenee.) Carnap attempted to use this pattern, derived from Frege's distinction between sense and reference , to resolve the same set of problems that Frege confronted: problems concerning identity, intentional contexts, predication, logical concepts , reference and meaning, arithmetic, necessity. Carnap sought to deal with these problems in the context of a system of modal logic. Modal contexts are intensional in the same sense that intentional contexts are intensional. One cannot replace sentences in modal contexts by sentences with the same truth value ( extension) any more than one can make such replacements in contexts embodying intentional verbs. Properties are thought, by some, to pose similar problems. In addition, some philosophers, like Quine, feel that there are no clear cut identity criteria for determining when predicates stand for the same property, as there are supposedly no unproblematic criteria for determining when sentences state or express the same proposition or terms have the same meaning. Thus, so-called intensional, and , of course , intentional, entities become suspect. This is easily understood when we recall the role 'identity criteria' play in ontology for philosophers like Quine. Quine thus rejects intensional entities. Russell , too, rejected propositional entities and had , very early, offered a ·relational' analysis of belief, which made belief contexts extensional. Quine has recently attempted to revive Russell's early analysis. (These issues are discussed in ' Belief and Intention'.) But , unlike Russell, Quine rejects not only propositions but all intensional entities . Hence, he rejects universal properties on such grounds, and thus he has another motive for his nominalism beside the desire to avoid 'abstract objects'. The attempt to analyze intentional contexts so that 34

they are extensional rather than intensional thus goes hand in hand with the rej ection of intensional entities like universal properties. The current fashion of calling properties 'intensional entities' reveals a confusion , fostered by Carnap's analysis , which has clouded much recent discussion. For Carnap, a predicate term had no referent, while it did have an extension , a class, and an intension, a property. Much recent discussion takes a predicate to be an extensional term if it is held to refer to a class, and an intensional term if it refers to a property. Yet, if predicates are taken to refer to properties, they may be said to be extensional terms if they may be replaced by other predicates that refer to the same property, in appropriate contexts. In another sense, they are not extensional terms if one does not allow for their replacement by other predicates that are ·coextensive , but which refer to different properties. Mixing these senses of ·extensional', some confusedly come to think of properties as intensional entities, as opposed to classes, which are said to be extensional. All this is aided by the belief that there are no clear cut identity criteria for properties (as there are supposedly not such criteria for othe r intensional entities, such as propositions and meanings or senses), while classes are identical if 'they' have the same members . Thus , as the term is used, in one sense the extension of an expression is its refe rent , while in another sense it is given by what the term applies to. Hence , in the case of predicates we can have properties as extensions, or referents, and classes as extensions, as collections of the things to which the predicates apply. The issues are complicated further since properties are often said to be the meanings of predicates, if predicates are taken to refer to properties, and properties are also frequently taken to be concepts, which, like propositions, are intensional entities. Moreover, properties may be taken to be extensional entities in two senses. They are the referents of predicates, and hence extensions in one sense. And, if a predicate may replace another that refers to the same property, and preserve truth value in certain contexts, the predicates, and derivatively their referents, are said to be extensional in a second sense . Properties are not extensional, in a third sense, in that the same class may be determined by two properties. But, then, if one rej ects properties as intensional entities, he does so simply because identity of extension (class) does not imply property identity. It is as if the very terminology Carnap propounded has led contemporary nominalists to believe that they have found a further reason for rej ecting properties. Historically , Carnap's terminology is derived from Frege 's pattern . For, recall that while Frege's fundamental distinction is between concepts (functions) and all complete ' obj ects' , he did link concepts along with propositions and senses, and not with truth values and obj ects, which were the referents of sentences and singular terms. Thus concepts, along with senses, were

35

taken by Frege to be constituents of propositions, which have become the prime examples of intensional entities. Contemporary nominalists who do not rej ect intentional contexts have taken behaviorism to offer a way of treating such contexts without appealing to properties. If one attempts a relational analysis, along Russell's lines, one uses predicates in subject place. For, on Russell's pattern, in the analysis of 'p thinks that a is F' the Fregean (S) TH(p, Fa) , we considered earlier, would be replaced by (R) TH(p, F, a) , with 'TH' standing for a three-term relation with a person , a property, and an obj ect as terms. Thus, one avoids propositions, as they are employed by Frege in (S), at the price of accepting universal properties. Moreover, (R) is also inconsistent with the fundamental Fregean theme , that Russell and Wittgenstein adopted, whereby predicate terms may not function as subj ect terms in any contexts. Russell was plagued by this problem and abandoned his relational analysis. Contemporary nominalists, like Quine and Sellars, having adopted the basic Fregean stricture on the use of predicates, cannot employ (R) to avoid the propositional entities appealed to by the use of (S).An apparent way out is offered by the basic nominalistic pattern whereby one treats predicates as "true of' objects, with true of taken as a basic relation between predicates and obj ects, just as reference is a basic relation between singular terms and obj ects. Thus, (R) is transformed into :

(0) TH(p, 'Fx' , a) , with 'TH' read as ' . . . thinks that . . . is true of . . .'. Aside from the transparent artificiality of taking TH as a primitive or basic relation, while understanding it in terms of thinks that and is true of, such an analysis clearly presupposes the viability of analyzing predication in terms of being true of. If this analysis is not cogent, ( 0 ) is pointless. (A number of the essays in this book raise arguments against various versions of such a nominalistic analysis.) Aside from the failure of the general nominalistic pattern on which (Q) is based, the purported analysis of intentional contexts that employs patterns like ( 0 ) fails for another reason. It must take a thought that a is F to be intrinsically connected with a sign , say ' Fx' , but one cannot avoid recognizing that the sign must be used in a relevant way, namely to stand for F. Various 36

attempts to avoid this by Quine, Davidson, and Sellars have offered little beyond complicated and esoteric linguistic devices. 7 A crucial part of such attempts has been the construal of thoughts in terms of dispositions to verbal behavior, which reveals another lin k between nominalism and materialism. Yet, it is clear that verbal behavior must be construed in terms of the production of verbal tokens, and it appears obvious that tokens of predicates can play the role they must only if they are taken to stand for certain properties. The nominalistic behaviorists must analyze such apparently referential uses of predicates in a way that avoids the recognition of properties. They attempt to do so by offering behavioral analyses of what takes place when a verbal token is said to refer. If such analyses were viable they would also provide a pattern for the behavioral analysis of the referential use of tokens of proper names and demonstratives. For, if talking about dispositions to utter predicates, and utterances of predicates, enables us to avoid taking predicates to stand for properties, we should be able, in a similar manner, to avoid talking about singular terms referring to obj ects. This would amount to a behavioral analysis of reference. This is hopeless. It is as hopeless as corresponding attempts to furnish behavioral analyses of other mental states, of feeling, perceiving, and so on. (These issues are taken up in ·B elief and Intention', ' Of Mind and Myth', and 'Physicalism, Behaviorism and Phenomena' . ) The essential pattern behind such behavioral analyses is simple and familiar, though it is not always presented in simple and straight-forward terms. The apparent fact that two tokens of a predicate are used to refer to the same property is construed ( analyzed) in terms of the occurrence of the two tokens functioning as causes and effects in similar ways. Hence one dispenses with the purported reference to properties by talking about the ·causal framework' for the utterance of, and dispositions to utter, tokens of words. The attempt to provide such analyses is what lies behind much of the currently fashionable talk of 'causal theories' of reference. Such theories reveal a con fused failure to distinguish between philosophical problems and analyses and scientific questions and answers. Providing necessary and sufficient causal conditions for the referential use of terms no more provides an analysis of the phenomenon of reference than providing the necessary and sufficient physiological conditions for pain resolves the mind-bod y problem or establishes materialism. To put it another way, there is a difference between causally necessary and sufficient conditions and analytically necessary and sufficient conditions. So-called ' causal theories' are also presented by those who seek to provide the 'link' between an obj ect and a term, without advocating any kind of materialist account. Thus, Kripke says:

37

A rough statement of a theory might be the following: An initial baptism takes place . Here the obj ect may be named by ostension , or the reference of the name may be fixed by a description . When the 'name is passed from link to link ' , the receiver of the name must , I think, intend when he learns it to use it with the same reference as the man from whom he heard it . 8

But this is not merely a 'rough statement', it is no statement at all as far as a philosophical analysis is concerned. For, the problem is to specify just what it is to 'intend' or 'intend to refer to' an object when one uses a token of a term, or hears a token , for that matter. Kripke is aware of the problem, as he continues: Notice that the preceding outline hardly eliminates the notion of refere nce ; on the contrary , it takes the notion of intending to u se the same reference as given .

and concludes: To repeat , I may not have presented a theory , but I do think that I have presented a better picture than that given by description theorists.

But, in not presenting a theory , what he does is reiterate , in different words , Russell's standard distinction between logically proper names that refer and definite descriptions that denote and repeat Russell 's claim that such names cannot be replaced, in Quine's fashion, by descriptions . though, he, following Strawson, disputes with Russell about what terms are genuine names. Unfortunately, Russell's empiricism , which is logically independent of his 'theories' of reference and description , complicated matters for later philosophers who derive their 'theories' from his writings. Russell held that acquaintance furnished the " link' between a name and its referent ( or a predicate and the property it referred to) in that one must be acquainted with an object to refer to it by the use of a demonstrative term. Holding that we were not acquainted with physical objects, on standard empiricist grounds. he also held that we could not refer to physical objects by names or other demonstrative terms. Thus, we were forced to denote them by definite descriptions. This view has led many to place Russell among those philosophers who , like Quine , have advocated the replacement of names by definite descriptions. This he did not do. (These issues are taken up in · on Pegasizing' , " Professor Quine , Pegasus, and Dr. Cartwright' , and 'Strawson and Russell on Reference and Description' . ) Moreover , many philosophers , influenced by the development of ·ordinary language' philosophy , reject acquaintance with an object as a suitable 'link' for using a term to refer to that object. After all , we normally refer to objects - Plato , Aristotle and the Sphinx -that are not objects of acquaintance, in

Russell's sense . Thus, the causal chain whereby the name "is passed from link to link' enters the picture . But this merely replaces a deep philosophical concern of Russell's , about reference and intentionality , with the commonsensical triviality that we learn who Aristotle and Plato were and what the Sphinx is from books or statements using tokens of those terms , derived from other books or statements using those terms, etc. , until one goes back to a situation that plays the role of Russell's experience of the object. But the two fundamental ph ilosophical questions involved are either not rai sed or judiciously avoided. One question we already noted: what is the analysis of reference, or as one commonly says , the concept of reference? The, second, related question , concerns the justification for the claim that one refers to an object, say Aristotle , when one uses a token of the term 'Aristotle' , in the same sense that one refers to a prnsented object by a token of a demonstrative term. This question, like many classical questions , is conveniently avoided by the uncritical acceptance of ordinary usage. For , as we commonly speak . we say that we refer to Aristotle, rather than Plato , and this object , rather than that one. Since we speak of referring to Aristotle and Plato , how can one then sensibly question whether or not we really do so refer? After all, have not "ordinary language' philosophers , influenced by their interpretation of Wittgenstein' s later philosophy, di smissed the philosophical problems of perception by assuring us that we do perceive physical objects, since we speak as if we do? What goes for the problems of perception goes for the problems posed by reference as considered by Kripke. Thus , when he speaks of providing a set of necessary and sufficient conditions in order to advance a theory of reference, he doesn't really mean what he says. What he means is that one provides a set of necessary and sufficient conditions that will pick out those cases where we normally talk about a reference being made. To do this is neither to pick out those cases where, on a viable notion of reference , we actually refer to objects nor to show that cases where we normally speak of referring are cases where we actually do refer. It is one thing to set out criteria that would pick out cases where we ordinarily speak about knowing something to be true or of something being the same object. It is quite another thing to offer analyses of knowledge, truth and identity in terms of which we can distinguish cases where we may viably say that we know something and that an object is the same as ' another'. Unfortunately, the legacy of "the linguistic turn' has led many philosophers to uncritically accept common speech as the embodiment of philosophical truth. The philosopher's task is then seen to be the proposing of 'criteria ' which pick out the same items that we commonly apply certain words to - cases of reference, perception, knowledge, identity, justice , etc. The offering of such co-extensive

39

criteria is then identified with the achieving of a philosophical analysis of the ordinary concept , since we are provided with purportedly necessary and sufficient conditions for the application of the original terms. Russell , by contrast, challenged common assumptions (and usage) regarding reference, judgment , and predication , just as Berkeley and Hume had challenged assumptions about causality and perception. He sought to justify claims about reference and not merely look for conditions obtaining in cases where people commmonly say that a reference is made. In short , Russell undertook a traditional philosophical or critical analysis that would resolve traditional problems about the nature of judgment and about the connection between thoughts and their objects. He did not seek to set forth a report on the usage of 'refer'. Kripke and other causal theorists , at best , do the latter . In this they follow in the footsteps of Strawson's early attack on Russell's account of reference and description, notwithstanding the fact that in some respects Strawson is criticized and Russell defended. Strawson 's critique of Russell reflected a more general position on the nature of philosophical analysis involving his celebrated attack on ·revisionary metaphysics' and his advocacy of 'descriptive metaphysics'. This , in turn , was part of the development, within analytic philosophy , of the ordinary language movement that stemmed from Moore's early attack on idealism and subsequent defense of common sense realism. (These questions are discussed in 'Strawson and Russell on Reference and Description' , 'Belief and Intention' , and in my article · strawson , Scepticism and Metaphysics' in Theoria, XLI I , 1 978.) There is , of course, a point to 'ordinary language' philosophy. In order to provide adequate philosophical analyses of reference, truth , perception, identity and so on, one must account for the features that lie behind our ordinarily speaking , in a broad , loose and philosophically problematic sense , of reference, perception , etc . in a wide variety of circumstances. A philosopher who denies that physical objects are really perceived , or that we can really kno w that a physical object perceived at one time is identical with a physical object perceived at another time , must distinguish cases of veridical perception from hallucinations and explain in what senses we may be said to perceive physical objects and to recognize an object to be the same as one previously perceived. Likewise , in giving an account of reference one must distinguish between uses of ·Aristotle' , of "Pegasus' and of logically proper names and reveal in what senses one may, and in what senses one may not , viably speak of reference in such cases. Thus , a philosophical analysis will have to accommodate , though not uncritical ly , common sense convictions that we do , on occasion , talk about Aristotle , rather than Socrates, even though we do not refer to Aristotle in the way one refers to a presented object.

40

The realist, as an anti-i dealist and as a proponent of universal properties, faces hi s own fundamental problems. The most seri ous kind of problem lies i n the ad diti onal entities reali stic philosophers find themselves obliged to recognize or which their cri tics foster upon them. Outstanding among such entities are the ad ditions to the category of facts. R ussell felt forced to recognize non-atomic facts such as negative facts and general facts. Wittgenstei n spoke as if he accepted possible facts, while Russell once accepted fictions and Moore seemed to waver between the recogni tion of non-existent facts and proposi tions as the correlates of false sentences. But neither Moore nor Russell nor Wittgenstein resolved the problem. And, it may be that faced with the acceptance of possible facts, as implied by the Logical Atomism of the Tractatus, Wittgenstei n abandoned ontologi cal analysi s and set out on the development of his " later' philosophy, with its critical preoccupati on with the reference theory of meaning that had been crucial i n his and Russeir s earli er work . The issue was to ari se again in Bergmann's attempt to analyze i ntentional contexts i n the context of his realism, in both senses, and anti materialism. In hi s important paper ' Intentionality' , Bergmann proposed to analyze · p thi nks that a is F' i n terms of three kinds of entities. He introduced propositional characters , represented in a perspicuous language by sentences in quotes ( replaced by corners in later works) , pa_rticulars which exemplify such characters, and generic properties, which such particulars also exemplified. The generic properties provided the ground for di fferenti ating various kinds of mental states, such as thinking, doubting, perceiving, etc. , while the propositional characters provided the basis for a mental state having its content. Such characters furnished the content to particular mental states exemplifying them because the characters stood in a unique logical relati on, means , to facts. Thus, with "M' standing for the meaning relati on,

was held to yield a well formed sentence when the dots were replaced by a well formed sentence. The quoted expression was taken to be a first order predicate standi ng for a propositional character of mental states. The relation M was held to be logi cal for two reasons. First, if one replaces both occurrences of the dots in (B1 )with tokens of the same sentence, Bergmann claimed the resulting sentence was a logi cal truth, while if different sentences were employed the resulting sentence was a logical falsehood . Thus, by analogy with tautologous patterns lik e 'q v ---- q', i nstances of (B1 ) are taken as true or false in virtue of their ' form'. The quoting operator and 'M' thus seem to be on a par with "v' and'----', for they

41

function somewhat like logical constants. Second, following Wittgenstein, Bergmann did not take logical constants like 'v' and '&' to be signs for relations. Where 'Rab' is an atomic sentence, if it is true, then for a Logical Atomist there is a fact consisting of a relation obtaining between the particulars a and b. By contrast, if 'Rab & Fa' is true , one need not hold that there is a relation obtaining between the facts stated by 'Rab' and 'Fa'.There is no fact with a constituent denoted by '& ' as there is a fact with a constituent denoted by "R'. Thus, '&' does not represent a constituent of facts that relates other constituents. It is, therefore, not a sign for a relation. This is one matter involved in speaking of conjunction as a logical relation or a 'pseudorelation'. In short, there are no logical relations recognized as entities in Bergmann's or Wittgenstein's ontology. This denial of ontological import to logical signs has an important consequence for Bergmann's handling of intentional contexts. 'M', being a logical sign, does not stand for a relational universal. Since 'M' is a logical sign, sentences of the pattern of (B 1 ), such as

are logical, not factual, truths. Thus (B1)is true irrespective of whether there is a fact in virtue of which 'Fa' would be true. Since (BJ does not state that a relation obtains between the propositional character 'Fa' and something else, that statement is true irrespective of whether or not the sentence 'Fa' corresponds to anything. If M were taken to be a relation and (B2) a statement of a relation obtaining between a propositional character and something else, then Bergmann's account would commit him to recognizing a correspondent of the sentence 'Fa', whether that sentence were true or false. This would mean that he recognized Russell's fictions, Wittgenstein's possibilities, or Moore's non-existent facts as correlates of false atomic sentences. Bergmann avoided such entities by holding that 'M' was a logical sign, like '&'. He appealed to the status of (B2) as a logical truth to account for how a propositional character could 'pick out' a specific fact, as a truth condition, without having to acknowledge correlates of false atomic sentences. In ( B2), only the expression to the left of "M' stands for an entity, since ( B ., ) is true as a matter of form, and not because two entities are related by M. The problem of possible facts is resolved by transforming an ontological question into a logical one, in the context of a view that denies the existence of logical relations. Bergmann did not stay satisfied with his solution. Faced with Bradley's paradox he, like Johnson, Wisdom, and Strawson , came to hold that we must recognize exemplification as a special tie or nexus. Such a tie required no further tie to tie it to its terms. In this respect a nexus differs

42

from a relation , which requires a further connection , a nexus . The crucial difference is · shown , by a perspicuous formal language in that the nexus of exemplification is represented by the j uxtaposition of terms , the type disti nction , and , implicitly , by the formation rules , and not by an additional relational sign . Thus, one naturally comes to think of exemplification as a formal or logical relation . But , in recognizing exemplification as the connection in facts , one then recognizes that a logical connection exists . Logical form is thus given ontological significance . In keeping with this , Bergmann proceeded to acknowle dge that M had ontological status . He would later proceed to recognize conj unction , generality , particularity , and universality. Giving M ontological status , Bergmann is forced to enlarge his ontological inventory . For, in (B-. ,) , •M' now stands for a relation al entity and not a pseudo-relation . What , then , is the second term of the relation M when ·Fa' is false ? The classical question receives a classical answer . When the sentence is false it stands for a possible fact . Bergmann thus recognizes both actual and possible facts . In addition to the difficulties posed by the recognition of possibilities , another problem arises . Bergmann , though a realist , is not a Platonist, since he accepts a principle of exemplification and holds that universals exist only if they are exemplified . Propositional characters are universals and , therefore , exist only if exemplified by mental states . Yet , in his paper ·Intentionality' , he gave as formation rules for "M' and the quoting operator (in his ideal language L) : a) Every sentence of L surrounded by quotes becomes a non-relational first-order predicate (type : f) with all the syntactical properties of a primitive descriptive predicate . b) Every sente nce of the form 'fMp' is well formed . 9

It is thus clear that every sentence of L surrounded by quotes stands for a propositional character irrespective of whether or not such a character has ever been exemplified by a particular mental state . That is , such propositional characters exist irrespective of whether anyone has ever had the appropriate thought . This , of course , brings him into conflict with his acceptance of a principle of exemplification . This conflict is not indicative of a trivial error or oversight . If he were to take a principle of exemplification as a guide in his treatment of predicates standing for propositional characters , he would have to add a proviso to the first of the above quoted formation rules for L: provided that there is a property of a particular thought that the predicate stands for . But, with such a modified rule , there is no point in using sentential expressions within quotes as predicates. On the meaning as referent pattern Bergmann accepts , one 43

provides meaning for a primitive descriptive sign, whether a predicate or a name, by coordinating the sign to its referent. Bergmann sought to have the propositional predicates, like sentences, have their meaning determined by their constituent signs and structure. This was essential to his claim that sentences of the pattern (B1 ) were logical truths or falsehoods. For, if one uses a primitive predicate, say 'F 1 ' , in place of a quoted sentence, in (B2 ) , then it becomes blatantly problematic to claim that (B2 ) , so modified, is a logical truth. For

is hardly analogous to 'q v � q' in the way that (B2) may be claimed to be. On the other hand, to retain (B2 ) and quoted sentential expressions as predicates makes Bergmann 's claim that propositional characters are simple universals untenable. This dilemma is a serious defect of Bergmann' s analysis. (It is explored in 'Belief and Intention' . ) Bergmann will eventually employ predicates like 'F 1 ' for his propositional characters, but he will not convincingly argue that sentences like (R,.) are logical truths. For that matter he will not convincingly argue that he really treats such predicates as primitive predicates and not abbreviatons of definite descriptions of the form '('f) (f M Fa)'.Yet, for all its problems. Bergmann's analysis, in recognizing a fundamental intentional relation, by contrast with the attempts of nominalistic-materialists to dispense with such a relation, is suggestive for resolving the problems of intentionality within a realistic framework. Moreover, while Bergmann has never attempted to provide analyses of the various puzzles involved in intentional contexts that occupy so much of the current literature, his pattern provides easy resolutions for many of them. In "Belief and Intention' I develop a modification of Bergmann ' s pattern and employ it to resolve a number of such puzzles. Bergmann 's analysis is a modification of the view Moore proposed in Some Main Problems of Philosophy.1 0 Yet, while, like Moore, Bergmann attempted to avoid propositional entities of a Fregean kind, unlike Moore, he came to explicitly acknowledge that his realistic philosophy forced him to accept possible facts. However, in so far as his view conflicts with a principle of exemplification, in that propositional characters exist whether or not they are exemplified by particular mental states, Bergmann also accepts Fregean propositions in the guise of universal properties. In this respect, one should keep in mind that Frege ' s propositions (thoughts) denote one of two truth values, the True and the False. Thus, in his way, Frege acknowledges entities corresponding to the possible facts forced upon the correspondence theory of truth. 1 1 Frege and Russell will always be linked with the logistic analysis of 44

elementary arithmetic. This analysis involved yet another application of Frege's basic pattern concerning predication and functions. For, he took numbers to be classes of concepts. This apparently conflicts with his basic idea that concepts cannot be objects. But , just as concepts can be arguments for "quantified functions' , so Frege would construe '(!) is a unit concept' as " (3x) (y) (y y = x) ', as Carnap did, and thereby not use · (l)x' as an incomplete subject term. Russell would adopt, in his lectures on Logical Atomism, this way of treating all higher level descriptive predicates, and thus advocate elementarism , the view that there are no higher level descriptive universals. (This issue and related questions are discussed in "Elementarism, Independence, and Ontology'.) Sellars, in turn, would attempt to use features of the pattern to argue for nominalism. ( " Mapping , Meaning, and Metaphysics' and 'Sellars and Goodman on Predicates , Properties, and Truth' take up such matters.) Whether Frege 's construal of such apparent uses of concept terms in subject place is viable, on his own terms, is another matter. Be that as it may , for Russell the logistic thesis had a deep ontological motive. He thought logicism enabled him to accomodate the truths of arithmetic, within the framework of a correspondence theory of truth and an empiricist account of meaning, without having to recognize any mathematical entities to ground such truths or the meaning of mathematical terms. Logicism thus enabled him to be an empiricist, while grounding the truths of arithmetic in logic, and thus avoiding Platonism, on the one hand, and conventionalism, on the other, as a phi losophy of arithmetic. In the Tractatus, Wittgenstein attacked the logistic thesis in a manner that was to foreshadow the development of his later philosophy. This would interject , in yet another way, a holistic theme into analytic philosophy, since Wittgenstein would emphasize context and the role of the linguistic framework in arguing against meaning as reference. Thus, the importance of the role of the "language game' and "linguistic use' would replace the concern with ontology and, in effect, would be used to reject classical ontology. It is interesting that an extended debate has arisen as to whether the later Wittgenstein's philosophy of mathematics can properly be taken to be a form of extreme conventionalism. So taken, it illustrates, once again, the linking of holism with a form of idealism. ' Russell 's Reduction of Arithmetic to Logic', 'Wittgenstein on Arithmetic and Propositional Form' , and 'Properties, Abstracts, and the Axiom of Infinity' take up these issues and attempt to defend the logistic analysis as offering an adequate ontological basis for accomodating the truths of elementary arithmetic. Taken together, the following essays argue against nominalism and for realism. They defend a correspondence theory of truth and the

=

45

recognition of facts, as entities, against the attacks of Frege, Sellars, Quine, Davidson and others. In doing so they also present an account of thought and intentionality, of reference and predication, and of negation and false belief within a framework of philosophical analysis stemming from the writings of Frege, Russell , and Moore. While written within the analytic tradition and 'the linguistic turn,' the essays focus on the ontological problems posed by a number of fundamental problems in the philosophy of logic and language. The notion of ontology employed is essentially linked to the recognition of facts and a correspondence theory of truth, since a philosophical analysis is taken either to specify the structure and constituents of the facts which comprise the situation analyzed or to show how such purported facts and entities may be avoided . The analyses attempt to remain true to the empiricist's 'sense of reality' that guided Russell and Moore and which led to Russell 's adoption of a principle ofacquaintance as a guide in ontological analysis. Thus, possible worlds, fictional entities, and mathematical objects (along with impossible objects, sets of possible worlds, truth values, functions onto possible worlds, and other bizarre entities) , which are so freely recognized in some contemporary ontological discourse , are avoided as philosophical excesses. Yet, by comparison with the views of some nominalists and materialists, the ontological position argued for in these essays is not sparse. In dealing with ontological reduction , Feigl contrasted ' nothing but' and 'something more· philosophies and Bergmann vividly compared 'desert' and 'jungle' ontological 'landscapes' . These essays reflect a taste for ontological 'elimination · and 'reduction', but they argue that the facts of experience force us to reject the three major reductive trends of our times: universals to particulars, the mental to the physical, and matters of fact and truth to matters of human behavior and language . In short , they reject nominalism , materialism, and idealism.

Notes 1

For a detailed analysis of Moore's early views see my ' Moore 's On tology and Non-Natural Properties' , reprinted in Essays in Ontology, ed. E. Allaire , (The H ague: Martin us Nij hoff, 1 963 ) .

2

G . E . Moore , 'The Nature of J udgmen t ' , Mind, 30 , 1 899 , p . 1 83 . W e can then see why ( l ) may be viewed a s a sentence, stating that ce rtain concepts are combi ned , while representin g an object , a. G . E . Moore , 'The Nature of J udgment ' . p . 1 80 . G . Frege , 'The Thought : A Logical I nquiry ' , t ranslated b y A . M . a n d M . Quinton , repri nted in Essays on Frege, e d . by E . Klemke , (Urbana: University of I l li nois Press , 1 968), p. 5 1 0.

3

4

5

46

6 7

8 9

10 11

I have attempted to do so in my book Thought, Fact and Reference: The Origins and Ontology of Logical Atomism, (Minneapolis: University of Minnesota Press , 1978) . For a detailed critique of Sellars' analysis of intentional contexts, which is the most systematic and careful attempt to present a nominalistic and materialistic analysis, see Thought, Fact and Reference, Chapters 8 and 1 0. S . Kripke , Naming and Necessity (Cambridge , 1980) , pp . 96-97. G. Bergmann, ·Intentionality' , reprinted in Meaning and Existence, (Madison : University of Wisconsin Press, 1960) , p. 32. For a comparison and analysis of the two views see Thought, Fact and Reference. Bradley, by contrast, recognized one fundamental entity , Reality, which accepted (or contained) some complex concepts while rejecting others. Thus, whereas Frege has two fundamental entities , the True and the False, which are denoted by propositions , and Bergmann has actual and possible facts which are intended (meant) by propositional characters, Bradley has only one fundamental ground of truth but two relations between propositional contents and that ground. In this way Frege, a correspondence theorist who acknowledges possible facts , and an absolute idealist like Bradley adhere to a common structural theme.

Frege on Concepts as Functions : a Fundamental Ambiguity

Frege's view that concepts are functions poses a problem of interpretation. Some have treated his doctrine that predicate signs like 'red' are unsaturated signs for functions or concepts as a claim that predicates are only connected with a function or concept, while saturated expressions, such as names an d sentences, have both a referent (an object) and a sense. Hence, if we use the term 'thing' in a broad enough sense to cover such diverse categories as tables, truth values, senses , meanings, propositions, and functions, then predicate signs are taken as being connected with only one thing. Saturated expressions, by contrast, are connected with two things, one of which is an object. Since Frege sometimes speaks of predicates as referring to concepts or functions (as he speaks of names referring to objects and of sentences denoting truth values) , while at other times he speaks of predicates as expressing a concept (as a name expresses a sense or a sentence expresses a thought), a dispute arises as to the status of concepts in Frege 's ontology. Such a dispute arises, in part, due to the belief that if predicates are taken to refer to concepts, then the latter are given more ontological status than if they are expressed by predicates. A second standard way of treating Frege's view is to hold that predicates stand fo r unsaturated things , while names and sentences stand for saturated things , i.e. , objects. The linguistic distinction is thus based on , or at least linked with, an ontological one . A third interpretation involves holding that Frege makes the sense-refer­ ence distinction for predicate terms, as well as for names and sentences. Predicates thus refer to a concept or function and express a sense. One then faces a problem about the difference between the various kinds of senses expressed by the different linguistic expressions (names, predicates, and sentences) and especially that between saturated expressions on the one hand and unsaturated ones on the other. 1 Here I wish to argue that the three types of interpretation are all oversimplified , and , consequently , that the disputes engendered by them are pointless . In seeing why this is so we shall also see how divergent aspects of Frege 's philosophy hang together as well as note how certain unresolved problems of Frege's view become more apparent. Consider a case where we may truly say of an object O that it has the property F. Then we supposedly have, on either the first or second interpretation , five things, in the broad sense of the term 'thing'

48

mentioned above: the object O, the sense of the term ' O' , the truth value T, the thought (proposition) expressed by the sentence 'O is F' , and the function indicated by the term 'F'. But if we consider the role of the function F , we immediately see that this account is incomplete. The funct ion F maps the object O onto the truth value T , and it also maps the sense of the term ' O' onto the proposition expressed by the sentence 'O is F'. In short, there are two functions involved. One function, hereafter F 1 , operates on things in the domain of objects (including truth values) ; the other , hereafter F.,, operates on things in the domain of senses and propositions . Hence , a predicate like 'F' is also connected with two things , just as a name, which has a referent, and a sentence are connected with two things . Moreover , just as in these latter cases , one of the things the sign is associated with belongs to the domain of objects, since it maps objects onto objects , while the other belongs to the realm of senses or meanings , since it maps senses onto thoughts. We may even speak of the sign 'F' referring to or denoting the one function, F 1 , and expressing the other , F.,. The key point about function terms li ke 'F' is neither that they are connected with only one thing, a meaning, and do not refer nor that they refer to functions but do not express a meaning. Function terms may be taken to both refer to and express functions, and each such term is connected to two functions. For, it will not do to resolve the terminological difficulty by holding that a term li ke "F' both refers to and expresses a function , i . e. , that F 1 is the same as F2 , since we would then have one function with two distinct roles and two distinct sets of arguments and values . The crucial distinction of Frege's scheme is then not between objects, on the one hand, and senses, meanings , concepts, and thoughts, on the other. but between functions and all these other kinds. To call such a distinction 'crucial ' is not to deny either the distinction between objects and senses or its importance ; it is merely to point out that the analysis of the difference between subject terms and predicate terms in a subject-pre­ dicate sentence, and the resulting account of predication that Frege suggests, depends on the distinction between functions and other things. It cannot be accounted for merely in terms of the distinction between objects and meanings and the corresponding distinction between sense and reference. In view of Frege's use of the notions saturated and unsaturated, we might then call both functions connected to a predicate unsaturated things and the predicate an unsaturated sign, while objects like 0, the True , and the False as well as the sense of ' O' and the thoughts expressed by sentences are all saturated things. Correspondingly , names and sentences are saturated signs. Frege has not, so far as I know, put his doctrine of functions explicitly in the form I have outlined above. What I am claiming is that his

49

discussion of functions and their difference from obj ects and thoughts entails the association of two functions with each predicate term. Moreover, there are reasons why Frege may not have realized this. He is generally quite clear about the difference between a name, its sense, and its referent , as well as about the difference between a sentence, the thought it expresses and its truth value. This clarity is usually evident when he discusses functions and function signs. But he sometimes confuses a function with the sign for it : Suppose that a simple or complex symbol occurs in one or more places in an expression (whose content need not be a possible content of j udgment . ) If we imagine this symbol as replaceable by another (the same one each time) at one or more of its occurrences, then the part of the expression that shows itself invariant under such replacement is called the function ; and the replaceable part the argument of the function . 2

Occasionally confusing a function with the sign for it could well lead one to overlook the fact that two functions need be associated with a function sign, for there is only one such sign. A far more significant reason for his overlooking the need for two functions for each predicate could be his taking mathematical functions as a sort of prototype of a function and treating sentential ones as basically similar to such prototypes. In fact Frege defines a concept as a function which takes truth values for its values. 3 If we consider a function like is the square ofwe may well take it to correlate numbers, like 2 and 4 , but we might not take it , as a mathematical function , to be something which maps the sense of "2' onto the sense of '4'. Even in the case of relations like is the father of, where we might think of the function (relation) as mapping individuals onto individuals, so that John and Henry are paired if John is the father of Henry, we would find it awkward to speak of mapping the sense of · John' onto the sense of 'Henry'. It is only when sentential functions are considered that the distinction sensibly emerges. Then , when we consider is the father ofas a function correlating the pair of obj ects Henry and John to the truth value the True, we can also speak of the function which correlates the senses of the terms to the thought that John is the father of Henry. Treating is th e father of as correlating John to Henry is not to treat that relation as a predicative function in Frege 's style, for , as a relation , it is a function which takes the obj ects John and Henry as arguments and truth values as values. Yet , just as mathematical functions correlate numbers , there is an obvious sense in which is the father of correlates objects other than truth values. In a way, then , there is an inherent ambiguity in the notion that the phrase "is the father of stands for a function. And , it is precisely this ambiguity which could prevent one from seeing the dual role of predicative functions. For , as a correlator of

50

J oh n and Henry, the relation is th e father of does not correlate the senses of the terms ·John· and · Henry' . Yet, on Frege's vi ew , the relation is really the function correla ting the pai r J ohn and Henry to the True. Hence, he may well overlook the fact tha t having so construed the relation, he has also i ntroduced another function which maps the senses of the two terms onto the thought ex pressed by the sentence ·John i s the father of Hen ry'. Aside from the compli cati on introduced by relational functi ons, thinking of nonrelational predicative functi ons, li ke is-red on the model of mathematical functions, li ke is the square of, can lead one to take a term for the former to un equivocally stand for one functi on. One may then treat such a function as either connecting an object to a truth value or a sense to a thought, depending on the context, without realizi ng that two functions are i n volved , since only one i s i nvolved in the ari thmetical case. A termi nological matter might also aid in obscuring the fact that a predi cate i s connected wi th two functi ons. Speaking of concepts along Fregean lines, one says that i ndivi dual objects 'fall under' con cepts. An object O falls under a concept i ndicated by the predi cate 'F' i f and only if the sentence ·o i s F' refers to the True. To speak of an object falling und er a concept can then he taken to be another way of saying that the function F correlates the object O to the True. Moreover, one does not characteristically speak of the sense of the term "O' falling under a concept. Hen ce, the use of the phrase 'falling under' as an alternative way of expressing the idea of a fun ction mapping something onto something else can lead one, if he thi nks of concepts as functi ons, to explicitly consider only the function whi ch correlates objects and truth values. The dual nature of the predi cative function becomes even more obvious when we recall that Frege thinks of negation as he does of predi cative functions: The thought does not , by its make-up, stand in any need of completion ; it is self­ sufficient . Negation on the other hand needs to be completed by a thought . 4

Just as predicates li ke 'is-red ' indicate something incomplete or unsaturated so does the ex pressi on 'not' or ' the negation of . . . ' . In the case of the negati on sign, or any truth functi on, we clearly have two function s i nvolved in Frege' s di scussion: there is the function whi ch coordinates one though t or propositi on to another and the function which coordina tes one truth value to an other. This corresponds to the correlation of a sense to a thought by F2 and the correla ti on of the object 0 to the True by F 1 • • . The compari son with n egation not only helps make more explicit the dual role of predi cati ve functions, it also helps to reveal a further

51

ambiguity and fundamental problem involved in Frege's notion of a concept as a function. Consider the role of negation in the realm of senses. It is a function which coordinates one thought or proposition to another . Yet it is also a part or constituent of a thought, for the negative proposition is taken to have the positive proposition and negation as constituents. Consequently the thought that contradicts another thought appears as made up of that thought and negation . (I do not mean by this, the act of denial ) . But the words ' made up of , 'consist of' , 'component' , 'part' may lead to our looking at it the wrong way . If we choose to speak of parts in this connexion , all the same these parts are not mutually independent in the way that we are elsewhere used to find when we h ave parts of a whole . . . . The two components , if we choose to employ this expression , are quite different in kind and contribute quite differently towards the formation of the whole . One completes, the other is complete d . 5

Thus the term 'not' stands for a function which maps propositions onto propositions and it also stands for a component of one of the propositions involved in the mapping. Negation is an unsaturated "thing' (and the sign for negation is an unsaturated sign) in two senses. It is unsaturated in one sense since it is a function which maps one saturated thing , a proposition , onto another. It is unsaturated in another sense since it is an incomplete component of a proposition. It is an incomplete component of a proposition in that the other component is true or false in itself, while negation is not ; but it becomes part of something which is true or false when it joins with or is completed by a proposition. Thus the whole proposition , of which negation is a part , is fundamentally like the other component of the negative proposition - the positive proposition - and fundamentally different from the concept of negation. Insofar as one_ takes negation to map one proposition onto another , it is a function: insofar as one thinks of it as an incomplete part of a proposition, it is a concept. This dual role or double sense of 'unsaturated' is there in addition to the unproblematic double use whereby both a sign and what it stands for are spoken of as unsaturated. It is also quite distinct from the double role the negation sign plays as a function sign which stands for a function mapping propositions onto propositions and for a function mapping truth values onto truth values . Perhaps this latter ambiguity is a cause of the former. Truth values , as objects, do not appear to have constituents. As a function mapping truth values onto truth values , negation cannot then be confused with a part of a truth value. Propositions have components . Hence, negation, as a function which maps a proposition onto another may be confused with negation as an unsaturated component of a proposition, and this may well be aided by not having been clear to start with about the difference between negation 52

as a function correlating truth values and as a fun ction correlating propositions. The same ambiguities mar Frege ' s attempt to deal with the problems of predication. We can see why if we first consider a traditional alternative to the way Frege speaks. Suppose one holds that the sentence 'O is F' indicates a fact when true and that he takes that fact to be analyzable into a particular, a universal property, and an asymmetrical relation of exemplification holding between the particular and the universal. In short, he treats the term "O' as designating the particular, the term 'F' as referring to the un iversal. and the predicative " is', together with the grammatical arrangement of the terms, as standing for the relation of exemplification. He may then speak of O and F as parts or constituents of the fact, j ust as the signs "O' and "F are parts of the sentence 'O is F' . Such an account may seem misguided and problematic since properties and obj ects, on the one hand , and names and predicates, on the other, are treated too much alike in that we take both names and predicates to refer to or labe 1 objects and properties respectively. This obscures the distinct roles names and predicates have in language and leads one to think of properties as special objects, namely, universals. Moreover, speaking of a relation of exemplification is both mysterious and problematic. It is problematic in that it invites an infinite regress in the fashion associated with B radley' s name. It is mysterious in that it is generally held to be u nlike al l other relations and is sometimes spoken of as being a tie or nexus in order to emphasize the difference. One might seek to avoid such probl ems by holding that the name ' O' refers to an obj ect or particular and the sentence refers to a fact, but the predicate "F' does not refer to a constituent of the fact. Rather than holding that the pred icate refers to any obj ect or any special constituent of a fact that is not an object, one can suggest that it stands for a function, which connects objects with facts or maps them onto facts. This treatment of a predicate not only reflects and explains the fundamental difference between names and predicates, but it avoids the problems associated with speakin g of a relation of exemplification between constituents of the fact and of entities which are universals. The proponent of such an acco unt may still speak of the object as a constituent of the fact, but if he does then he must hold that the fact is more than the obj ect and he is then obliged to specify how the fact differs from its constituent obj ect. On e attempt to deal with this question would involve holding that the function in dicated by the pred icate is also part of the fact but, being an incomplete part, no relation of exemplification is required to join it to the obj ect in the fact. Alternatively, on e may hold that the function is not part of the fact and that while the fact has the obj ect as its only con stituent it is somehow more than that constitu ent. Finally, one could hold that the object is not a constituent of the fact, but that the 53

fact and the object are things connected by the function. Here I am not concerned with the details of such gambits nor with their cogency or failings or intelligibility. Instead, I merely wish to call attention to the fact that on the first gambit, which holds the function to be a part of the fact, the function plays a dual role that is parallel to the case of negation which we considered above. On the one hand the function maps an object onto a fact ; on the other hand it joins with the object as a constituent, albeit an incomplete one, of the fact. In its first role it is like the function square of which correlates numbers but which is not, in any reasonable sense, a part of a number. In its second role, it functions like a universal which is held to be a constituent of a fact and not a function mapping things onto facts. Frege takes the sentence to indicate or stand for a truth value. Hence , just as he does not have a problem about the analysis of facts , he does not face the question of the dual role of the function which maps objects onto the True. In one move , by introducing functions, Frege removes all of the problems raised by having universal objects connected to particular objects to constitute facts. Not being objects, his predicative functions need not be connected to other objects in a fact. They map ordinary objects onto the two distinctive objects, the True and the False. Since he has such distinctive objects, and not facts, he need not consider the role a predicate plays as standing for a constituent of a fact. He need only consider its role as a correlator of objects to fruth values. In short , since he does not have facts, Frege seems to escape from the problems associated with predication and the ambiguity caused by the dual role his functions play, if we confine ourselves to the realm of objects . But , as Frege recognizes propositions or thoughts , corresponding problems arise in the realm of senses. Predicative functions , recall , also map senses onto propositions. In this connection, Frege speaks of a concept as a part of a thought or proposition . Expressions for concepts can b e s o constructed that ma rks o f a concept are given b y adj ective clauses as , i n our example , b y the clause 'which i s smaller than O' . It is evident that such an adj ective clause cannot have a thought as sense or a truth value as reference , any more than the noun clause could . I ts se nse . which can also be expressed in many cases by a single adj ective , is only a part of a thought . 6

It is also clear that Frege is concerned about the problem of the predicative tie in connection with propositions . For not all parts of a thought can be complete ; at least one must be · unsaturated' . or predicative ; otherwise they would not hold together. For example , the sense of the ph rase 'the number 2' does not hold together with that of the expression 'the concept prime num ber' without a link . We apply such a link in the sentence 'the number 2 falls unde r the concept prime number' ; it is contained in the words 'falls

54

unde r' , which need to be completed in two ways - by a subj ect and an accu sative ; and only because their sense is thus · unsaturated' are they capab le of serving as a l i nk . 7

He thus solves the problem about the nature of the predicative link by having two distinct kinds of constituents of propositions : satu­ rated senses and unsaturated concepts. A predicate (relational or nonrelational ) stands for such an unsaturated constituent. But, in so solving the problem of predication , Frege abandons the idea of a concept as a function correlating saturated things to saturated things or is forced to consider concepts ambiguously: once as such functions, once as constituents of propositions. The last quoted passage clearly brings out Frege's way of being concerned with Bradley's regress and his attempt at a solution. If we think of the problem as, first , involving a query as to what connects the constituents of the proposition into a complex entity and, second , as rej ecting a further constituent to furnish the tie since it will have to be connected in turn to the other constituents, then Frege's solution is simply to have the tie furnished by one of the constituents being unsaturated . This can be accomplished by taking the concept F to be unsaturated, and one does not then require a special unsaturated relation to bind F to the sense of "O' in the proposition expressed by 'O is F'. Even though he speaks of the concept expressed by the phrase 'falls under' in a manner similar to the way some have spoken of the special relation or tie of exemplification, his point is that an unsaturated concept is needed and, moreover , "falls under' stands for a concept just as an expression like "is the father of does. What is crucial is not that "falls under' stands for a special nexus or tie but that it stands for a concept or unsaturated 'entity' . Hence, due to such an unsaturated constituent the various components involved hold together in the proposition expressed by the sentence 'the number 2 fal ls under the concept prime number , just as the relevant senses and concepts hold together in the proposition expressed by 'John is the father of Henry' due to the unsaturated constituent denoted by 'is the father of . Frege's belief that the elements of a proposition cannot be linked, if they are all saturated, corresponds to Bradley's claim that if the relation of exemplification is a distinct entity from the terms it relates, it must, in turn , be related to them by a further relation. It is thus the solution to the problem of predication which forces Frege to take concepts as constituents of propositions and , hence, not as functions. Perhaps this dual use of the notion of a concept by Frege, as a constituent of a proposition and as a function mapping senses onto propositions, may help to explain why the earlier noted double role of predicate signs, as signs for two functions (like F 1 and F2 ) , is not explicit in his discussion.

55

Thinking of the predicate as a sign for an incomplete constituent o f a proposition, one would naturally not take it as a sign for two distinct functions. As an incomplete constituent of a proposition the concept a predicate sign stands fo r is unique. In fact when he considers the concept as such an incomplete constituent of propositions Frege does not seem to think of it as a function at all. The idea of a function comes in when he speaks of the domain of obj ects, of things like O and the True. Part of the problem of interpreting Frege stems from the fact that he sometimes speaks as if the concept, which is a constituent of the proposition, is one and the same thing as the function which correlates an obj ect to a truth value, whereas some other things he says would lead one to hold that he recognizes two entities and takes one to be the sense and the other the referent of a predicate like F. The constituent of the proposition would then be the sense while the function F 1 would be the referent of the predicate 'F'. Frege never seems to explicitly recognize the function F2 . It is obvious why he would not. Since a proposition is a complex of constituents, one of which is a concept, he thinks of pred icates as standing for or expressing such entities, when he discusses propositions. Since a Truth value does not have constituents, as a proposition does, there is not, in Frege's system, any complex in which F 1 may be a constituent. Hence, when he talks of obj ects and truth values, Frege thinks of the predicate 'F' as standing for the function F 1 , since there is no incomplete constituent to confuse with the function. Thus, talk of functions goes naturally with discussions of the realm of obj ects, while talk of incomplete concepts goes j ust as naturally with the discussions of senses and propositions. But the two distinct senses of " unsaturated' which result - the sense in which functions are unsaturated (as requiring completion by arguments to determine values) and the sense in which concepts are incomplete constituents of propositions - are consequently never clearly distinguished by Frege. To clearly distinguish them would involve the explicit recognition of F 1 , F2 , and the incomplete constituent of propositions which 'F' also denotes8 . All this shows why the problem is more transparent in the case of negation than in the case of a concept like F. For in the case of negation, one j ust as readily speaks, in Frege 's terms, of a truth function and of a propositional function. Yet Frege also speaks, as we saw, of negation as an incomplete constituent of propositions. Frege's analysis faces two further problem s. Dum mett has held that the sense of a pred icate term is not a function or concept, but should be an ·obj ect' of the same kind as the sense of a si ngular term. 9 However , such an obj ect will not do as a constituent of Fregea n propositions or thoughts expressed by sentences like 'Fa· . For as an obj ect, the sense of ' F' ( or · Fx') is co mplete or saturated, and a proposition cannot consist of only complete ·obj ects'. Thus, on Dummetf s ·emendation' of Frege's view, 56

one must end up with four entities associated with the predicate 'F' : the functions F 1 and F2 , the concept F, and the sense of the predicate "F' . Frege ' s view also faces a problem i n dealing with so-called oblique contexts . On Frege ' s view , in the sentence

(S 1 )

George IV believes that Scott is an author

the sentence (S i )

Scott is an author

does not stand for the True but for the proposition the sentence (S 2) normally expresses . Thus, one cannot replace (S 2 ) by (SJ

Scott is Scott

in (S 1 ) to obtain

(S 4 )

George IV believes that Scott is Scott

since (S 3 ) does not express the proposition expressed by (S 2 ) . Similarly o ne cannot replace the name 'Scott' in (S 4) by 'the author of Waverley' _ smce (S 5 )

Scott is the author of Waverley

does not express the same proposition as (S i ) . (S,) and (S i ) do not express the same proposition since the senses o-f 'Scott' and- "the author of Waverley' are different . This means that , since the referent of a complex expression is determined by the referents of its constituent expressions, the term "Scott' in (S 1 ) refers to its normal sense. Hence , questions arise about the sense of ( S 2 ) , as it occurs in (S 1 ) , and about the senses of 'Scott' and 'is an author' as they occur in (S 1 ) . If the sense of the sentence (S) , as i t occurs in (S 1 ) , is not a proposition , as Frege suggests at one place , then, on Frege's pattern , in the sentence (S 6 )

I believe that George IV believes that Scott is an author

a problem arises . On one theme of Frege 's, the sentence ( S 2 ) , as it occurs in (S 6 ) , should refer to the sense that the sentence has as it occurs in (S 1 ) . If that is not a proposition , then (S 2 ) does not refer to a proposition as it occurs in (S 6 ) . This is awkward on several counts . First , in some cases sentences will neither refer to truth values nor to propositions . Second , 57

the proposition expressed by (S2) is not a constituent of the propositions expressed by either (S 1 ) or (S6 ) . Third, a constituent of the proposition expressed by (S 1 ) will be a sense that denotes the proposition expressed by (S2 ) . Call such a sense 'SP' , and let 'p' refer to the proposition expressed by (S 2 ) . A constituent of the proposition expressed by (S6 ) will be a sense that denotes the proposition expressed by (S 1 ) . Such a sense will contain a constituent sense that denotes SP. Call this latter sense ' SSP'. Constituents of SSP will denote constituents of SP , which in turn will contain constituents that denote the constituents of the proposition expressed by (S2) - the concept is an author and the sense of 'Scott'. But, then , it is as if Frege has several 'sense functions' that serve to generate senses of senses, as it were, and to combine senses into compound senses. Thus, the constituents of SP, a sense that denotes the sense of'Scott' and a sense that denotes the concept is an author, will be combined by a sense function or concept into SP, a sense that denotes p. A similiar situation arises in the case of SSP. Such a pattern not only involves us in the ad hoc and arbitrary generation of an infi nity of entities from every sense, but another problem emerges that Russell noted. 1 0 It is as if we have a sense function which generates a sense which denotes another sense that is the argument for the function. Thus , with SP as argument we get a sense that denotes SP and which is determined by the function "sense of. But this would mean that the denotation of a sense determines that sense - which, as we shall see, Russell correctly argued that it could not. 1 1 Equally problematic is the function that would serve to compound senses like the sense of the sense of 'Scott' and a sense denoting the concept is an author into a compound sense , SP. Such a sense function must not only serve to 'combine' senses , as it were, into compound senses but indicate a predicative con nection so that the compound sense denotes a proposition. If we consider relational propositions . rather than monadic propositions like p, we can see further problems about order emerging. If. in (S6 ) , (S2 ) is taken to refer to the proposition it normally expresses. while sentences are held to always express propositions rather than 'other senses' , then (S 2 ) will express and refer to the same thing. On Frege·s pattern, this means that the proposition expressed by (S 1 ) will denote itself as well as a truth value. For . recall, Fregean propositions and senses , as well as phrases and sentences . denote. Then, not only will expressions shift reference in oblique contexts, but there will be , so to speak, a shift in the denotation relation among entities . Hence , no longer will one have a uni vocal determi nation of reference on Frege's view. For , either the propositional entities denote more than one "thing' or they, li ke words, shi ft reference. It is difficult to see what the latter claim could mean , and the former abandons a fundamental Fregean theme. 58

Notes 1

3 4

6 7 8

9 1 11 11

These interpretations are represented by a number of articles in Essays on Frege , ed. by E. D. Klemke, (Urbana : University of Illinois Press, 1968 ) . G . Frege, Translations from the philosophical writings of Gottlob Frege, ed . by P. Geach and M. Black , (Basil Blackwell, 1966), p. 1 3. Ibid. , p. 1 5 5 . Ibid. , p. 1 3 1 . Ibid. , pp. 1 3 1-32. Ibid. , p. 7 1 . Ibid. , p. 54. In 'Frege on sense-functions·. reprinted in Klemke , op. cit. , pp . 37er38 1 , H. Jackson cites passages from unpublished manuscripts to support the claim that Frege explicitly recognized the two functions F 1 and F2 . At most the passages show that Frege sometimes speaks as if the sense of a function sign is a function, which is also apparent from the published material . The question is whether Frege clearly distinguished F 1 and F2 • Jackson's argument is not convincing, especially as he, following Frege, confuses F2 with the incomplete constituent of a proposition. Of course one could say 'they' are really one and the same . j ust as one could say that 'they' are identical with F 1 • Whether one chooses to speak of one entity with three roles or of three entities, each with a unique role, is not worth disputing about . The point is to note Frege's three-fold use of the notion of a concept. That Frege j umbled, rather than distinguished, the three entities ( or roles) is neatly revealed by his definition of a concept as a fu nction which takes truth values for its values ( note 3 above). M. Dummett, Frege: Philosophy of Language, (London : Harper & Row , 1973) , pp . 26869, 294 . See 'Russell's Attack on Frege's Theory of Meaning' , in this volume , pp. 60 f. In 'Russell's Attack on Frege's Theory of Meaning' . The point is that we do not have a genuine function since many senses may have the same denotation. But , to arrive at a genuine function by taking the argument to be the sign 'SP' , rather than the sense SP , makes the function "linguistic" in an obvious way. This points to a problem with Frege's analysis that is central to Russell's critique , as we shall see.

Russell's Attack on Frege's Theory of Meaning

Russell's attack on Frege's theory of meaning occurs in one of the most puzzling passages in the 1 905 paper that introduced his theory of descriptions. The passage has been the occasion for much commentary and abuse. According to Searle 1 , Russell's discussion is unclear, obscure, confusing, slipshod, either superfluous or irrelevant to Frege's argument, inconsistent, involving a misstatement of Frege's position 'combined with a persistent confusion between the notions of occurring as a part of a proposition . . . and being referred to by a proposition' 2 , nonsensical, and, finally, Russel l mistakenly states the negation of Frege 's theory instead of Frege's theory. 3 Geach agrees that Searle has shown the irrelevance of Russell's criticism of Frege and explains it as an excusable confusion, on Russell's part, between Frege's theory and Russell's earlier views in The Principles of Mathematics. 4 Thus Russell, according to Geach, wasn't really arguing against Frege. He was led to think he was since 'Russell, like Aristotle, so often distorts others' thought into his own mould . . . '5 Geach advises the reader of "On Denoting' to ignore the ·use of Frege 's name'. To Church, Russell is merely confused about use and mention. 6 Carnap, politely and correctly , finds Russel l obscure. 7 Chrystine Cassin, in an article pleasantly free from rhetorical abuse , develops Geach's suggestion and holds that Russell was criticizing his own earlier view and that there is 'no direct textual support'8 for the claim that Russell was arguing against Frege's view. According to Cassin, the sole textual basis for treating Russell's arguments as criticisms of Frege would be Russell's use of 'meaning and denotation' and the assumption that such use corresponds to Frege 's use of ·sense and reference'. 9 Cassin is thus gentler with Russell than are Searle, Geach, Church and others. While Geach seems to believe that Russell distorted Frege's thought, thus confusing Frege's views with his own, Cassin doubts that Russell was intending to argue against Frege's views. Hence, as she sees it, Russell can be saved from the charges of not understanding Frege or of distorting Frege's views. I do not know what Cassin would count as direct textual support but there is abundant textual support to show that Russell believes that he is disputing Frege's views. A few pages before the passage that concerns us he speaks of Frege 's theory involving the distinction between meaning and denotation. 10 The footnote is to the now celebrated paper generally

60

translated as 'On Sense and Reference' . A second footnote states that: Frege distinguishes the two elements of meaning and denotation everywhere , and not only in complex denoting phrases . Thus it is the meanings of the constituents of a denoting complex that enter into its meaning , not their denotation . In the proposition ' Mont Blanc is over 1 .000 metres high' , it is , according to him , the meaning of ' Mont Blanc' , not the actual mountain , that is a constituent of the meaning of the proposition . 1 1

and a third footnote, immediately following, reads: In this theory , we shall say that the denoting phrase expresses a meaning ; and we shall say both of the phrase and of the meaning that they denote a denotation . In the other theory , which I advocate , there is no meaning , and only sometimes a denotation . 1 2

I think that it is perfectly clear that Russell is taking the notions of meaning and denotation to be those involved in Frege' s paper, irrespective of any differences between his own earlier notion of a denoting complex and a Fregean sense or of his own earlier notion of a proposition and a Fregean proposition or thought. The attempt to save Russell from criticism like Geach' s and Searle' s by holding that he didn't intend to argue against Frege' s view is simply not plausible. Thus, if one believes that the view Russell is arguing against is not Frege' s, he will have to hold that Russell either misunderstood Frege or distorted , perhaps knowingly, Frege' s view. I shall argue that, properly understood , Russell ' s arguments are not only directed against Frege' s view but are fatal to it. That is, I shall attempt to establish the astounding and unfashionable claims that (A) Russell understood Frege; (B) Russell also understood his own earlier view and how it was related to Frege' s view; (C) Russell's arguments are, as Russell clearly believed them to be, directed against Frege; (D) Russell's statement of Frege's view is correct and neither a distortion nor the result of a confusion; and , finally, (E) Russell' s arguments are cogent. From the opening sentence of the supposedly unintelligible discussion it is clear that Russell is concerned with the connection between the meaning and the denotation on a view which recognizes both sorts of entities. As he puts it: The relation of the meaning to the denotation involves certain rather curious difficulties, which seem in themselves sufficient to prove that the theory which leads to such difficulties must be wrong. 1 3

This delineates quite clearly the target of his argument. Russel pro­ ceeds to state that a natural way of talking about the meaning of a de­ noting phrase is to put the phrase in inverted commas, whereas we de-

61

note the denotation by a straightforward use of the phrase. Thus 1) The O is U . 2) 'The O' is a meaning or a denoting complex and not a U . or, to use his example, 1 ') The first line of Gray's Elegy states a proposition. 2') 'The first line of Gray's Elegy' does not state a proposition. All {2') says is that the meaning of the phrase 'the first line of Gray's Elegy' does not state a proposition, as meanings ofphrases are not the sort of things that state propositions. The trouble begins with Russell's next sentence. He writes: Thus taking any denoting phrase , say C, we wish to consider the relation between C and 'C' , where the difference of the two is of the kind exemplifi e d in the above two instances . 1 4

A contemporary reader may be led to think that when Russell speaks of the relation between C and'C' he is talking about the relation between a phrase and something else. He is not doing that ; he is raising the question I mentioned above - a question about the relation between the denotation and the meaning of the phrase. Keeping this simple point in mind will help us to unravel and follow the argument. The modern reader would no doubt be helped if Russell had put the first occurrence of the letter C in inverted commas or double quotes. But note two things; first, he naturally would not put it in inverted commas since he has just cited the convention whereby the putting of a phrase in commas forms a sign that denotes the meaning of the original phrase and not the phrase. Second, he need not do so to be clear enough, just as I did not do so in the above sentence when I spoke of 'the letter C'.If he had introduced the use of double quotes to speak about the phrase, the passage would run as follows: Thus taking any denoting phrase , say '•C" , we wish to consider the relation between C and 'C' , where the . . .

But this wouldn't help in that he is using the letter ·c· as a variable and does not want to talk about the letter, but about any denoting phrase . In short, part of the problem is precisely the kind of difficulty associated with Quine's use of corner quotes and with a problem that led to Tarski's results in connection with the definition of a truth predicate. The simplest way of putting, it, given Russell's concern, is the way he does put it, assuming that the reader have, not charity, but understanding. 62

Russell then reiterates that when sentences like ( 1) and ( 1 ') are used we are speaking about the denotation and when sentences like (2) and (2') occur we are speaking about the meaning. He then states that . . . the relation of me aning and denotation is not merely linguistic through the phrase : t he re must be a l ogical relation involve d , which we express by saying that the meaning denotes the de notation . 1 5

He follows this assertion with a statement of the two-fold conclusion of his argument. But the difficulty which confronts us is that we cannot succeed in both preserving the connexion of meaning and denotation and preventing them from being one an d the same ; also that the me aning cannot b e got at except by means of denoting phrases . 1 6

Before looking at his argument, it behooves us to attempt to be clear about what it is he wishes to prove. On the view he opposes there is , associated with the phrase 'C', a meaning, which I will refer to as MC, and an object or denotation, which I will refer to as DC. Russell thinks there are two problems with such a view. The first problem is that there must be a relation between MC and DC that requires explanation and it is not explained by speaking of the merely linguistic relations between the phrase 'C', on the one hand, and the entities, MC and DC, on the other hand . That is, suppose we hold that (I)

"C' denotes DC

and (II)

'C' means MC

with the letter C in quotes, as it occurs in (I) and (II), standing for the phrase. We also have (III) R (MC, DC) . If we now ask what relation does the sign 'R' stand for in (III), one response might be that (III) is elliptical for the conjunction of (I) and (II). In short, a meaning, MC, and a denotation, DC, stand in R if and only if there is a denoting phrase which both means MC and denotes DC. Call this response or claim (F). That (F) will not do is one thing Russell wishes to establish. It should be clear that what I mean by saying that (F) will not do is that it will not do as a proposed explication or analysis of (III). It should also be clear why it is that Russell speaks of (F) as

63

establishing a 'merely linguistic' relation between MC and DC. A relation between two nonlinguistic items is merely linguistic if the predicate indicating the relation is defined in terms of relational predicates standing for relations between a linguistic item and a non­ linguistic item.Thus, if R (x,y) = df (3z) (z means x & z denotes y) provides a definition of 'R', then R is a merely linguistic relation, since the conjuncts in the defining phrase employ only relations between signs and things, and not relations between non-linguistic items.Similarly, we may say that an entity has been specified in a 'merely linguistic' way if it is specified solely by denoting phrases mentioning a sign such as "the meaning of 'C"' and "the denotation of 'C "' . It should already be evident that the comments of some commentators as to the ineptness of Russell's understanding of Frege or as to the irrelevance of Russell's discussion to the views of Frege are gratuitous. Whatever else Frege's view involves, it certainly involves a relation between the meaning or sense of a denoting phrase and the denotation of the phrase. In fact, as Russell sees it, any view that recognizes a meaning (or something playing the role of a meaning) and a denotation, which differs from the meaning, must recognize such a relation. In this essential respect his own earlier view can be taken as similar to Frege's. What he will argue is that one can only give an inadequate account of that relation and, hence, any such view must be inadequate. Consequently, both Frege's view and Russell's earlier view must be inadequate. Russell, then, is clearly arguing against Frege 's view, as he himself explicitly said and obviously thought.Insofar as we have grasped merely the initial phase of Russell's argument, as set out above, we understand that to dismiss Russell as not understanding Frege, or as distorting Frege, or to save Russell by claiming that he does not intend to argue against Frege, but against himself , or to claim that his argument is unintelligible reveals a failure to comprehend a basic and elementary aspect of Russell's argument : namely , what it is he argues against. To repeat, since , in view of what has been written, it bears repetition , the argument is that any view that recognizes both a meaning and a denotation for denoting phrases must account for the relation between the meaning and denotation and that (F) does not furnish an adequate account, since (F) establishes a merely linguistic relation between the two entities, MC and DC. Let it be clear that I have not, so far, considered Russell's argument, but merely one thesis he wants to argue for: an account of R must be given and (F) is not an adequate account.What I have said is not that his critics have not understood his arguments, but that they have not understood

64

this basic thesis that he wishes to argue for, though , in due course , I will assert that they also have not understood the specific arguments he proceeds to give. Setting it out the way I have also suggests what one of Russell's lines of argument is. A Fregean type view involves a relation like R. An account of R like (F) will not do, but no other account is feasible. Hence a Fregean type view will not do. More generally, the argument is that on any view which recognizes something like MC, as well as something like DC, where there is a denoting phrase , C, there must also be a relation, R, between MC and DC. This relation will either be merely linguistic or not. It is not cogent to hold that it is merely linguistic. To hold that it is not is doomed to failure . Therefore , no view recognizing MC, in addition to DC , will do. This is Russell's approach. It is obscured since Russell almost takes it for granted that if R is merely linguistic the view introducing MC is inadequate. We must then fill out why he thinks this. But, since he does think this is so , we can understand why he tries to show that a Fregean style view must take R as merely linguistic by arguing that attempting to avoid taking it so introduces unpalatable consequences. His detailed arguments thus attempt to show that if R is not merely linguistic, the re is an insoluble problem for the view he attacks. The Fregean must retreat to the inadequacy of a merely linguistic connection between MC and DC. Such a connection is inadequate since MC and DC are entities , on the Fregean view. Therefore, the relation between them ought to hold independently of our assigning signs to entities. Thus, (F) does not constitute an adequate explication of R. To put it anothe r way, a sign which means MC and denotes DC ought to be connected to one of the entities in virtue of a relation that holds between MC and DC. In short, R ought to be involved in the explanation of either the connection between the term and its meaning or the term and its denotation. But this precludes attempting to explicate R by appeal to (F). To explicate "means' or 'expresses' in terms of R and denotes, where the latter relation is taken as the relation between a sign and its denotation, is not feasible . It is not because several meanings can stand in R to one and the same referent. Thus one could not specify MC as the M such that R (M , DC). Consequently, one cannot specify 'means' as standing for a relation holding between a name and a meaning, M, where the name denotes the x such that R ( M , x) . As Russell put it ; . . there is no backward road from denotation to meaning, because every object can be denoted by an infinite number of different denoting phrases . 1 7

The alternative, to explicate denotes in terms of R and means, is also not feasible. It would involve holding, for example, that 65

(I)

'C' denotes DC

is to be understood as elliptical for (IV) ('C' means MC) and R (MC, DC). But (IV) presupposes that we have specified what a meaning, like MC, is and which meaning MC is. The only way we have of doing so is by stating that MC is the meaning expressed or mea nt by the phrase 'C' . Doing this involves two problems. First, we still specify MC in a 'merely linguistic' way, for, MC is specified as the meaning of the sign 'C'. By contrast , whereas we can denote DC by the denoting phrase. "The denotation of the sign 'C' ", this is not the only, nor , indeed, the fundamental way of denoting the object DC. It is, in fact , parasitic on our having ways of connecting the sign 'C' with the object DC without using a phrase like 'the denotation of the sign . . . '. In the case of MC there are no alternatives ; we must specify MC as the meaning of the sign 'C', and hence specify it in a merely linguistic way or 'through the sign', as Russell puts it. It is worth noting that Frege apparently thought it was perfectly all right to specify MC in this way: In order to speak of the sense of an expression 'A' one may simply use the phrase 'the sense of the expression "A' " . 18

If we could specify MC by means of R and the denoting relation between a sign and an object this problem would not arise. It would not arise since we can specify the denotation, say DC, of a sign by other than 'merely linguistic' means ; by the use of denoting phrases that do not indicate the object as the denotation of a sign or phrase. But this alternative is not open since, as Russell put it, we cannot go backward from denotation to meaning. In short , this would return us to the attempt to explicate means in terms of denotes and R. The second problem with the present alternative is that R remains a mysterious relation, as it would on the first alternative ; and there seems to be no hope of explicating it solely in terms of the denoting relation between a sign and its denotation. R must be taken as basic or primitive on the view Russell attacks. But then , just as the meaning MC is a mysterious and parasitic entity , since it is indicated solely as 'the meaning of a sign' , so is the relation R , since it seems to be postulated as a basic relation between the mysterious entity MC and the object DC. All this ought to be evident since we can only speak of MC as the meaning of a phrase and of R as the relation between the meaning and the denotation of a phrase. In a way the point stems from the peculiarity of the relation

66

means or expresses . We have a sign ·c· , a meaning MC, and a relation means holding between them. B ut both MC and the relation are

dependent on the term and specified by reference to a phrase or term. What is MC? It is what the phrase ·c· means. What is the meaning relation? It is what holds between a phrase and its meaning. The case of denotation is different in that, first, we can specify denoted obj ects independently of mentioning phrases which denote them, and, second, such obj ects are often familiar obj ects of experience in an unproblematic sense of · experience'. Meanings appear to be problematic by contrast. But, even if one maintains that he is familiar with such entities and hence holds that means is no more problematic than denotes , he is left with R as a basic and problematic relation, which must be specified 'through the , phrase , i. e. , as holding between a meaning and a denotation when some phrase means the one and denotes the other. Before pursuing the question of the mysterious status of R and MC, it is significant to note that Frege appears to agree with the contention that R and means are involved in the explication of denotes , as a relation between a sign and its denotation. This would seem to be the import of his statement: The regular connection between a sign , its sense , and its reference is of such a kind that to the sign there corresponds a definite sense and to that in turn a definite reference , while to a given reference (an obj ect) there does not belong only a single sign . 1 9

This suggests that Frege takes the referent to be determined by the sense, and hence that he agrees with Russell that R should not be taken as a merely linguistic relation. For, if the sense determines the referent one need not appeal to (F) to explicate R. That portion of Russell' s argument that we have considered so far is, I take it, unquestionably intelligible. Its cogency appears to rest on the force of the claim that R is mysterious, if it is not merely linguistic. One could seek to refute the claim by construing the meaning in such a way that R ceases to be problematic or 'mysterious'. Taking the meaning to be a property which only the denotation has would be one possible initial move. R would then become the exemplification relation between a property and an obj ect. If pursued, however, this gambit would amount to no m ore than the view Russell is developing in connection with his theory of description in 'On Denoting' . Other alternatives were attempted by Moore and Russell at the turn of the century and Frege can be interpreted along similar lines. While I believe all such views fail, I will not attempt to substantiate that claim here, though such arguments clearly are a necessary part of an overall consideration of the issue. The

67

reason I will not take up the issue here is that Russell develops, later in the passage, a different line of argument to establish that R is problematic and, in this paper, I am primarily concerned with unpacking Russell's attack on a Fregean style view. We will postpone for the time, consideration of Russell's argument against R. Before proceeding with Russell's discussion, there is a feature of his use of the term 'mysterious' that requires comment. To hold, as Russell does, that R is mysterious if it is taken as primitive or basic or, perhaps more accurately, if it is taken to be indicated by a primitive term, does not mean that Russell holds that all relations indicated by primitive terms are mysterious. Consider a referring relation as holding between a name and an object, when I stipulate that the name is to refer to or denote that object. Assume that we accept a Russellian type principle of acquaintance to the effect that primitive predicates must stand for (1) a property or relation that is instantiated and (2) a property or relation that is experienced in at least one instantiation by the user of the predicate. A primitive predicate is mysterious if it does not fulfill condition (2) while it is assumed that it fulfills condition ( 1) . R is purportedly mysterious in this sense. Reference , as I used it in the case of the naming of an object just above, is not mysterious. One experiences the relation of referring to just as one perceives the relation ofleft of when one sees one object to the left of another. Let me explain . Moore introduced a relation of direct acquaintance in his analysis of perception. 20 Roughly , such a relation held between a mental act, a perception , and an object. A sense datum would be one example of such an object. He held that he knew what such a relation was since there were other acts of direct acquaintance which experience the relation of direct acquaintance holding between an act and an object. So we have two acts , ma 1 and ma2 , a sense datum, sd 1 , and the relation of direct acquaintance , DA, such that (I) DA (ma 1 , sd 1 ) (11) DA (ma2 , DA) or , perhaps (11 1) DA (ma2 , DA (ma 1 , sd 1 ) ) . (II) would state that ma2 stands in the relation DA to that relation, whereas (III) would state that ma 2 stands in the relation DA to the fact that ma 1 stands in DA to sd 1 • In either case we know what we mean when we speak of a relation of direct acquaintance because such a relation is 'experienced' by acts like ma 2 . In a similar way, we know what we mean when we say '"let 'C' refer to that" on a given occasion . We know that without necessarily being able to give an explication of the term "refer' and without needing to give such an explication. We do not so know what we mean when we say ' R (MC , DC)'. R is parasitic on stipulative reference, but as such is not what it ought to be in the Fregean-style

68

scheme. This is a theme I take to be involved in Russell's rej ection of Frege' s scheme via his rej ection of the "' mysterious" relation R. A second matter I wish to discuss before proceeding to Russell' s argument that R is problematic is Ru ssell's argument that MC cannot be gotten at except by means of denoting phrases of the kind " the meaning of . . . ' , where the dots are replaced by a phrase denoting or mentioning a sign or phrase. This argument purports to establish that we must indicate MC in such a way as to make R parasit ic on reference, as a relation between a sign and its denotation and hence turn R into a merely linguistic relation. In other words to be able to indicate MC in some other way would be to indicate it and R in such a way that "R (MC, DC)' could be expressed as a genuine relation between entities and not as a mere ellipsis for "" "C' means MC and denotes DC" . The point is simple enough, but Russell' s way of putting it requires explication. Russell says: But if we speak of "the meaning of C' , that gives us the meaning if any of the denotation . 2 1

It is clear, in our terms, that the expression he uses is (C 1 )

the meaning of the object C

so that we would be talking about the meaning of DC, the denotation. There would be such a thing only if DC had a 'meaning' in the sense that it stood to something in the same relation that the sign 'C' purportedly stands to MC. So, to get at MC, we cannot use (C1 ) , and , hence, we cannot use the phrase 'the meaning of C . ' Russell con tinues, . . . in order to get the meaning we want , we must speak not of 'the meaning of C' , but of 'the meaning of "C'" , which is the same as 'C' by itself. 22

Here, recall, we indicate the meaning of the phrase 'C' by putting the phrase in inverted commas. So, Russell is quite clear and quite correct. The point is simple. We are forced to indicate the meaning, MC, by speaking of the word s or phrases that , so to speak, mean the meaning. This reduces R, as Russell sees it , to a merely linguistic relation, something that is parasitic on talk about words as opposed to things like MC and DC. Nor should it be puzzling that I have referred to MC all along using the phrase 'MC' , if we remember how that phrase was introduced. What Russell intends is clear enough, but an example he uses in the course of the discussion may be a source of confusion for the reader. He illustrates the point by writing:

69

But if we speak of 'the meaning of C' , that gives us the meaning (if any) of the denotation . 'The meaning of the first line of Gray's Elegy ' is the same as 'The meaning of "The curfew tolls the knell of parting day" ' , and is not the same as 'The meaning of "the first line of Gray's Elegy"' . 23

The problem is posed by his having introduced the inverted commas for a specialized purpose, denoting an entity like MC, and yet he also uses them in a more ordinary way: to mark off a phrase. All he is saying here is that the phrase (A) The meaning of "the first line of Gray' s Elegy" does not refer to the same thing as the phrase (B)

The meaning of the first line of Gray's Elegy

and that the latter phrase does indicate the same thing as the phrase (C)

The meaning of "The curfew tolls the knell of parting day" .

Thus, if we want to refer to the meaning of the phrase The first line of Gray's Elegy we must use (A) or the expression (D )

The first line of Gray's Elegy'

with the special convention about inverted commas. Thus, the example is, as he obviously intends, merely an illustration of his claim about the expression 'the meaning of C' . The next part of Russell's paragraph is not really essential to the discussion for he simply applies the point which he has just made about the expression 'the meaning' to the expression 'the denotation'. But, as an example he uses appears to be responsible for much of the abuse heaped upon him, it is worth explicating the passage.The passage reads as follows: Similarly ' the denotation of C' does not mean the denotation we want, but means something which , if it denotes at all . de notes what is denoted by the denotation we wan t . For example , let 'C' be 'the denoting complex occurring in the second of the above instances' . Then C = 'the fi rst line of Gray's Elegy· , and

70

the denotation of C = The curfew tolls the knell of parting day . B ut what w e meant to have a s the denotation was 'the first line o f Gray's Elegy' . Thus we have failed to get what we wanted. ::! 4

The first sentence of the passage, stating Russell's point, is unproblematic. He is simply stating that the phrase 'the denotation of C' denotes, if it denotes at all , what the object C denotes , for such a phrase refers to the denotation of C, the object , not to the denotation of 'C', the phrase. But , once again , the example he gives complicates the matter, for it also involves a use of inverted commas that is distinct from the announced convention whereby the meaning of the phrase 'C' will be indicated by that phrase in inverted commas. But, it is understandable, for we must recall the sign ·c is not a phrase but a variable . Russell is asking us to consider the "denoting complex' (a) the denotation of C where the occurrence of ' C' is replaced by the phrase 'the denoting complex occurring in the second of the above instances', so we get (/3) the denotation of the denoting complex occurring in the second of the above instances from (a) by the appropriate replacement . Here, it makes no difference whether we think of the denoting complex as a phrase or some sort of entity associated with the phrase ( expressed by the phrase), though we shall return to the double sense of 'denoting complex' shortly. I think , however , that it is quite clear that Russell is here taking the phrase to be the denoting complex . He is simply specifying a definite phrase to replace the letter 'C'. Russell assumes that we want to denote the denoting complex The first line of Gray's Elegy just as, in the immediately preceeding sentences, he assumed we wanted to denote the meaning of 'The first line of Gray's Elegy'. His point here is that we cannot do so by the use of (/3). We cannot because the expression 'the denoting complex occurring in the second of the above instances' denotes 'The first line of Gray's Elegy' and, hence, (/3) denotes the denotation of the denoting complex 'The first line of Gray's Elegy' , and does not denote that denoting complex. Thus, we end up denoting The curfew tolls the knell of parting day

71

and not The first line of Gray's Elegy , just as in the earlier example we ended up referring to the meaning of ' The curfew tolls the knell of parting day' and not to the meaning of 'The first line of Gray' s Elegy ' by using the phrase 'The meaning of the first line of Gray's Elegy' . A ll thi s merely illustrates, in an overly complicated way, that one cannot denote anything, say O, by the use of the phrase 'the denotation of O', since we then denote what the obj ect O denotes and do not denote the obj ect 0. What has caused all the confusion is Russell' s use of the inverted commas to specify which denoting complexes he i s talking about, after he has introduced the inverted commas as a device to indicate the meanings of phrases. B ut, there are several things worth noting: first, it is clear that he is using the commas to emphatically mark off the denoting complexes through the whole paragraph as is evident from his use of a whole series of expressions employing them: 'the meaning of C' 'The meaning of the first line of Gray's Elegy' etc. Second, he has used inverted commas throughout the paper in a variety of ways, for emphasis, to mark off an expression, to speak about an expression, etc. Third , Russell has in a way announced this ambiguous use, for he is treating both meanings and phrases as denoting complexes. Critics who find him confused as to whether a denoting complex is a phrase or an entity might find it helpful to recall what Russell writes earlier in the paper. In this theory , we shall say that the denoting phrase expresses a meaning ; and we shall say both of the phrase and of the meaning that they denote a denotation . 25

Thus, both phrases and meanings are denoting complexes. Moreover, it makes no difference whether one takes hi m to be talking about the phrases, as the denoting complexes, or about the entities associated with the phrases- the meanings. Russell's point applies in both cases. It is only if one takes hi m to be talking about the phrases at some points and about the entities at other points that he appears to be confused . Just as he uses the inverted commas in a number of ways, Russell uses the term ' means' in two ways. Thus, when he says:

72

Similarly 'the denotation of C' does not mean the denotation we want , but means something which , if it denotes at all , denotes what is denoted by the denotation we wan t .

he is stating what could be more cumbersomely, but more clearly stated as: Similarly 'the denotation of C' does not denote the denotation we want , but denotes something which , if it denotes at all , denotes what is denoted by the denotation we want.

He is clearly using ' means' in the sense of • 1 meant this book, not that one' , as he finishes the paragraph with: B ut what we meant to have as the denotation was 'the first line of Gray's Elegy' . Thus we have failed to get what we wanted .

Hence, again, one can accuse him of being confused instead of expending minimal effort to understand what he says. Looked at, as I have suggested, Russell is hardly confused , simply not, here, a master of literary style. The contrast between the simplicity of the line of argument and the unfortunate complexity of style comes to a head in the next passage which is clearly, from the context, meant to sum up what has been said . Russell writes: The difficulty in speaking of the meaning of a denoting complex may be stated thus : The moment we put the complex in a proposition , the proposition is about the denotation ; and ifwe make a proposition in which the subj ect is 'the meaning of C' , then the subj ect is the meaning (if any) of the denotation , which was not intended . 26

If we equate the proposition with the sentence and thus take the denoting complex to be the verbal expression, there is no problem of interpretation. Russell is merely summing up what he has said. A problem is only created if one, who is aware of Russell' s earlier view about denoting complexes, insists on taking him to be speaking of some sort of entity, rather than a linguistic phrase, and of a proposition as what is 'expressed' by a sentence rather than the sentence itself. One can then proceed to poke fun at Russell. But we should note first, that Russell clearly distinguishes between a proposition and its verbal expression and then, as one may easily do, employs the term " proposition" for the verbal expression; and, second, that Russell explicitly states in this paper that both the verbal expressions and the entities expressed by them will be spoken of as denoting the denotation. 73

B ut , even if we take him to be talking about the non-verbal proposition , expressed by the sentence , and the entity expressed by the denoting phrase his point is still clear . Take MC as the denoting complex and 'O(MC) ' to stand for a proposition which contains that complex as a constituent , i . e . , as its subj ect . The propositional entity , O(MC) , is about the denotation of MC and not about MC, j ust as the phrase 'the meaning of MC' is not about MC but about something which MC means. So , we cannot consider a proposition (non-verbal) which contains a meaning, MC, as its subj ect to be about that meaning . We must recall that Russell , i n The Principles of Mathematics, considered denoting complexes or concepts to be constituents of propositions , which were thus about the objects denoted by such complexes . . . . That is to say , when a man occurs in a proposition (e. g. " I met a man in the street") , the proposition is not about the concept a man, but about something quite different , some actual biped denoted by the concept . Thus concepts of this kind h ave meaning in a non-psychological sense . And in this sense , when we say "this is a man" , we are making a proposition in which a concept is in some sense attached to what is not a concept . But when meaning is thus understood , the entity indicated by John does not have meaning , as Mr. Bradley contends [ Logic , Book I, Chap . I , 1 7 , 1 8 (pp . 58-60) . ] ; and even among concepts, i t i s only those that denote that have meaning. The confu sion is largely due , I believe , to the notion that words occur in propositions , which in turn is due to the notion that propositions are essentially mental and are to be ide ntified with cognitions . 2 7

Russell's view in The Principles has puzzled his commentators for a very simple reason . For Russell , when a sentence contains a denoting phrase , such as 'the author of Waverley' , the corresponding proposition contains a denoting complex . However, when the sentence contains a name (a referring as opposed to a denoting expression) the corresponding proposition contains the referent as a constituent . For Russell , at this stage , Scott is a constituent of the proposition expressed by the sentence 'Scott is a man ' : whatever may be an obj ect o f thought , o r may occur i n any true o r false proposition , or can be counted as one, I call a term . This, then , is the widest word in the philosophical vocabulary . I shall use as synonymous with it the words unit , individual , and entity. 28 . . . But a proposition, unless it happens to be linguistic , does not itself contain words : it contains the entities indicated by words . 2 9

Thus , Searle's shocked tone in his discussion of Russell's ' confusion' regarding the constitue nts of propositions may indicate a lack of familiarity , rather than a lack of sympathy , with Russell's earlier views . We can also see how the earlier view will provide a basis for the theory of

74

descriptions, since that theory is designed to illuminate the 'logical' difference between sentences like " Scott is tall' and sentences like 'The author of Waverley is tall' , where the latter, but not the former, contains a denoting phrase. As we will see, one of Russell's basic criticisms of Frege is that Frege's theory does not preserve, let alone explicate, the difference between names and descriptions. It is true that Russell speaks of both the object and the name for it being constituents of propositions. · A proper name, when it occurs in a proposition . . .' , 30 but this is not merely due to his confusing use and mention. He clearly distinguishes, at a number of places, between a proposition as a nonverbal entity and a proposition as a verbal expression of the former. Rather than seeing his view as a result of such a simple confusion it would appear more reasonable to take his view to lead him to write, at places , as if he confuses a sign with a thing.As he later writes: It may be expressed as the distinction between verbs and substantives, or, more correctly. between the obj ects denoted by ve rbs and the obj ects denoted by substantives . (Since this more correct expression is long and cumbrous, I shall ge nerally use the shorter phrase to mean the same thing . Thus , when I speak of verbs . I mean the obj ects denoted by ve rbs . . . 3 1

At the time of The Principles Russell recognized three senses of ·proposition'. Propositions R , as entities, which contain constituent objects, like Scott, and concepts like is human.Propositionsf ' as entities, which contain denoting complexes, or denoting concepts, like the denoting complex the man, as well as concepts like is tall.Propositionsv , as linguistic expressions, which contain words, either names or denoting phrases, predicates, etc. The second sort of propositional entities, propositionsf are very like the Fregean propositions. In "On Denoting' ' Russell is moving toward the elimination of such entities by eliminating denoting complexes as entities. So far, then, a simple result has been argued for. Expressions like 'the meaning of C' and "the denotation of C' cannot be used to denote, respectively, that which is expressed by the term "C' and that which is denoted by the term: the meaning of the term in the first case and the referent in the second. About this Russell is clearly correct. He is really concerned with the first case: to hold that we cannot use the expression "the meaning of C' to denote what the term "C' expresses, on the view he attacks. Recall that, for Russell, to speak of the meaning of the term 'C' is not acceptable, since it involves us in treating the denotation relation R, between a meaning and a denotation, as merely linguistic.We can then understand Russell's bothering to argue that we cannot use the phrase 'the meaning of C' to denote what we want to denote. He is implicitly 75

claiming that since we cannot use either denoting phrase, "the meaning of C' or "the meaning of 'C "' , for different reasons, we have no unproblematic way of denoting MC. What we must do is simply stipulate that by distinguishing meaning from denotation we are, in a way, dealing with the meaning. That is, in holding that a meaning as well as a denotation is connected with a denoting phrase we introduce a way of talking about meaning. This is one point involved in Russell's next statement: This leads us to say that , when we distinguish meaning and denotation , we must be dealing with the meaning : the meaning has denotation and is a complex , and there is not something other than the meaning, which can be called the complex and be said to have both meaning and denotation 32 .

Russell is not conceding that such a move is legitimate. It still involves us in treating R as a merely linguistic relation.But he wants to move on to another argument. Forgetting the supposed inadequate treatment of R, on a view that has MC, DC, and R, we can consider such a view to construe MC as something which denotes DC but which does not itself have both a meaning and a denotation.That is both the phrase 'C' and the meaning, MC, denote, but while the phrase both means and denotes, the meaning only denotes.One kind of denoting complex, a phrase, has both meaning and denotation: the other kind of denoting complex has a denotation and is a meaning . So Russell concludes this section by saying: The right phrase , on the view in question , is that some meanings have denotations. 33

There is second aspect to his claim. On the Fregean view , the connection between a denoting phrase and its denotation is via the meaning or sense of the phrase. The basic denoting relation is between the meaning and the latter's denotation, i. e. , the relation R. What stands in R to a denotation does not also express a meaning.A phrase expresses a meaning but does not stand in R to anything. One may then say that the phrase is directly connected only with the meaning. Hence, if we introduce meanings on such a pattern, a phrase, by itself, determines only a meaning, which in turn denotes.Given the phrases, we are 'dealing with the meaning' and this latter 'has denotation and is a complex' . The phrase, in short, should no longer be taken to be a denoting complex that has both meaning and denotation , for to say it has the latter is to say it has a meaning which stands in R to a denotation. If one agrees that a stipulated referential connection between a sign and an object is not a myterious relation, whereas R , as a connection between entities is, then the Fregean view can be looked at as one which abandons an 76

unproblematic connection between a sign and an object in favor of two problematic notions: expresses (or means) and R. All this leads to a further , and fundamental, objection to the Fregean view and the relation R. Russell 's argument , which is one of the most misunderstood parts of his paper, begins But this only makes our difficulty in speaking of meanings more evident . For suppose C is our complex ; then we are to say t hat C is the meaning of the complex . Neve rtheless , whenever C occurs without inverted commas , what is said is not true of the meaning , but only of the denotation , as when we say ; The centre of mass of the solar system is a point . Thus to speak of C itself, i . e . , to make a proposition about the meaning . . . 34

Having introduced meanings, irrespective of the difficulty about R, we should be able to talk about them. The attempt to do so will lead to further problems. The first mistake an interpreter of the above passage is prone to make is to consider , in spite of what Russell says, that he is using the term "C' as he used it earlier in the discussion, to denote something other than a meaning . ·c is a variable denoting phrase, which Russell specifies differently at different points in the paper. If one reads the second and last sentences of the just quoted passage with care, it is clear that Russell is telling us that he is now taking C to be the meaning MC . C is the denoting complex , and a meaning , recall, is taken to be something that can have a denotation. C , then, is a meaning, MC, say, and Russell is now concerned with stating something about it. A problem of interpretation arises since Russell speaks , first of C being the meaning of the complex, after saying that C is the denoting complex, and, second, of C occurring without inverted commas. Phrases, not meanings, are the sorts of things that appear with or without commas. But, once again, what he is arguing is clear enough and, once we understand what that is, we can see that the alternative possible interpretations make no difference to his argument. The argument is simple enough. To state something about the meaning involves denoting it. Hence, there will be a non-verbal proposition containing "something' which denotes that meaning ; just as the verbal expression of such a proposition will contain a denoting phrase that means such a "thing' and denotes the meaning we talk about. Hence, there will be, on the view he attacks, a further meaning which denotes it. MC denotes DC and is involved in our talk about DC. Our talk about MC involves us with another entity, say MMC, which denotes MC. The point involves a familiar objection to a Fregean view. If we refer to a meaning by a denoting sign or phrase, then such a sign will have a meaning as well as a referent. Some have felt that we thus introduce a mechanism for producing an unending series of entities. Russell raises a related question. He asks whether MMC can be identified with MC or whether it

77

must be a new entity. He is concerned to prove that they must be different. The heart of this argument is the previously established point that, for any denoting complex O, the denoting complex - the denotation ofO -cannot be taken to denote what O denotes.If MMCwere identical with MC, then MC would denote itself, for, recall, MMC denotes MC.Perhaps a more familiar way of putting matters will help.Let the term 'MC' denote the entity MC.Let the entity MMC be the meaning or what is expressed by the term 'MC'.Thus, the term'MC' denotes MC and means or expresses MMC. But, MMC stands in R, or denotes, MC. Hence, if MMC is identified with MC, then MC will stand in R to itself: i. e., it will denote itself.Now, however, we must recall that on the pattern in question MC denotes DC. Thus, consider a proposition, in the sense of a Fregean thought, which is a proposition about the meaning MC.A constituent of the thought or proposition will be something which denotes M C, namely MMC. After all, that is what MMC was introduced for in the first place . But, if MMC is identified with MC, then the relevant constituent of the proposition will be what the proposition is supposed to be about. Hence, the proposition will be about DC, not MC, because DC is what is denoted by MC. This is Russell's argument. Understanding it, we can then readily understand Russell's continuation of the just cited passage : Thus to speak of C itself, i . e . to make a proposition about the meaning , our subj ect must not be C, but something which denotes C. Thus 'C' , which is what we use when we want to speak of the meaning, must be not the meaning , but something which denotes the meaning. And C must not be a constituent of this complex ( as it is of 'the meaning of C' ) ; for if C occurs in the complex , it will be its denotation , not its meaning , that will occur , and there is no backward road from denotations to meanings , because every obj ect can be denoted by an infinite number of different denoting phrases . 35

When he uses the letter C in inverted commas he is speaking about MMC, distinguishing such an entity from MC.Moreover, he is claiming that MC cannot be taken as a constituent of MMC.What he has in mind by this claim can be understood as follows.MMC denotes MC, hence it occurs in a proposition (Fregean thought) that is about MC.Suppose MC is taken to be a constituent of MMC , so that the latter is composed of MC and a further sense, (or sense function) associated with the phrase "the meaning of ... '. Thus, as Russell sees it , MMC would not denote MC, but the meaning of whatever MC denotes.In short, he is reiterating the same argument we discussed earlier; only here , he is putting it in terms of his views, from The Principles of Mathematics, of meanings (or denoting complexes) as compounds of entities. Thus, the meaning or denoting complex associated with the phrase 'the father of the author of Waverley' would contain , as a constituent, the meaning or denoting complex 78

associated with the phrase 'the author of Waverley' j ust as the latter, in turn, might be taken to have as a constituent the concept of being an author. In the case of the denoting complex the father of the author of Waverley, no problem would arise, since, the obj ect we wish to denote is, so to speak , determined by the object denoted by the complex: the author of Waverley. Thus, what the denoting complex or Fregean meaning the father of the author of Waverley denotes can be taken as a function of what the denoting complex or Fregean meaning, the author of Waverley, denotes , without our losing what we want the former to denote. This is not the case with the two denoting complexes or meanings: MMC and MC. Here, what the former denotes cannot be taken as a function of what the latter denotes. 36 Again, Russell's exposition is overly repetitive and complicated, but that is not the same as being confused , mistaken, and ignorant. Understanding what Russell' s argument is we can then see why he proceeds to assert that MC is not to be identified with MMC. Thus, he goes on to conclude: Thus it would seem that ·c and C are different entities such that 'C' denotes C ; but this cannot be an explanation , because the relation of 'C' to C remains wholly myste rious ; and where are we to find the denoting complex 'C' which is to denote C?31

Here he is taking C and "C' to be the MC and MMC of my exposition: A meaning and a further entity that denotes the meaning (i.e., a 'second level' meaning) . So, since we are forced to distinguish MC from MMC we must recogn ize a new double mystery. First, MMC is a new mysterious entity, even more so than MC, since in the case of MC we at least have an anchor to an obj ect we are familiar with, DC. In the case of MMC, we have a denoting relation holding between two things, both meanings, neither of which we are familiar with, and therefore, second, the relation R, as holding between MMC and MC, is even more mysterious than as a relation between MC and DC. Moreover, it is clear that we not only have an infinite series of such meanings, but that the series is built on the merely linguistic denoting phrase "the meaning of the phrase 'C'" . So concludes Russell' s argument against the relation R which, as he sees it, is a necessary ingredient of the Fregean view. Let us now return to the two expository problems we noted above. C is a meaning or denoting complex. What could it then mean to say ·c is the meaning of the complex' ? A simple explanation is that Russell momentarily mixes the two senses of denoting complex, the entity and the phrase, and speaks of the entity as the meaning of the phrase. This would explain his going on to speak of C occurring without commas, and thus simply solve our second problem of i nterpretation as well. There is a 79

more complicated interpretation we might consider, in spite of our having noted Russell's tendency to use the same term for both the term itself and its referent. Recall that j ust before the problematic passage, Russell had concluded that the meaning of a phrase could be said to have a denotation but it was not something that had both a meaning and a denotation. But if the meaning, MC, does not, in turn, have a meaning, what can be the constituent of a non-verbal proposition that will be about MC? The only candidate seems to be MC itself. Thus, such a proposition will contain MC. The role of a meaning here, we remember, is to occur in a proposition as a denoting complex: it stands in R to that entity the proposition is about and about which we speak when we utter a sentence (verbal proposition) which expresses the (non-verbal) proposition. In this sense, the meaning of the complex MC - that which occurs in a proposition and which denotes the complex - must be MC itself. That it cannot be is what Russell proceeds to argue. Even his statement about C occurring without inverted commas need not be puzzling. It is just a way of saying that if MC occurred in a (non-verbal) proposition that proposition is about DC, not MC. To be about MC a proposition must contain an entity such as ' MC', i. e. , MMC. This leads directly to the obj ection to such an entity that we consid ered above. Cassin, as I understand her paper, bases much of her discussion on Russell's supposed concern with a denoting phrase like 'The denoting complex MC'. The idea is that Russell is trying to show that such a phrase, and the denoting complex it expresses, is paradoxical. In terms of our discussion, such a phrase could be taken to express MMC. Then, if MC is taken as a constituent of MMC there is the problem of having MMC denote MC (stand in R to MC) while being a function of what MC denotes. That is, let MMC be composed of the sense of the expression 'the denoting complex' plus (in some unspecified sense of " plus') the entity MC. Then, as the entity MC denotes DC, MMC will supposedly denote something determined by DC. But , DC cannot determine anythin g in the realm of meanings, since no meaning is, so to speak, a function of it. Thus, the denoting complex MMC will not denote what we wan t it to denote, since MMC cannot both denote MC and denote what it does as a function of MC's denoting what it does, DC. Ifcorrect, the point would be that the phrase "" the denoting complex MC" presents us with a specific case where MMC cannot contain MC, but, apparently, should . However, the point is not correct . What should be taken as a constituent of MMC is not MC but the meaning expressed by the sign 'MC'. Such a meaning would, like MMC, den ote MC, and MMC could denote what it does as a function of what that meaning would denote. All one need do here is to take the phrase 'the denoting complex' to express an identity 80

functi on. The case is n ot different from that where we take'the individual Scott' and " Scott' to stan d for the same thing. Cassin also thinks that Russell's discussi on is confused since he fails to distinguish between, in our terms, a proposi tionF containing MC, and a ' proposition R , contai nin g DC . If she is correct that would account for some of the termi nology and the obscurity. B ut such a confusion would neither touch Russell' s main arguments nor n eed it be attributed to hi m. The relevant passages in The Principles do n ot indicate that Russell either confuses or clearly separates the two proposi tion al entities in the case of a propositionv with a denoti ng phrase. 38 At this point Russell has ended one li ne of argument. He proceeds to another line , which leads directly to his theory of defi nite descriptions. The final argument against a Fregean style theory of meaning is stated as follows: Moreover, when C occurs in a proposition , it is not only the denotation that occurs (as we shall see in the next paragraph) ; yet , on the view in question , C is only the denotatio n , the meaning being wholly relegated to 'C' . This is an inextricable tangle , and seems to prove that the whole distinction of meaning and denotation has been wrongly conceived . 39

To prevent obvious misun derstanding, it should first be n oted that all Russell means by the claim that ''it is not on ly the den otation that occurs" is that the den otation is not the only thing that is relevan t . This is clear from three assertions i n the next paragraph: That the meaning is relevant where a denoting phrase occurs in a proposition is formally proved by . . . . . . hence the meaning of 'the author of Waverley' , must be relevant as well as the denotation . . . . . . we are compelled to hold that only the denotation can be relevant . 40

With this simple observati on in mind we can grasp Russell' s argument and see that he is not stating the '" negation" of Frege' s view in place of Frege' s vi ew. However, he has n ot used the term'proposition' univocally throughout the passages we have consi dered. In the present passage he is clearly using the term in the sense of a 'verbal expression', i . e. , a meanin gful sentence, rather than in the sense of a Fregean thought or as a proposition in the sense of his discussion in The Principles of Mathematics. This ambiguous use can lead one to see Russell as confusing 'being a part of a proposition' with 'being referred to by a proposition' and of use with mention. For , as we noted earlier, in The Principles Russell thinks of a sentence like 'Scott is tall' as expressing a proposition which contain s the object Scott as a constituen t . However, even if he is thinking 81

in terms of his old view, and hence speaks about a propositionR in the above passage, the point of his argument can still be put in the simple terms he uses in the fol lowing paragraph regarding the relevance of the meaning. If we then ask 'relevant for what? ', the answer is not very elusive. Granting that the argument is intelligible to this point, it behooves us to attempt to seek to understand the remainder of his argument, rather than taking the easy way of dismissal on the basis of an understandable variable use of the term 'proposition'. In fact we do not have to look far or try very hard in order to understand Russell. What Russell is getting at is the following. Consider the sentence (S 1 )

Scott = the author of Waverley.

Suppose we raise a question about the truth condition for the corresponding 'proposition' , i. e., what must obtain for the proposition to be true? On the Fregean pattern there are only two possible answers, if we ignore the reply that the proposition denotes the True, which is no answer at all. The latter is no answer since it merely invites a repetition of the question in another form. One answer is that the denoted object is self-identical, which would be the same condition that obtains if we considered (S2)

Scott = Scott.

A second reply is that the meaning of the denoting phrase 'the author of Waverley' denotes what the name 'Scott' , or the meaning of the name, denotes. This means that meanings enter into the truth-condition. Suppose we held that propositions , to be significantly employed, must state or express or indicate what truth conditions obtain if they are to be true. What Russell is claiming is first, that the conditions for (S 1 ) and (SJ must be different. The first reply, on the Fregean view , does not allow for that. If we offer the second reply, then the proposition somehow states that a constituent of it, a meaning, denotes an object (or that the object 'Falls under' the sense or meaning). Such a proposition is then about a meaning which is a constituent of it. But , it cannot be; or so Russell has argued in the preceeding paragraphs. Thus the Fregean view fails. The assumption is that a Fregean proposition must connect with or indicate a truth condition: what obtains in order for the proposition to be true. This assumption Russell does not argue for , nor will I argue for it here, except to claim that Fregean propositions are useless as entities expressed by sentences if they do not , in turn , indicate such conditions. Frege, by talk of the True and the False , avoids the question in much of his writing. Given this assumption , Russell's criticism is cogent. As Russell sees it (S 2 )

82

will be true if a certain object is self-identical while (S 1 ) will be true if that object has the property of being the only author of Waverley. Thus, an adequate view must construe (S 1 ) and (S�) in a way that reflects that difference. The theory of descriptions he proceeds to offer does exactly that. For, on the theory Russell proceeds to present, verbal propositions like (S 1 ) will differ logically from verbal propositions like (S ')), since they will express different truth conditions (or propositions in -the sense of propositions R) ; what Russell will come to speak of as facts. Frege's theory does not capture the difference since both names and definite descriptions express senses and denote objects. Thus, the Fregean propositions expressed by (S 1 ) and (SJ are of the same kind though they differ in ·constituent' senses or meanings. Frege 's theory does not, for Russell, capture the purely referential or indicating function of names. This is a theme that not only lies behind his criticism of Frege but which links the earlier theory of 1 903 with the new theory of 1 905. Thus, when Russell holds that the meaning must be relevant what he means is that the denotation of "the author of Waverley' is not all that is relevant to the truth of (S 1 ) , whereas on the Fregean view one must either hold that it isor hold that the sense of the denoting phrase enters into the truth condition. As Russell sees it, either alternative is problematic. To insist that Russell's discussion reveals a misunderstanding of Frege or that Russell discusses the '"negation" of Frege's view is to miss the point. Of course, on Frege's view, (S 1 ) and (S2 ) express different propositions. The question is how, on Frege's view, one shows that the truth conditions are different , assuming that they must be. One interpretation of Frege acknowledges that they are not different . The other interpretation makes the correspondence between senses and objects the truth condition. This leads Russell to hold that a Fregean proposition, in stating what must obtain for the proposition to be true, is about its constituent senses or meanings, and not just about the denotations of the phrases (or the senses , for that matter). It may help to think of Russell's argument as somewhat analogous to the claim that identity statements like 'Scott = Sir Walter' cannot be treated metalinguistically, as the claim that the two names name one thing, since then the sentence would be about words that refer and not just about the referent of the words. In Russell's critique of Frege the claim is stronger , since Russell finds the Fregean view paradoxical in that the non-verbal proposition cannot be about its constituent senses. By contrast, on the metalinguistic treatment of identity statements, one advocates the replacement of the sentence by another. Such a move is not open to Frege without basically changing his analysis. One important difference that Russell's theory of descriptions, set forth in ' On Denoting' , will have from the Fregean account will be, of 83

course, that entities like senses will be rejected along with Russell's earlier denoting concepts . But, an equally important difference will be that the mysterious relation R will be replaced by exemplification as a relation between properties and objects. Thus, given the sentence (S 3 ) The author of Waverley is tall, the phrase 'the author of Waverley' will be about an object since that object will have the property indicated by the predicate x wrote Waverley and (y) (y wrote Waverley if and only if y = x). Russellian definite descriptions thus provide a means for linking verbal expressions , or thoughts employing them, to denoted objects. In that they do so they function like Fregean senses. Yet, such descriptions do not involve an appeal to such Fregean entities or to the relation R. Associated with a definite description is a property, not a sense , which is exemplified by an object that the expression or thought employing it is then about. Notes 1

2 3 4

5 6 7 8 9 10 11

12 13 14

15 16

17 18

J . R. Searle, ' Russell's Obj ections to Frege's Theory of Sense and Reference· , reprinted in Essays on Frege, ed . by E . D . Klemke ( Urbana : University of I llinois Press , 1 968) . pp . 337345 . Ibid. , p . 342. Ibid. , p . 344 . P .T. Geach , 'Russell on Meaning and Denoting ' , reprinted in Essays on Bertrand Russell, ed. by E . D . Klemke (Urbana : University of Illinois Press , 1970) . pp . 209-2 1 2 . Ibid. , p . 2 1 2 . A . Church , ' Carnap's Introduction t o Semantics' , Philosophical Rei·iel·V ( 5 2 . 1943 ) , p . 302 . R . Carnap , Meaning and Necessity, (Chicago : University of Chicago Press , 1970) . p . 140. Ch rystine E. Cassin , ' Russell's Discussion of Meaning and D enotation' , in Essays on Bertrand Russell, pp. 256-272 . Ibid. 27 1 . B . Russell, 'On Denoting' , reprinted in Essays on Logic and Kno wledge. ed. R . C . Marsh (New York : Allen and Unwi n , 1968) , pp . 45-55 . Ibid. , p. 46 . Ibid. , p . 46 . Ibid. , p. 48. Ibid . . p . 49 . Ibid. , p . 49 . Ibid. , p. 49 . Ibid. , p . 50. G. Frege , ' On Sense and Reference ' . in Translations from the Philosoph ical Writings of Gottlob Frege (Oxford : Blackwell. 1970) , p . 59.

84

19

20 21

22 23 24 25 26

27

28 29

30 31 32 33 34 35 36

37

38

39 40

Ibid. , p. 58. G .E . Moore , Some Main Problems of Philosophy, ( London : Allen and Unwin , 1958) . Russell, op. cit. p. 49 . Ibid. p. 49. Ibid. , p . 49. Ibid. , p. 49. See note 12 above. Ibid. , p. 49 . B . Russell , The Principles of Mathematics, ( London: Allen and Unwin, 1956) , p. 47 . Ibid. , p. 43. Ibid. , p. 47 . Ibid. , p . 43. B. Russell . ·on the Relations of Universals and Particulars' , reprinted in Logic and Knowledge, ed. R.C. Marsh , ( London: 1956) , pp. 107-8. On Denoting, op . cit. pp. 49-50. Ibid. , p. 50. Ibid. , p. 50. Ibid. , p. 50. The re ason , to repeat, why the denotation of MMC cannot be a function of the denotation of MC or, to put in another way, why MMC cannot be a function of MC , is that we want MMC to denote MC , but a denotation does not uniquely determine any meaning which denotes it. Thus, we could use the phrase 'the denoting complex (meaning) which denotes DC' to express MMC and denote MC only if there was a unique meaning which denoted DC. There is not. Hence, Russell ends th e just quoted passage by pointing that out. Moreover , all Russell means by saying 'the denotation occurs' is that if MC is a constituent of MMC , then MMC denotes DC or something uniquely determined by DC. Hence , MMC does not denote MC . Ibid. , p. 50. It is worth noting that in the 1918 'Philosophy of Logical Atom ism' , Russell holds that "the constituents of propositions, of course , are the same as the constituents of the corresponding facts .. . " , op . cit. , p. 248. By then, however, denoting phrases do not indicate constituents of anything. On Denoting, op . cit. , p. 50 . Ibid. , pp . 50-5 1 .

Professor Quine, Pegasus , and Dr. Cartwright

Classical philosophers seriously assert such propositions as "there are no physical objects'. But they also tell us that we do not in the ordinary sense err when we use language as we ordinarily do, saying, for instance, that we perceive physical objects such as houses and guided missiles. As Hume noted, outside their closets philosophers themselves say and believe such things. They thus use language in two ways - in its ordinary sense and in one that is puzzling and problematic. If we are then to decide whether what they say is true we must first find out what they are saying, i. e. , we must explicate the problematic propositions. We also note (a) that the conclusions one draws depend on the grammatical form of the statements that express the premises , and (b) that statements such as 'Pegasus is not real' and 'George is not tall' share the same grammatical form though they do not share the same logical form. This suggests that philosophers are led to say some of the things that they do say because they rely on grammatical and not on logical form. Consistently pursued , the notion of logical form leads to that of an ideal language in which logical and grammatical form coincide. If the classical philosophical propositions are indeed spawned by problematic use and grammar, then these problematic and puzzling propositions could not be stated in an ideal language. On the other hand , such a language would have to con­ tain the transcriptions of all ordinary nonproblematic statements . Furthermore, one might attempt to explicate. or reconstruct, the meaningful core of the classical philosophical propositions as statements about the ideal language. In thus speaking about the structure and interpretation of an ideal language, one would speak commonsensically (use words ordinarily or . what amounts to the same thing, nonphilosophically). Ordinary discourse of this kind about an ideal language and the classical propositions is , on this view, the heart of philosophical analysis. Let us now look more closely at the notion of an ideal language. Consider first the notion of-syntax. Written signs (for our purposes we need not consider others) are instances of geometrical shapes . Syntax is concerned only with the shape and order of signs. Hence, it deals with signs and sign sequences only as geometrical patterns. This is what is meant by speaking of syntax as 'formal' and syntactically constructed schemata as formal languages. Philosophers, being neither geometricians nor logicians, are not

interested in such schemata for their own sake. Rather, they are concerned with the potentiality of such schemata for serving, upon interpretation, as ideal languages. As an illustration of a schema with such potentialities we may consider the schema of Principia Mathematica (with ramification suppressed) supplemented by the addition of certain classes of descriptive constants to its primitive signs and the so-called axiom of extensionality to its primitive sentences. 1 The schema is constructed syntactically, without any reference to its prospective use , to force attention on what may otherwise go unnoticed and thus to avoid building philosophical commitments into the very construction of the schema. One first selects certain shapes as its elements or signs. The elements are then divided into categories, the division again being based on shape and nothing else. Signs are either descriptive or logical. The descriptive signs, upon interpretation, are either proper names or predicates (monadic, etc.) of the various types. Logical signs are either connectives, or quantifiers, or variables. Certain sequences of signs are selected as sentences (well-formed formulae). The formation rules are thus syntactical or formal. This latter fact, together with the possibility of syntactically distinguishing a sub-set of well­ formed formulae which, upon interpretation , become what philosophers called logical truths, is a critical point in the explication of 'analyticity'. 2 The shapes originally selected are called the undefined (primitive) signs of the schema. The reason for setting them apart is that we may wish to provide a machinery for adding new signs. To each sign added corresponds one special sentence, called its definition, everything being so arranged that this sentence is analytic. This has two consequences. First , the definitions of the 'language' which the schema becomes on interpretation are all nominal. Second , interpretation of the undefined signs automatically interprets all others. One way to interpret a syntactical schema is to pair its undefined signs one by one with words or expressions of our natural language . In the case of undefined descriptive signs of the schema this involves making them name or refer to the same things as the English terms or expressions. An interpreted schema is still only in principle a language since it is neither rich nor flexible enough to function like a natural language. An ideal language is an interpreted syntactical schema. This does not mean that every such schema is an ideal language. To be one it must have three properties. (a) It can serve as an adequate tool for the analysis of all philosophical problems. That is, all philosophical propositions can be reconstructed as statements about its syntax and interpretation. (/3) It must, in principle, contain transcriptions of anything nonphilosophical we might want to say. Thus it must enable us to account for all areas of our experience. For example , it must not only contain (schematically) the

87

way in which scientific behaviorists speak about mental contents but it must also reflect the different way in which one speaks about his own experience and that of others. And, it must then show how these two j ibe. (y) It must not contain the transcriptions of any of those problematic propositions and uses which give rise to the philosophical problems. Let me illustrate what is involved in (a) . Consider the classical theses of nominalism and sensationism. Reconstructed, the first becomes the assertion that the ideal language contains no undefined descriptive signs except proper names; the second becomes the assertion that the ideal language contains no undefined descriptive predicates except nonrelational ones of the first order, referring to characters exemplified by sense data. These reconstructed statements, though almost certainly false, are yet not absurd, as so many of the classical theses are. Notice two more points. First, in reconstructing a classical proposition one does not assert it ( either affirmatively or negatively). One merely suggests what it could be taken to mean. In practicing the method one always speaks commonsensically, (i. e. , without using any word philosophically) . Thus when I used 'name' and ·refer' above, I used them commonsensically - in a manner and on an occasion where they do not give rise to philosophical troubles. This is why one can, without circularity, clarify those uses that do give rise to philosophical problems. It is also important to note that, since we no longer assert any of the classical theses, we need no longer choose among traditional positions. Each of the classical answers to each of the classical questions has a commonsense core. To safeguard his insight each of the classical thinkers was driven to deny or distort those of others. Reconstructed these ·cores' are no longer incompatible. One may thus recover the several insights of the several classical philosophers without either assserting the absurd classical propositions or accepting the extravagances of the traditional positions. With this schematic account of an ideal schema in mind , we can now turn to the issues raised by my previous paper. I said that ""we cannot literally state, within a clarified language , either that Pegasus exists or that he does not exist" . Dr. Cartwright takes issue with this assertion by holding that I must be 'unaware' of the fact that one can , in a formal language, write 'E ! ('x) (WHx)' , 3 which states that there is one and only one winged horse. That is , if 'Pegasus' is construed as a description, one can make existential statements about the winged horse. Before considering my awareness or lack of such, I shall first recall some points from my paper in order to provide a setting for the discussion. I had said that in terms of a clarified ( ideal) language one can distinguish between philosophical (ontological) uses of · exist' (ontology) and ordinary uses of that term . I had further insisted that this distinction was needed for a successful analysis of the ontological questions. A

88

philosopher's answer to the ontological question "What exists?' is, I suggested , reconstructed as a question about the undefined descriptive terms of his explicit or implicit ideal (clarified) language.4 An existent, for him, is what is named by one of these terms. Yet if this clarified language contains no predicate of existence, then he cannot literally in it say about an existent that it exists, for one can of course not existentially quantify over descriptive constants . That existents exist shows itself by the occurrence in the clarified language of the descriptive constants that name them. Furthermore, in a clarified language of this sort there will not occur undefined descriptive constants that do not name anything (designate). Thus , since zero level constants are all undefined , there will be no zero level constant signs that do not name anything. Similarly, one cannot literally , in the clarified language, say about •something' that has no positive ontological status that "it' does not exist. I also pointed out that the ordinary uses of 'exist' are reflected in the existential operator of a clarified language and that, consequently, statements containing this operator have no ontological significance . Thus one does not within a clarified language reproduce problematic philosophical assertions (in this case ontological ones) by using the existential operator. This is as it ought to be . For, we saw , upon reconstruction the philosophical questions become questions about the structure and interpretation of an ideal language ; both questions and answers taking place in ordinary commonsensical discourse . In addition, I contended that the attempt to state literally, in a clarified language, with the philosophical (ontological) sense of ·exist' that Pegasus does not exist leads to trouble. Since I said all this very clearly, it should have been clear that when I said that neither "Pegasus exists' nor "Pegasus does not exist' can literally be stated in a clarified language I meant two things and did not mean a third . I did mean and did say , first, that there is in a properly clarified language no undefined descriptive constant that could take the place of (correspond to) the English term ·Pegasus' in a transcription into this language of the English sentence "Pegasus exists' . I did mean and did say, second, that this latter English sentence has no transcription at all in the clarified language and that in this language the existence or nonexistence of Pegasus merely shows itself by the (possible) occurrence or (necessary) nonoccurrence of a certain undefined descriptive constant . 5 I did not mean and did not say, third, that one could not , in an interpreted formal language or in an ideal language (the two notions, we saw, are not the same), assert that there is one and only one winged horse . Moreover, I also said more than once that if the English term 'Pegasus' is construed as a description, then the difficulties I pointed out do not arise . I even used as an illustration an existential statement (in symbols) about the

89

winged horse (to the effect that it is white) . I also said, in criticizing Quine, that the sentence 'There is such a thing as appendicitis' is an object language statement. Thus I find Dr. Cartwright' s supposition about my lack of awareness rather surprising, particularly since what I said is neither original nor very radical. Russell, whose views Dr. Cartwright tries to contrast with mine, wrote long ago: For the present let us merely note the fact that , though it is correct to say "men exist ' , it is incorrect , or rather meaningless , to ascribe existence to a given particular x who happens to be a man . Generally , 'terms satisfying 0x exist' means '0x is sometimes true' ; but 'a exists' (where a is a term sati sfying 0x) is a mere noise or shape , devoid of significance . [ 19 1 9 , eighth impression : 1 953 , p. 1 65] As regards the actual things there are in the world, there i s nothing at all you can say about them that in any way corresponds to this notion of existence . It is a sheer mistake to say that there is anything analogous to existence you can say about them . [ Mo, vol . XXIX, no . 2, 1 9 1 9 , p. 206]

Dr. Cartwright thinks that my unawareness clashes with the view, which he ascribes to me, that every grammatically proper name is synonymous with some description, and hence, I take it, in principle eliminable. This, by the way, he also attributes to Russell. His evidence for my holding this view is that I urged ' Pegasus' be treated as a description! I hardly know what to say about this imputation and this evidence, especially since I never used 'synonymous' . Actually, there are two possible interpretations of what he says. Since (a) in a clarified language grammatical and logical form coincide and (b) he believes that one can, in such a language, say about an existent that it exists, he could mean (q1 ) that all logically proper names are replaceable by descriptions. This view (a later one of Quine' s) I do not hold . 6 Nor, for that matter, do I believe that Russell held this view. I quote Russell: It may be possible to invent a language without names , but for my part I am totally incapable of imagining such a language . This is not a conclusive argume nt , except subj ectively : it puts an end to my power of discussing the question . [ 1 940 , p. 1 17] In connection with certain problems it may be important to know whether our terms can be an alyzed , but in connection with names this is not importan t . The only way in which any analogous question enters into the disussion of names is in connection with descriptions , which often masquerade as names. But whenever we have a se ntence of the form 'The x satisfying 0x satisfies '4,x· we presuppose the existence of sentences of the forms "0a" and "'!'a " , where "' a " is a name . . . . Per contra , though some among indispensable words are , I believe , to

90

be classified as names. these are , all of them , words not traditionally so classed [ 1 940 . pp . 1 1 8-1 19] . 7

Dr. Cartwri ght' s '"version .. of Russell' s view stems, I suggest, from a blur i n the way he states it. At one place he speaks of this view as involving the elimination of all grammatically proper names. At another place he says that according to Russell the proper names of ordinary discourse are all eliminable. These two sentences need not say the same thing. As Russell and, following him, I use the phrase, a grammatically proper name of an ideal language is an undefined zerolevel constant. Russell never held the view that such logically proper names are expendable. The view he did hold, at least at certai n stages of his thought, is that the proper names of ordinary discourse are not (do not correspond to) logically proper names and are, therefore, all eliminable. Since I was, throughout my paper, not concerned with the proper names of ordinary di scourse and spoke always in terms of an ideal language, I assumed throughout, with Russell, that this language contains, i nexpendably, logically proper names. Again, Russell himself has been most explicit about this matter of proper names in an ideal language: In a logically perfect language , there will be one word and no more for every simple obj ect . and everything that is not simple will be expressed by a combination of words . . . A language of that sort . . . will show at a glance the logical structure of the facts asserted or denied . . . all the names that it would use would be private to that speaker . . . [Mo, vol . XXVIII , no. 4, 1 9 1 8 , p. 520]

However, Dr. Cartwright could simply mean (q2 ) that the proper names of ordinary language can all be construed as descriptions in a PM type symboli sm. He may then think that this vi ew conflicts with my contention that ontologi cal assertions cannot be made in an ideal language since he believes that (q2 ) involves the acceptance of both (r) existential statements, using '3', can be made in a PM type symbolism, and (s) such statements involve ontological commitments. (r) is, of course, not at i ssue and does go with (q2 ).( s) , however, is at i ssue and certai nly does not follow from (qJ. Dr. Cartwright has some further things to say about Russell. These, too, he thinks, are to the detriment of my paper. Let us see. So it seems clear that on Russell 's view of the meaning of grammatically proper names singular existe nce statements can be expressed in L * -type languages . . . Where 'N' is any grammatically proper name , the statement that N exists can be rendered by a state ment of the form (3x) (y) [ . . . y . . . = (y = x) ] , where ' . . . y . . . ' is any open sentence such that ' ('y){ . . . y . . . )' is synonymous with ' N' . [ Psc, vol . 23 , no . 3 , 1 956 , p . 262]

91

If we wish to say 'N exists' all we have to do, according to Dr. Cartwright, is to find a description that is 'synonymous' with 'N' . If 'N' is a logically proper name, and, therefore, a primitive zero level constant of a clarified language it is not a defined sign. Surely, a definite description of what 'N' names cannot then be taken as a definition of 'N' . If 'N' is a proper name of ordinary discourse then to say it is an abbreviation for an expression in an ideal language is to mix English with the formalism. I do, therefore, not know what 'synonymous' means in this passage. Dr. Cartwright does nothing to explain to us how he uses this philosophically problematic term. We have here come across a good example of how a completely unanalyzed notion of synonymy operates in his argument. If 'synonymous' is used as we use it when we say that definiens and definiendum are synonymous 'by definition' then there is no question about the sense of the passage. 8 Only, in this case, it is certainly false. What then I wonder is Dr. Cartwright's notion of synonymy? Is it such that synonymy can be established by synthetic identities? 9 Be that as it may, since he appeals to Russell, I shall quote once more : We have , then , two things to compare : ( 1 ) a name , which is a simple symbol , directly designating an individual which is its meaning, and having this me aning in its own right, independently of the meaning of all other words� (2) a description , which consists o f several words , whose meanings are already fixed , and from which results whatever is to be taken as the 'meaning' of the descriptio n . [ 1 9 1 9 , eighth impression : 1953 , p. 174] 1 °

Quine insists on putting 'Pegasus does not exist' literally into his ideal language. This, I argued, gets him into two kinds of difficulties. (a) He fails to distinguish and is led to collapse the ordinary and the ontological uses of 'exist', representing them both by the existential operator . Yet this distinction is indispensable for an adequate analysis of classical ontology. The reason why it is indispensable is that if one neglects it, one will, as Quine does, merely restate, quite traditionally even though in symbols, one of the traditional ontological positions - instead of , by means of a symbolisrn, explicating all of them without accepting any of them in their traditional form. (b) Quine 's insistence on putting "Pegasus does not exist' literally in his would-be ideal language presents him with certain puzzles about inference. In order to solve these puzzles he alters the rules of inference so that they are no longer syntactical . The main purpose of my paper was to analyze the confusions which led to those unwarranted alterations . Let us see once more. Consider a language structurally simil ar to PM with primitive zero-level constants ·a 1 ' , 'a 2 , , ·a -' ' . . . Suppose that the language also contains the incomplete symbol ' ('x) (WHx) ' and , furthermore , that we want to introduce an

92

abbreviation for it. Since Quine continually mixes ordinary English and his symbolism, uncritically jumping from the one to the other , he picks 'Pegasus' to serve as this abbreviation. One who is aware of the dangers of all this mixing and jumping may ask whether Quine, if we limit him to symbols, would pick one of the ·ai' to serve as an "abbreviation' for " ( ,x) (WHx) '. The answer is that he would. This I shall now show. Consider ( 1 ) · (z) - (z = Pegasus)', ( 1 ' ) · (z) - (z = an)', and (2) ' (3y) (z) ­ (z = y) '. The conventional rules permit the inference from ( 1 ') to (2). Quine holds ( 1 ) to be true and to be one way of saying, in his would-be ideal language , that Pegasus does not exist. 1 1 (2) is a contradiction. Thus he must keep us from inferring (2) from ( 1) . He does so by saying that the crucial term (here ·Pegasus' ) in such inference patterns must in fact designate. Thus he alters the purely syntactical character of the traditional rules of inference. This shows conclusively that he does not really construe the English ·Pegasus' as a description of his would-be ideal language, in spite of what he himself and Dr. Cartwright may say to the contrary. To grasp that point, we merely have to remember that , if we replace ·Pegasus' in the •mixed' jumble ( 1) by a (symbolic) description, we are already kept from making the critical inference by the conventional rules of inference. 1 2 This is indeed the point of * 14. 1 8 : . E ! ( 1x) (0x). =:) : (x) '!J x. =:) . '!J ( 'x) (0x) [ 1 9 1 0--- 1 9 1 3, second edition: 1 950, vol. 1 , p. 1 80] which shows that if 'Pegasus' in ( 1) is replaced by a description the inference from ( 1) to (2) requires an additional premiss. Thus, if Quine "construed" 'Pegasus ' as a description, he would not need to modify the conventional rules of inference. I conclude that he does not '"construe" ·Pegasus' as a description but as an undefined zero-level constant. In other words he treats ( 1 ) as if it were ( 1 '). Since (2) does, by the conventional rules, follow from ( 1 ') but must not follow from ( 1) in the case of Pegasus and other nonexistents, he divides the zero-level constants into two classes - those which merelypurport to name and those which do in fact name something. For the latter the conventional inference holds ; for the former it does not . This is the new rule. Clearly it destroys the syntactical character of inference. One major source of all the unnecessary commotion and confusion is the introduction of "Pegasus' into the clarified language. So one may wonder why both Quine and Dr. Cartwright are so eager to do that. Quine I believe , and previously said , wants to get as close as he can to restating literally, in the clarified language, 'Pegasus does not exist'. Dr. Cartwright is misled by his uncritical use of the philosophically problematic 'synonymy'. If 'Pegasus' and the description are 93

synonymous and if the latter is in the ideal language, so should or, even, so must the former. The ensuing confusion is not noticed because of an inadequate conception of the role a clarified language plays in philosophical analysis. Let me explain. Take two signs, 'Pegasuso ' and 'Pegasus/ . Let the first be the English word ; the second a sign introduced into a formal language. Consider next the English sentence 'Pegasuso does not exist'. This sentence, we all know, is the source of certain philosophical problems. To solve these problems some philosophers construct a clarified or ideal language. Suppose that in such a language we have '(1x) (WHx) ' . We thus have three expressions; one English, two in the clarified language. The English term 'Pegasus/ gives rise to the problematic use. We discover that we can solve the problem (or problems) by means of one of the two formal expressions, '('x) (WHx) ' , without making use of the other, 'Pegasus/ . As a mere abbreviation of '(,x) (WHx)', 'Pegasus/ is worthless. 13 If we introduce 'Pegasus/ as a logically proper name, then we undo the clarification. In particular we may then be tempted to say, in an entirely unanalyzed sense of 'synonymous' , that 'Pegasus/ and '('x) (WHx) ' are synonymous. Or, again, one may say that 'Pegasus/ is an abbreviation for '('x) (WHx) '. N ow, to say that a logically proper name is an abbreviation, either of a definite description, or of anythin g else, is paten tly absurd . What one should say, if one wants to make some sense, is, therefore, that 'Pegasuso' is an abbreviation of ' ('x)(WHx)'. But if one says that , then one scrambles English with the symbolism. Besides, of what possible use could 'Pegasus/ be. The only on e I can think of is to seemingly reproduce, in the ideal language, as literally as possible the sentence which gives rise to the problem. This problem we have solved already, without 'Pegasus/ . An d it is, at least in this case, part of the solution to have shown that 'Pegasus/ need not occur in the ideal language. Furthermore, there is always the danger that one will confuse 'Pegasus/ with 'Pegasus/ even if one does not treat 'Pegasus/ as a logically proper name. This, I suggest , is on e reason for Dr. Cartwright' s being sure that 'Pegasus exists' can be literally stated in a clarified language, since, as he sees it , that sentence is " synon ymous' with 'E! ('x)(WHx)' [ p. 262] . Since ' ('x)(WHx)' is the definition of 'Pegasus' ('Pegasus/ ) these expressions have the same meaning and , consequently, are 'synon ymous'. Hence, ' Pegasus exists' ('Pegasuso exists') is synonymous with'E! ('x)(WHx)', and therefore one can literally say, in a clarified language, that Pegasus exists! Dr. Cartwright reminds us that Quine actually says 'Pegasus' is a description. He even quotes him to that effect. I also cited such passages. This Dr. Cartwright acknowledges but interprets as merely an inconsisten cy on my part . Then he goes on to say that in a thorough search 94

through Quine 's wntmgs he could not find a single passage which suggests to him that Quine "" envisions a language in which "Pegasus' will be construed neither as a description nor as a denoting individual ,, constant. So he considers himself as performing the desirable task of , ""set(ting) the record straight. . Concerning Dr . Cartwright's task two things need be said. First, one sometimes has to consider not only what a philosopher says but also what is involved in it and what is a consequence of it ; for, unfortunately, the two do not always jibe with each other. Thus, in order to see how Quine thinks of "Pegasus' we must investigate how he treats the term , as well as note what he explicitly says about it. I trust I have now once more shown cause for my assertion that Quine, whatever he may say, did not really treat "Pegasus' as a description. Second, Dr. Cartwright could have saved himself the trouble of searching . I did, after all . give specific references to passages where Quine explicitly defends names (not descriptions !) that do not name. I transcribed the sentence 'Nothing is identical with Pegasus', with "Pegasus' as a definite description built on the property WH, as (3x) [WH(x) · (y) (WH(y)

= (y = x)) · (z) - (z = x) ]

and I said that on this interpretation it was false . Dr. Cartwright says this transcription is incorrect . There is, he claims, a rule about expanding sentences which contain descriptions - "begin by expanding the smallest subsentence containing "Pegasus', or '( 1 x)(WHx)' , then expand the next smallest such subsentence, etc. " Consequently the sentence should be transcribed by (z) - (3x) [WH(x) · (y) (WH(y) - (y = x)) · (z = x)] . The issue is thus the scope of the description. About this matter I shall say three things. First, what Dr. Cartwright says is l iterally false. Second, I was nevertheless careless in my interpretation of Quine and did in fact err . But my error is one of omission rather than of commission. Third, my error was, as it were, a happy sin, since it not only does not in the least detract from my point, but also affords me an opportunity to strengthen it. There is no 'rule' about first expanding the smallest subsentence containing the description. The ambiguity of scope is, in PM, taken care of by the use of a scope operator, to indicate the scope intended. Russell merely adopts as a notational convention the omission of a scope operator when the use of the smallest scope is intended. Quine, I believe, was in fact not concerned with matters of scope, for he never expands the description but always uses 'Pegasus'. I erred nevertheless by not

95

considering the alternatives presented by the ambiguity of scope. Taking the scope as primary, as I did, the sentence in question is false, as I said it was. In this case Quine's problem , namely , how to exclude undesirable inferences, does not even arise ! But even if the scope is taken as secondary there is still no such problem. As long as 'Pegasus' is really treated as a description , it cannot operate in inference patterns as either a zero level descriptive constant or as a free zero level variable. This I have shown before when I referred to * 14. 1 8. It is only when one mistakes a zero level constant for an 'abbreviation' of a description or , perhaps even worse, for a 'synonym' of it , that one begins to entertain Quine's problem. The question of scope is thus neither critical nor even relevant. Dr. Cartwright professes to be puzzled by the distinction between the philosophical and the ordinary uses of 'exist' (note that he , not I , italicizes the definite article). 14 And he complains that I did not explain it. Yet the distinction is familiar to everybody familiar with what went on in analytic philosophy in this century. Nor do I really think that Dr. Cartwright wants me to explain it. Rather , he wants me to defend it or , perhaps, to defend the conclusions which, following many others , I draw from it. Even so, since he complains , I will remind him of two things. (A) People say such things as ( 1 ) 'Physical objects do (or do not) exisf . People also say such things as (2) 'Beergardens exist in Vienna' , meaning no more nor less than 'There are beergardens in Vienna·. ( 1 ) is an example of the philosophical use , (2) is an example of the ordinary use. As we noted earlier . the reason for the distinction is that only the former gives , in the familiar manner , rise to the philosophical problems. The founders of the analytic movement made the distinction because they considered it important , and it has, within that tradition . been considered important ever since. (B) Ideal languages have been constructed , as tools of philosophical analysis, ever since the beginning of the analytic movement. The first to construct one was Russell ; he was followed by the Wittgenstein of the Tractatus. These ideal languages were al l so constructed that while the ordinary uses of 'exist' can be transcribed in them . the philosophical uses cannot be so transcribed . The latter , we saw, are then explicated as questions and answers about the structure and interpretation of an ideal language. Also . the explications of ontology proposed by analytic philosophers all hinge on this difference. (A ) and (B) are certainly familiar. But this does not mean that they are right. More precisely, it must be argued that the distinction in (A) actually is important and that the explication of ontology in (B) is satisfactory. This, I think , is the defense (not explanation !) that Dr. Cartwright wants me to produce. In this paper I have tried to indicate why I accept (A) and (B). For the rest , I do not think his request quite fair. I

96

wrote, as we all must in a short paper , and as he himself writes, within a certain frame of reference - namely , the one which accepts (A) and (B). Nobody does or can do anything else. Even so, I shall add two comments. Nobody can defend a pattern of analysis or a frame of reference directly. This sort of thing is j udged by its fruits. (This should be doubly clear to one who , as I believe Dr. Cartwright does, hails intel lectually from Oxford.) Thus I am not unfair, on my part , when I refer Dr. Cartwright to the literature. If he wil l only look for them , he will , I think, find a few very able "defenses'. This is my first comment. As it happens, though, I did in fact contribute to a defense , in that indirect way in which it can and must be undertaken , by showing that Quine's deviation from the patterns (A) and (B) is unnecessary , confused, and confusing. This is my second comment. Dr. Cartwright , of course , knows and understands all this. For , if he did not , how could he , as he does , argue against (B). His argument is as succinct as it is incredibl e - in both tone and content. Why, he asks should one not interpret the existential operator of the 'formalism' as reflecting the philosophical ( ontological) sense of ' exist'? And he writes as if I had foolishly overlooked this obvious alternative, despite the fact that I devoted a whole paper to show that Quine chose j ust this alternative with disastrous consequences. I think I can explain why Dr. Cartwright says such strange things. Before explaining it , though, it will pay if I first cal l attention to another thing, also very strange , which he also says. Why, he asks , should we not, if that be necessary, construct a 'formalism' that contains two 'operators', each reflecting one of the two uses of 'exist', and thus again obtain the distinction on which I insist . The reason why Dr. Cartwright says all these strange and irrelevant things is, I suggest, that he does not appreciate the difference between a 'formalism ' or even an 'inte rpreted formalism' and an 'ideal language'. In other words, he seems to believe that any formalism, just because it is a formalism, is automatically the (or a !) ideal language. I repeat , therefore, that an interpreted formalism is an ideal language if and only if it has the three properties (a), (/3) , and (y) . Generally, it seems that most, if not all, of Dr. Cartwright's misunderstandings are caused by his not paying sufficient attention to a large and significant body of recent analytical thought. In recent years, Kripke has j oined those who rej ect the construal of definite descriptions as 'synonyms' for names (in Cartwright's fashion) or as capabl e of replacing names (in Quine 's manner) or as requiring backing-up descriptions. In so doing he has introduced the notion of a 'rigid designator' with its problematic dependence on talk of possibl e worlds and its equally problematic connection with the notion of an essential property. The idea involved in the notion of a rigid designator is

97

easily expressible in Russell's terms. Given a name, say 'a ' , there are derivative descriptions, such as ' ('x) (x =a) ' , which are such that sentences like 'a = ('x ) (x = a) ' and 'E ! ('x) (x = a) ' are logically equivalent to 'a = a' and '(3x) (x = a)' in a PM type schema. (See the point made above in note 9 in criticism of Cartwright). Let 'a-izes' abbreviate 'x = a' and be taken as a predicate standing for a property of a . It is clear why one who introduces such a property will think of it as an essential property of a , since a , and only a , can and must have that property . Assume that a-izes is the only essential property of a . Then, it is equally clear that the description ' ('x) (x a-izes) ' is different from any other description that a fulfills. The sentences cited above show how and why. They show that there is a sense in which the description '('x) (x a-izes) ' , or '( 1x) (x = a) ', functions like the name 'a' : a sense in which no other description of a so functions. (I ignore trivial compounds such as '('x) (x = (1y) (y = a) ) '. ) The name 'a' and the description ' ('x) (x a-izes) ' thus form a special class of designators of a , if we lump names and descriptions together as "designators' and overlook the crucial ways in which they differ-see 'Russell's Attack on Frege 's Theory of Meaning', 'On Pegasizing', and "Strawson and Russell on Reference and Description'. Forming a special class, they deserve a special name. This is what the idea of a 'rigid designator' amounts to. There is another, philosophical, point involved. Talk of possible worlds and free logic is tied to the idea that we can make claims of existence and hold that things which do exist need not. Thus, that '(3x) (x = a) ' is a theorem of PM type schemata is taken to be problematic, since it should be a factual truth and not one that holds logically or 'in all possible worlds. ' Moore raised the same problem and the same objection long ago. He found Russell's claim that 'a exists' is not a significant statement to be unacceptable, for such reasons. But , a Russellian type view must be understood in context. Given that proper names must be interpreted, it is redundant to use a name to ascribe existence, as Moore acknowledged. However, this does not mean that what is named must exist in the sense that it is a necessary existent. Rather, that something exists can be shown by the occurrence of a name for it, and need not be stated . That a language which, in principle, could be an ideal language , and hence serve as an instrument of ontological analysis , need not be an ideal language reveals that no particular existent need exist. One shows , as it were, what Moore and, following him , Quine and Kripke , want to state. That is why '(3x) (x = a)' does not state that a exists, but is a mere consequence of the occurrence of "a' as an interpreted sign.

98

Notes 1

See 'Properties, Abstracts, and the Axiom of lnfinity' for a discussion of the problems with such a claim , in this volume , pp. 339 f. 2 For such an explication see 'Two Cornerstones of Empiricism', reprinted in G. Bergmann , Metaphysics of Logical Positivism, (New York : 1955) . 3 Dr . Cartwright informs us that the expansion of this sentence is not a 'logical theorem' . For Dr . Cartwright there are ·extra-logical' as well as 'logical' theorems , and the latter can be either true or false upon interpretation . _. For a detailed exposition of this view of ontology see G. Bergmann, 'Particularity and the New Nominalism', Methodos, 1954, pp . 13 1-147 and Metaphysics of Logical Positivism ( 1955). 5 Russell put this in his own way as follows : It would seem that the word 'existence' cannot be significantly applied to subjects immediately given ; i . e . , not only does our definition give no meaning to ' E!x," but there is no reason in philosophy , to suppose that a meaning of existence could be found which would be applicable to immediately given subj ects . See B. Russell and A. N . Whitehead, Principia Mathematica, 2nd ed ., vol . 1 . (Cambridge : 1950) and elsewhere : You can assert 'The so-and-so exists', meaning that there is just one c which has those properties , but when you get hold of a c th at has them, you cannot say of this c that it exists, because that is nonsense : it is not false , but it has no meaning at all. See B. Russell, 'The Philosophy of Logical Atomism' , Monist. vol . 29, nos . 1,2,3, 19 19, pp. 32---68, 190---222, 345-388 . 6 For some reasons for not adhering to such a view see my 'On Pegasizing' , in this volume , pp. 101 f. 7 Russell also explicitly , and in some detail, considers the possibility of substituting descriptions for all names and decides against it. See B. Russell, Human Kno wledge: Its Scope and Limits (New York : 1948) . 8 It is interesting to note that if one considers 'Pegasus' as defined by ' ( , x) (WHx) ', 'Pegasus = ( ,x) (WHx) ' is not analytic. Some may consider this to be a reason why one can not intro­ duce, in the case of this veryparticular incomplete symbol, a further abbreviation for it. 9 Dr . Cartwright also asserts, in connection with his use of 'synonymy' : Hence , given any sentence containing a grammatically proper name , there is an equivalent sentence differ-ing syntactically from the given one only in having a description wherever the given one has the proper name . See R.L. Cartwright, 'Comments on Dr . Hochberg's Paper', Philosophy ofScience, vol 23 ,3, (July 1950) , pp . 260---265 . I do not propose to interpret wh at he means by 'equivalent'. I merely note that (a) ' N = N' , wh ere 'N' is a proper name, is analytic, while (b) ' (,x} (¢x) = (,x) (¢x)' is not. Hence (b) isnot analytically equivalent to (a). There are, however , certain descriptions for which such identity statements are analytic , for example , ' (,x) (x = N) = ( ,x) (x = N)' . They are those wh ere 'E!( ,x) ( . . . x . . . ) ' , '(,x)( . . . x. . . )' being the description, is analytic. Hence, they are not the descriptions involved in the replacement of proper names. 1 0 Russell also made this point in detail in the well-known case of 'Scott' and 'the author of Waverley'. See B. Russell, 'The Philosophy of Logical Atomism', Monist, vol . 29, nos. 1 ,2,3, 19 19, pp. 32---68, 190--222. 345-388. 1 1 One might note, as Dr. Cartwright does , that ' (3x) (x = Pegasus) ', where 'Pegasus' is a description , is analytically equivalent to ' E!('x) (WHx) ', which says that there is one and only one winged horse . In PM * 1 4 . 204 : E!(,x) (¢x) · = · (3b) · ( ,x) (¢x) = b · [p. 181] .

99

Perhaps this is the basis for Quine's use of the algorithm of identity as a key to existence. (See my 'The Ontological Operator' , Philosophy ofScience, vol. 23 , 3 (July 1950), pp. 250-259.) Moreover, in view of Dr . Cartwright's transcription of 'E! ('x)(WHx)' as ' Pegasus exists' , to say about something that it exists is analytically equivalent to saying that it is identical with something . Since ' (3x) (x = a) ' , where 'a' is a proper name , is analytic, it seems that to ascribe existence to something that is named is to utter an analytic statement. It is thus no wonder that both Quine and Dr. Cartwright are eager to abolish proper names in favor of descriptions. 1 2 Dr. Cartwright seems to understand this , but he fails to see what this point implies about Quine's discussion. 13 One attraction it holds may be found in the syntactical resemblance of '£! (Pegasus ( which / one could then write) to 'e(Pegasus/ with 'e' as a philosophically problematic predicate of existence. Thus one seemingly gets closer to literally expressing 'Pegasus exists' in the clarified language. This syntactical resemblance is especially suggestive if we recall that '0!(x)' is well-formed in PM. 14 Though using the definite article I did not mean to insist that there was j ust one use in each category. Since Quine collapses the two categories (philosophical and ordinary usage) into one, I was concerned in my paper to distinguish them as categories without being concerned with subdivisions. Furthermore, I did not insist that the distinction is in all cases sharp and clear. One of the tasks of analysis , as I conceive it, is to provide such clarity.

On Pegasizing

Professor Quine has claimed that all proper names , and would-be proper names , can be construed as definite descriptions. He does this in part to avoid the difficulties presented by singular terms that are alleged proper names ("Pegasus') ; in part to give more prominence to the ontological role of the existential operator , in keeping with his conception of ontological commitment. For , whereas one cannot literally say 'Pegasus exists' using a proper name and the existential operator one can say , for example , There is something that pegasizes': the latter type of statement being more conducive to the use of the logical sign '3' for ontological assertions about Pegasus. What I wish to examine here is Quine's apparent belief that his program for the elimination of all proper names is unproblematic, obvious , and trivial. In order thus to subsume a one-word name or alleged name such as ' Pegasus' under Russell ' s theory of descriptions , we must, of course , be able first to translate the word into a description . But this is no real restriction . If the notion of Pegasus had been so obscure or so basic a one that no pat translation into a descriptive phrase had offe red itself along familiar lines, we could still have availed ourselves of the following artificial and trivial-seeming device : we could have appealed to the ex hypothesi unan alyzable , irreducible attribute of being Pegasus, adopting, for its expression , the verb 'is-Pegasus' or 'pegasizes' . 1

Thus he apparently wishes to construe all proper names in terms of 'ordinary' (without 'izes' predicates) descriptive phrases. Where there is no obvious group of predicate terms available to construct such a phrase one employs the 'izing' device. Predicates created by this device , like 'pegasize s' , would be undefined terms ; the property of pegasizing being irreducible and unanalyzable. Before considering such peculiar predicate s let us first examine some consequences of a program for eliminating all proper names in terms of ordinary Russellian definite descriptions. To be able to use definite descriptions instead of proper names one would have to know that the uniqueness clause of the description is fulfilled. 2 Hence , supposing five predicate terms were used in constructing a certain definite description , Quine would have to know that one and only one thing in the world exemplified all five properties. This means that in order to know whether or not he could use one descriptive phrase to ascribe some property to an individual he would

101

have to know something about everything. As all individuals are to be signified by descriptive phrases, to know anything about any individual we would have to know something about every individual. One is thus led to an essentia)ly Hegelian outlook. This comes out rather clearly when one sees that the elimination of all proper names in favor of definite descriptions also eliminates atomic statements about individuals. None of this should be too surprising in view of Quine' s recent overtures toward the pragmatists, including his " abolition" of the synthetic-analytic dichotomy. In fact one consequence of his proposal for the exclusive use of definite descriptions can be said, in a sense, to tie in with his views on the synthetic-analytic issue. A proposition like'F(a) v - F(a) ' , whre 'a' is a proper name, is analytic. With only definite descriptions, we would no longer have such analytic propositions. For, upon expansion of

F(( ,x) ( Gx) ) v [ ( ,x) G (x)]. - F(( ,x) G (x) ) we would get

( 3x ) { G(x) · (y)[ G(y) ( 3x ) { G(x) · (y)[ G(x)

= (y = x)] · F(x) } v = (y = x)] · -F(x) }

which is not analytic. Similarly, all such propositions would become synthetic. Perhaps this might be considered a step in the abolition of the dichotomy. We might further note that whether or not there are a finite number of objects in the world would be a relevant factor for the success of Quine's program. The existence of an infinite number of individuals would pose more difficulties for determining the fulfillment of any uniqueness clause. Thus, whether or not his analysis of proper names into definite descriptions will be technically feasible, apart from the question of its philosophical significance, depends on the gathering of factual scientific information. Again we have a point of contact with the pragmatists. This time via the scientizing of philosophical analysis. Since Quine speaks of having undefined predicates like ' pegasizes' it might seem that there are some propositions about individuals which can be known without knowing something about all individuals. Two things may be said about this. First, the Hegelian consequences are not the reasons for Quine' s introduction of the 'izes' device. It is introduced, not for any philosophical reasons, but, rather, as an artificial technical device to enable him to glide over those cases where ordinary definite descriptions are not conspicuously convenient. Second , one would still have a uniqueness clause in the expansion of the descriptive phrase. At this point one might say that in the case of the ' izes' predicates one

102

automatically knows that the uniqueness clause is fulfilled. This invites us to take a closer look at the · peculiar predicates· . Though of no apparent problematic status t o Quine, the treatment of · pegasizes' as an undefined propert y term may well puzzle others. One might wonder how Quine knows the meaning of an undefined property term when the property is not exemplified. Quine apparently, nominalist though he be, is in communion with unexemplified , unanalyzable, irreducible properties. Such communion returns us to the problems that Russell' s principle of acquaintance was invoked to dissolve. Quine thus transplants us to a previous era in philosophy with all the consequent puzzles and perplexities. The situation is really no better when his peculiar properties are exemplified. Let us consider the predicate ·trumanizes'. This term, I believe, expresses , on Quine 's view, a property that is exemplified. B ut what kind of property is it ? How can one, looking at two things , determine whether one of them trumanizes - without covertly or overtly defining · trumanizes' ? Apart from the question of determining whether only one individual has that property, how are we to determine whether any has it ? One is reminded of Kant and Moore and the 'property' of existence. In this vein we might be tempted to ask whether or not in listing the descriptive properties that characterize Truman we would include trumanizes on the list. B ut the very asking of this question pointedly illustrates the peculiarity of this property. For •trumanizes' is really intended to represent the conj unction of all the descriptive properties that serve to individuate Truman. (We remember that 'izes' predicates are introduced when 'no pat translation into a descriptive phrase' is available. ) B ut this has to be done covertly by making it into the one (unanalyzable) individuating property which is then included among the properties of Truman. For to do it overtly would be to define 'trumanizes', and hence to furnish an 'ordinary' descriptive phrase. And , since it is the individuating property, one might then claim that nothing else has it. But then all this is to be expected. Since Quine gives up proper names , the purely indexical linguistic signs by means of which we refer to and indicate particulars , his peculiar undefined predicates take on the task of individuating obj ects. Thus they are not ordinary descriptive predicate terms . They reintroduce the medieval properties of individuation. For Scotus , we will recall, developed , in a manner of speaking, Truman-ness - the haecceitas in Truman. Once again we return to the past. This time to the metaphysical puzzles surrounding the problem of individuation and the attempt to individuate objects by ,, "special properties . Aside from such problems , the employment of terms like "pegasizes' , with a 'guarantee' (whatever that might mean) that they are truly

103

predicable of one and only one thing ( or of nothing), does nothing more than introduce "proper names" at the predicate level. The philosophical significance of such a procedure, even if it were unproblematic, is thus not apparent. 3 Its attraction for Quine can be understood in view of his conception of the ontological question. The 'izes' device provides him with a means of attaching the problem of ontology to the logical sign'3' in a literal manner. For, as we noted above, with proper names he could never literally say'George exists' in a formal language - one just doesn't quantify over proper names, and the existential operator is Quine's 'existence sign'. Terms like 'pegasizes' seem to resolve this problem. With the abolition of proper names, he can now make ontological assertions as literally as is possible by using the existential operator. (To say'There is something that pegasizes' is the way one says'Pegasus exists' in Quine's language without proper names). However, since he only recognizes one sense of the term 'exist' , that embodied in'3' (hence his pert phrase 'to be is to be the value of a variable'), he does not distinguish the peculiarly philosophical uses of that term from its ordinary uses. Quine thus returns us to a pre-analytic era in philosophy by uttering traditional metaphysical propositions, since 'there are' in 'There are musicians' and 'exist' in'Physical objects do not exist' are both packed indiscriminately into '3'. Moreover, just as he can make ontological pronouncements about what is, he can also literally speak about what is not ('There is no thing that pegasizes'). Only instead of doing so by using proper names that do not name, and consequently bringing in the problems that led to the theory of descriptions, he employs unexemplified, unanalyzable, irreducible properties. In avoiding one set of puzzles we run into another. Or, to put it another way, he re-introduces the problems of proper names that do not name at the predicate level. Again, that is to be expected, since his peculiar predicates have taken over the individuating functions of proper names. Notes 1

2

3

W. V . Quine , 'On What There Is' . reprinted in From a Logical Point of View, (Cambridge: 1 953), pp . 7-8. I am speaking of using definite descriptions to truly ascribe properties to individuals . That one can use unfulfi lled descriptions to make false statements or true compound statements is irrelevant here. For a discussion of some related issues see G. Bergmann, " Particularity and the New Nominalism", Metlwdos ( 1 954), pp . 13 1--4 7 .

Strawson and Russell on Referenee and Description

Strawson's attack on Russelrs theory of descriptions is also an attack on Russell's view of proper names and their function as referring signs ( 1950). In this paper I shall consider both of Strawson 's critiques and shall attempt to show that they are misleading and unfounded. I shall also attempt to show that Strawson's alternative view is philosophically more problematic than Russell's and that attempts to justify it as closer to ordinary usage are mistaken. First, we shall take up the question of names and reference and then turn to descriptions . 1. Naming and Referring The issue between Russell and Strawson can be seen in terms of a simple artificial situation. Consider the signs 'a', 'b' , 'F1 ', and 'F2'. Assume we have not provided any interpretation for them in the sense of specifying what objects or properties they may be used to refer to. But, assume that we have specified grammatical rules determining that 'F1 a' is well­ formed , that "F1 ' and "F2' are predicates that may be used to refer to properties, and that ·a' and 'b' are to be used as proper names of individuals, as opposed to properties or characteristics. Here 'name' may be used synonymously with 'zero-level constant' and, hence, a sign is a name only in that it belongs to a certain grammatical category and not in virtue of its referring to some object. In this sense names need not name. Knowing that the sign is a name and knowing the grammatical rules of our simple system, we may say that we know how to use the signs. Since we know what kind of sign it is and, perhaps , what sort of things we will use signs of that kind to refer to , even without knowing what specific references, if any, we will make, we may say we know the meaning of the sign. Suppose we then use the sign 'a' to refer to various objects on different occasions. One might then believe that the sign has been used to name them all and, hence, that the name, as such, does not name one thing, but may be used to name several different things at different times or even at the same time. We can distinguish between the sign, as a physical mark (or class of such) , and the sign together with its interpretation in a particular use. To use it to refer to an object is to give it a specific interpretation. The sign, taken as a type as opposed to a token , 1 05

can also be said to have several interpretations. We might then consider it to be a different name on the different occasions on which it is used to refer to different objects. This emphasizes the different senses of 'name' that are involved. First, we have a sign type being a name in the sense of belonging to a category of signs whose members are used to refer to objects. Second, there is the sense in which the sign type is a name of something in that it is used to refer to a specific object. This latter sense gives rise to the idea that the name is the sign as an interpreted sign or as a sign with a specific referential use. Thus a sign type, used to refer to different objects at different times, may be taken as being a different name on each occasion. Or, perhaps better, some tokens of the sign type would be different names while some tokens would be the same name. If we then raise questions as to whether a sign can be a name if it is not interpreted or as to whether two objects designated by the same sign type have the same name, we can give either affirmative or negative answers, depending on which sense of 'name' we use. In turn, these different senses of 'name' can give rise to different senses of 'meaning'. One may say he knows the meaning of a sign if he knows its grammatical category, the rules applying to it, and the kind of object it could be used to refer to. In this sense the meaning has nothing to do with the actual referent , on a given occasion, or even with the question of whether there is one. Alternatively, one may hold that he knows the meaning only if he knows the referent of the sign. The first sense of ' name' and 'meaning' sums up Strawson's view. Russell takes the second to be philosophically significant . Strawson holds that Russell' s view of meaning is confused . B ut it is, I think, clear that Russell did not confuse the two senses of 'meaning' we just noted. What one might claim is that speaking of 'meaning', 'referring' , and 'names' as Strawson does fits better with the way we commonly speak of such matters. One could then claim that Russell would be confused or wrong if he thought that his view recorded or reflected common usage of the relevant expressions. But Russell did not seem to think that he was recording such usage, in spite of the fact that he sometimes speaks as if he is. Rather, what Russell seems to have been concerned with were certain problems that arise in connection with such ordinary usage and what he proposed was a solution to those philosophical problems. Consider one apparent problem about names. Let ' a' an d 'F1 ' refer, respectively, to a white square patch and to the color white. Given appropriate syntactical or grammatical rules, we may then use the sign sequence 'F1a' to assert that the object designated by 'a' has the color indicated by ' F, ' . I f it were asked what was said or meant by uttering or writing a token of that sentence type, a reply could be made in just that manner. If the questioner knew the interpretation of the signs, 1 06

he wouldn't have to ask . Suppose the sequence is uttered without an interpretation being made of the name "a' . It is clear that we would not then say anything in the sense in which something was said the first time. Both Strawson and Russell agree about this. The forme r holds that the sentence is significant and meaningful but that an assertion was not made. The latte r denies the sentence to be significant or meaningful. One motive for the disagreement could be Russell's desire to avoid introducing assertions , the propositions of earlier writers, in addition to interpreted sentences. Russell, aware of the ontological issues, would recognize that something, in addition to the signs and their refe rents, is being implicitly appealed to in order to account for meaning and the connection between language and what it is about. In short, talk of assertions along Strawsonian lines involves, for Russell , the introduction of assertions as entities in one's ontology in spite of what Strawson may say to the contrary . This issue aside, it might seem that there is only a terminological dispute about 'meaning'. Russell speaks of 'meaning' and "names' in one way, Strawson in another. But, in addition to avoiding propositional entities, Russell was pointing to a distinctive feature of signs used simply as labels or referring signs. In the case of a name which names , we are not forced to use descriptions to indicate what we are speaking about. We need not explain our use by employing descriptions. By contrast, where we use names that do not name, we are forced to use descriptions to indicate what we are talking 'about'. This seemingly trivial difference between names that name and names that do not has significant consequences, if one takes metaphysical or ontological questions seriously. We can get into those questions by noting another set of problems that arise with names that do not name. These are questions of the truth and falsity of sentences and the validity of inferences involving sentences containing non-designating names. Suppose 'a' is a nondesignating zero level constant. A question then arises about 'F1a V ,..._, F1 a' as an instance of a logical truth. To accept it as logically true in the standard sense is to accept either 'Fia' or ',..._, Fi a' as true. But to retain existential generalization as a valid infe rence pattern is to derive either '(3x)F1 x' or '(3x),..._, F1 x' on the basis ofa's being F1 or ,..._, F1 where a doesn't exist. This is unpalatable. Several alternative solutions are open . Russell, in effect, refused to acknowledge 'F1 a V ,..._,F1 a' , 'F1 a', and ' ---- F1 a' as sentences when 'a' did not designate. For a sign to be a name 'presupposed' that it did designate. H ere 'presuppose' is used in a quite straightforward and clear manner. To say that 'a' is a name presupposes that 'a' designates, is simply to say that the latter statement logically follows from the former by the requirements for a sign's being a name ; i. e. by the meaning of 'name'. Russell rules out the problematic sentences by a metalinguistic requirement. Alternatively, if one allows such signs and 1 07

sentences into a language then something else must be done. One may compl icate the logical machinery by permitting existential generalization only if the sign 'a' , in the above example, does designate. Along this line one could syntactically distinguish two kinds of zero l evel constants and require that onl y signs of the one kind designate. The problematic i nferences could then be avoided by restricting existential generali zation to sentences containing signs of that kind. In such a language we would have names, in one sense, that did not name. Another alternative would be to stipulate that sentences l ike ' F1 a' , ',__, F1 a' , and 'F1 a V ,__, F1 a' are neither true nor false, and hence no problem of an inference to either ' (3x)F1 x' or '(3x) ,__, F1 x' ari ses. In effect Strawson adopts this alternative via his distinction between sentences and assertions and the claim that the sentences 'F1 a' and ',__, F1 a' can be used to make true or false assertions only if 'a' does in fact designate. The use of either sentence to make either a true or false assertion i s said to 'presuppose' that 'a' does designate. In his way, by speaking of 'presupposing' , Strawson does something that amounts to what Russell does; he neutralizes the troublesome sentences. Only he does it not by banning the troublesome terms, and hence the sentences, but by holding that the sentences containing such terms cannot be used to make true or false assertions. The problematic i nferences are avoided , since even if one all ows l ogical inference patterns to apply to sentences, rather than asserti ons, we could still not infer that either existential sentence was true (or, perhaps, was used to make a true assertion) since neither' F1 a' nor ',__, F1 a' was or was so used . This would be so irrespective of whether or not we all owed the inferences from 'F1 a' and ',__, F1 a' to their existential generali zations. Thus Strawson's solution, l ike Russell's, avoids the problematic inferences. Russell does so by specifying a metalinguistic criteri on for a term' s being a name of a problem free l anguage, and from that criterion it logically follows that if •a' is a name then "a' designates. A Russellian' s claim that being a name presupposes that something be designated is then a tautology. As we noted above, " presupposes' here only means 'logically entails'. Strawson holds, in effect, that "F a' being either true or false presupposes that 'a' desi gnates. But this is not a tautol ogical consequence of a cri terion for a term ·s being a name. It is, for him, a logical truth of a different kind . Just what kind l ocates a problem in Strawson's analysis. It appears as if St rawson is l aying down a generalization about language and its use. Being neither an empirical generalization nor the consequence of a stipulation or definition, it may be characterized as a 'conceptual' or 'logical' truth. Thus Strawson is, in effect , led to introduce both an unanalyzed notion of 'implies' or 'presupposes' and of 'conceptual' or 'l ogical truth', for there does not seem to be any explication available for these notions as there is for the

1 08

standard formal notion of logical truth and inference. Even if one specified a set of axioms gove rning Strawson 's notion of ' presupposes' this would only corre spond, for example , to an axiomatization of propositional logic without any reference to the explicative device of the truth table s. We would the n have an axiomatization without any explication. Whatever explicatory problems there are about Strawson' s notion of " pre supposes', it may be held that Strawson's view is closer to and supporte d by our ordinary use of language , while Russell violates such usage, since we commonly use na mes that do not name . UnlikeStrawson, Russell apparently refuses to allow such signs into a logically proper langua ge . This implies that our ordinary language is not logically proper and hence mistaken. But the point that Russell was concerned to make can be ma de eve n if we allow nondesignating zero level constants, so long as we preserve a syntactical distinction between names that name and names that do not. One could , as noted above , avoid the problematic infe rences to existential statements in such a system. But there is another question regarding the truth and falsity of subj ect-predicate sentences containing nondesignating names. If one wished, unlike Strawson, to adhere to the twofold claim that all (indicative ) sentences are either true or false and that a sentence is true if and only if its negation is false, he could introduce a further stipulation that atomic sentences of the form ' 0a' are false, where 'a' stands for any name that does not name , while the negations of such sentences are true. Strawson, by his doctrine of assertions, denies that all indicative sentences are true or false but reflects the second claim that a sentence is true if and only if its negation is false. I say ·reflects' since , on his vie w, one may hold an assertion is true if and only if its negation is false. This shows that there are different kinds of assertions in two senses: in the sense of true and false and in the sense of positive and negative. Consequently we see how close Strawson's doctrine of assertions is to an earlier metaphysical claim that there are propositions. For Strawson, as on the earlier view, negation is properly a characteristic of assertions, not of sentence s. This brings us to a peculiarity of the tre atment of negation on the doctrine of assertions. In an early form of the view Strawson seems to have held that asse rtions were either true or false, but he later holds that some assertions may be neither [ T, vol. XXX , no. 2, 1 964, pp. 1 03, 1 05, 1 06] . On this later view, we could have two assertions such that one is the negation of the othe r with neither being true or false. This means that negation, in some contexts, is not truth functional. Moreover, on eithe r the earlier or the later view, one would recognize that some sentences with negation signs we re meaningful but not used to make assertions. Such uses of negation are also not truth functional. It would not do to reply that such uses are 1 09

truth functional since if contradictory sentences 1 were to be used to make assertions with truth values then one assertion would be true and the other false . For the use of the negation sign on such occasions is still not quite the same ; i. e. , once it is used to make a true or false assertion and once it is not. The latter use is then clearly not truth functional. Thus we have a correlate to Strawson's unexplicated notion of 'presupposes' - an unexplicated notion of 'negation'. We can consider one feature of Strawson 's talk about assertions, se ntences, truth and falsity to be reflected by a threevalued logic that applied to sentences. 2 A sentence ' Wa' could take one of three values , T, F, or N. If 'a' designated the n ' Wa' would be either T or F, otherwise N. Then '-' would have the following truth table

P- P F T T F

N N This shows in what sense Strawson does and in what sense he does not abandon the standard truth table for negation . All this is just another way of pointing out that Strawson denies the claim, (1 1 ) , that every indicative sentence is either true or false. His denial is put in terms of the doctrine of assertions and presuppositions. But we may ignore the question of assertions in getting at his difference with Russell. If we allow atomic sentences like ·F1 a' where ·a' is a nondesignating constant, we face, as noted above , a question about its truth value . In effect, Strawson's solution is that neither 'F1 a' nor · - F1 a' is either true or false. Whether we withhold a truth value or introduce a third value N makes no real difference. Another alternative we noted earlier is to adhere to { 1 1 ) and stipulate that 'F 1 a' is false and, hence, its negation true. A third alternative might be to stipulate that both ' F1 a' and '-F1 a' are false and thus not give up {1 1 ) but hold that for such sentences we alter the rules about negation and contradiction. This last alternative thus gives up the further claim, (L,), that an indicative sentence is true if and only if its negation is false. This third alternative may seem extre me. I me ntion it to point to a similarity it has with Strawson 's alternative. Both abandon the uniform use of the standard truth table for negation and both hold that a conjunction of a sentence and its negation need not be or reflect a contradiction. For the moment, let us consider the second alternative, which 'stipulates' that ' F1 a' is false and its negation true, to be Russe ll's . Supposedly Strawson 's view is close r to ordinary usage than Russell's 1 10

and to preserve this fit he is willing to pay the philosophical price of introducing assertions, secondary referents, an unexplicated notion of 'presupposes' and a non-truthfunctional use of negation. But the one supposed strength of Strawson' s view, and weakness of Russell' s, the match with ordinary usage, is specious. Consider the sign 'b' as belonging to a certain grammatical category, which we call that of zero-level constants or proper names. To separate issues let us forget about the possibility of the sign being used on many occasions to refer to many things and simply say, if it refers at all, it refers to one and only one thing. If it does so refer we will call it a referring name. We may establish this reference in many ways; by a gesture, a coordination procedure as in mathematics, or by giving a definite descpription, say '(1x)(0x)'. The gesture and the description do the same job. But there is a radical difference. We can assert · b = ('x ) (0x) ' but not that the gesture is identical with b or 'b = this' . This last sentence would be useless without being accompanied by the gesture (or an appropriate context). Given that "('x)(0x) ' serves to connect "b' to its referent, the sentence 'b = ( 1x)(0x) ' is peculiar. It plays exactly the role of the assertion "Let 'b' stand for this" , where that statement is accompanied by an appropriate context or gesture, or of "Let 'b' refer to a" , where we know what 'a' refers to. In short, we are dealing with a conventional assignment of a sign to an object. Thus 'b = ('x)(0x)' is perhaps better expressed by the rule (R1 ) '' Let "b' stand for ('x)(0x)" . The identity statement may then be taken as a consequence of (R 1 ). The peculiarity arises simply because 'b = ( x)(0x)' reflects or is a consequence of the rule of interpretation for 'b' and is thus not an ordinary statement of identity like ' Descartes is the most significant French philosopher'. The role of the descriptive phrase ' ('x)(0x)' is to connect 'b' to some object, just as a gesture or coordination procedure may do. Assume there actually is such an object and that the sign 'b' is then connected with it. Let us call the object the 'referent-mean­ ing' of the sign 'b'. This notion may then be contrasted with that of the 'grammatical-meaning' , which we may construe as being provided by the implicit or explicit rules governing the use of the sign, including those specifying what sort of referent-meaning it may have. Let us further call the descriptive phrase '('x)(0x)' the 'semantical-meaning' of the sign 'b' . Thus the object, (1x)(0x), i s the referent-meaning while the sign ' ('x) (0x)' is the semantical-meaning of the sign 'b' . Suppose, however, there is no referent of "b' since (1 x)(0x) does not exist. We then have a grammatical­ meaning and a semantical-meaning for 'b' but no referent-meaning. This merely points out that aside from the rules specifying the use of 'b' we have only the sign '(1 x)(0x)' and not the object ('x)(0x) to provide 'meaning' for 'b' . To put it another way, aside from 'b' and its grammar, we have only the descriptive phrase. Since there is no object, a Russellian 1

111

may claim that the descriptive phrase gives the meaning of 'b' . However, since there are the grammatical rules a Strawsonian may claim they provide the meaning. By contrast, since we have only the sign ' ( tx)(0x) ' , and not the object, to provide more than grammatical meaning, the Russellian may insist that 'b' is defined by ' ( 1x) (0x) ' . A definition is often characterized as the specification of meaning for a sign solely in terms of other signs; and { R 1 ), or 'b = ( 1x) (0x)', which introduces 'b' into the language, is remarkably like a definition in that respect. Consider a predicate 'F1 ' defined by 'F1 = ('F)(x)(Fx (G 1 x & H1 x))'. Such a sentence may be said to introduce ' Fi ' into the language by description , just as 'b' is introduced. The similarity may reinforce the idea that 'b = ( 1x) (0x) ' defines 'b' . The distinctions we have made between grammatical, semantical, and referential meaning should enable us to see what is reasonable in the claim. For, given the distinctions one need not deny that the descriptive phrase provides a definition , and hence semantical meaning, since the sign already has 'meaning' - grammatical meaning. Likewise , one need not deny that a definition is provided if the sign has a referent and, hence, has referential meaning. If there is a referent, a Russellian may rightly insist that the sign has the same meaning it would have if the referent did not exist, just as a sentence means the same thing whether it is true or false. What he has in mind is clearly semantical meaning, and, possibly, grammatical meaning as well. Thus , he may well hold that '(1x) (0x)' defines "b', whether there is a referent or not, since 'b = ('x) (0x)' introduces 'b' into the language. This merely reflects the difference between 'b' and a sign, say 'a' , which is not introduced by a definite description as 'b' is. In the case of 'a' there must be a referent, if the sign is to have more than merely grammatical meaning. Given referents for both 'a' and 'b' , they differ only in that 'b' has '('x) ( y = x) &'Vx] (3x) [ 0x& (y)(0y ::> y=x) & '4JX]�1V ( 'x) (0x)

for it must be a further pair of conceptual truths that if the left sides of (S2) and (S3 ) hold ( are true) so must the right sides. I n S we should then have the theorem

This correlates with PM's

Yet, (S4)'s being a theorem does not mean its left and right sides are truth­ functionally equivalent, for '�' and ' �' are not truth-functional notions. 6 Moreover, the left side of (S4 ) may take the value N, while the right side may not. All this just reflects the fact that, as Strawson uses the notion, if 'p' presupposes 'q', it does not 'follow' that ' not q' presupposes 'not p' , whereas the standard use of 'entails' does involve 'not q' entailing ' not p ' , if 'p' entails 'q'. Strawson has recently insisted that the basic difference between himself and Russell is that on Russell's view a statement like the 0 is 1V is only apparently a subject-predicate statement and really an existential statement. On Strawson's view, however, such statements are subject­ predicate statements, not existential statements, but their being true or false presupposes existential statements holding . (R2 ) being an analytic truth and theorem of PM expresses what it means to say that '1V (1x) (0x)' is an existential statement. What is involved is that (R 2 ) either is a, or follows from, axioms of Russell's system, and that system employs the standard truth functional notions. Russelrs axioms embody definitions or rules for definite descriptions. As such they are amo ng the " linguistic' truths or rules of the system. But as (S 1 ) , (S 2 ) and (S 3 ) reflect logical or conceptual truths for Strawson, one might wonder what the difference amounts to. Since ( R2 ) is a theorem of PM '1V (1x) (0x)' is held to be an

1 22

existential statement in PM , but it is not held to be an existential statement i n S even though (S4 ) is a theorem. One difference is that (R2 ) states or is a consequence of a defini tion. (S4 ) is not. But it is preci sely because of this that (R I ) an d (R2 ) may be held to be theorems in PM wi thout introduci ng an unexplicated notion of presupposes an d an extended notion of logical truth. It is because the theorems of S are not consequences of defini tions that ·�' is a problemati c (unexplicated) si gn. Even so, (S 1 ) , (SJ, and (S3 ) hold as axi oms or conceptual truths. A definiti on i n PM is, i n effect, a further axi om. Why do not (S 1 ) , (S2 ) , (S3 ) , and (S4 ) , taken as axi oms (theorems), imply that ''l!'('x) (0x)' i s an existential statement i n S? One obvious part of the answer lies in the di fference between ·�' and · �'. on the one hand, and entails , '=:) ', an d · = · . on the other, and, consequently, between S and PM. Since ' �' di ffers from logical equivalence, and '= ' , (S4 ) holds in S even though its left and right sides may take different truth values: N for the left when the right side is false. The left and right sides not bei ng truth functionally equivalent would naturally involve the clai m that they do not (or are not used to) state the same thing. For, it would seem awkward to hold that two sentences state the same thing i f they are not, at least, truth functi onally equi valent. But this is not all that is involved. Let us modi fy the treatment of descriptions i n S i n a way that resembles Russell's treatmen t of proper names. Russell, recall, holds . as a metalinguistic requi rement, that names name. Suppose we treat Strawson's use of 'presupposes' in a simi lar manner. Consider S to be a calculus all of whose senten ces are ei ther true or false, due to a metalinguistic requirement that all admissible descriptions do i n fact designate. Descriptions now function in S as names would in a Russellian ideal language. The comparison is especially suggestive i f we note that descriptions are not defined signs i n S, j ust as names would not be abbreviations for Russell. Call this modi fied system s : In S' we could sti ll have (S 1 ) , (S2 ) , (S3 ) , and (S4) as axioms or theorems. But what (SI ) states could be reflected in two other ways. On the one hand what it would state in S' could be reflected i n the metalinguistic cri terion for a defini te description being a sign of s: Alternatively, one could add an axiom ' (3x) [0x& (y) (0y =:) y=x)] ' for every predi cate like \�' used to bui ld a description. Thus '(3y) [y = ('x) (0x )]' would functi on in S' like '(3y) (y =a) ' functions in PM with 'a' as a name. Just as the latter is a theorem of PM (with cons tan ts) if 'a' is a name, so the former is a theorem or axiom of S' i f '(1x) (0x)' is a description. Strawson 's noti on of presupposes is now reflected in S' by the metalinguistic cri terion or the list of theorems of the form '(3x) [0x&(y)(0y =:) y=x)] ' . In S' we need no longer consider N a third value of sen tences. 7 We may then i nquire as to whether or not ''l!' ('x) (0x)' is an existential statement in S' , as i t is i n PM. If a Strawsonian would say 1 23

that it is, then it is clear that his critique of Russell reduces to his twofold claim that some sentences (statements) can take one of three values, and, consequently, that logical notions like negation and presupposition cannot be construed or explicated in the standard truth functional manner. If he holds that the sentence in question is still not an existential statement in S', then he has some further points in mind.He might claim that even though the statements ''¢Cx)(0 x)' and ' (3x)[0x& (y)(0 y => y=x)& '¢x ] ' are truth functionally equivalent in S' this is due to a 'logical' connection (other than mutual entailment) and not the consequence of a definition, for descriptions are not defined signs in S'.Thus even though the two sides of ( S 1 ) are truth functionally equivalent in S', the ground of that 'equivalence' is a unique logical connection reflected by' �'. The double arrow is thus not to be taken as the sign for either the biconditional or logical equivalence ; even though the statements it connects are held to be truth functionally equivalent. Thus, while ( R2 ) is an analytic proposition in a PM type schema under the standard interpretation, (S4 ) is a conceptual truth of S', where 'conceptual truth' does not mean analytic in the standard sense.Thus, the issue of the meaning of 'logical truth' arises even if we consider a schema in which all sentences are either true or false.One is forced to hold that ( S 1 ) etc., in S' , reflect a different kind of logical truth in order to hold that S' significantly differs from PM. 8 For, if S4 , as an axiom of S', expresses an ordinary logical equivalence it, like (R2 ) , may be taken as logically equivalent (in the ordinary sense) to a formula of the form 'p p'. That the relevant axiom of PM is called a definition, while the corresponding one of S' is not, is merely a verbal difference.What is crucial is that in both (S4 ) and (R2 ) the left and right sides must have identical truth values. To differ significantly from PM the statement of S' must then be taken as other than an ordinary logical equivalence. One ultimate point of difference between the advocates of PM and S' would then be that the different schemas reflect different conceptions of logical truth. Thus the Strawsonian critique would ultimately rest, in part, on the claim that Russell does not recognize that there is a unique and unexplicated logical connection other than ordinary entailment, which is quite different from the claim that Russell's analysis does not fit ordinary usage of descriptive phrases. A further point could be involved.Consider ·p => q' as an abbreviation for'----p V q'. Is'p => q' then a disjunction , really? In view of the definition of'=:>' one may reply in the affirmative ; in view of the use of ' =:> ' and the problems of intentionality one may also reply in the negative. For the sentence with the defined sign is not literally the same (or of the same form) as its defining sentence , and, when we use the defined sign, we may not intend or have in mind the definition of'=:>' in terms of disjunction and negation. Perhaps, Strawson's critique of Russell reflects, in part , this

=

124

point. 9 Thus, on a given use of a description one might not intend or have in mind an existential statement, and, hence, should not be said to mean one by definition. But the problems of intentionality affect Russell's theory only in so far as they are involved in any definition whatsoever. Once we have distinguished the different senses of meaning involved ( defining as opposed to intending, for example) one can only object to definitions as such , and not just to Russell's theory of descriptions. The question of meaning is at the heart of Strawson's critique. He is claiming that Russell is mistaken, since sentences containing descriptions do not, in general, mean the same thing as existential statements. But the crucial question is , what can be taken to support his claim? The answer seems to lie in ordinary usage. Consider the sentence (I) The present king of France is bald' . Strawson seems to claim that, in ordinary usage, in the year 1 968 we would consider neither that sentence nor its negation to be either true or false. But the evidence of ordinary usage is surely not the explicit claim, by ordinary people, to the effect that Russell is wrong, since the sentences are neither true nor false. Rather, it would seem that the sentences in question create puzzlement , ifone raises a question about their respective truth values. The evidence of ordinary usage is precisely this state of puzzlement which, prima facie, supports neither Strawson nor Russell. Strawson accounts for the puzzlement in one way, Russell in another. A Russellian may point to two features of Russell's solution. First, there is Russell's treatment of the negative sentence, 'The present king of France is not bald', in two ways; one as the denial of (I) and the other as the existential claim that there is one and only one nonbald present king of France. Since one is true and the other false and yet both can be taken to correspond to the ordinary statement on his theory , Russell's analysis reflects the puzzlement that arises. That is, the ordinary negative statement can be taken as either true or false, depending on which way we transcribe it into Russell's symbolism. Moreover, the same thing is true of the affirmative statement (I). The straightforward transcription of (I), on Russell's account, is as the assertion that there is one and only one present king of France and that he is bald. But consider the sentence 'Anything which is the present king of France is bald'. This may also serve to transcribe the ordinary statement into a PM type ideal language. Just as the ordinary negative sentence is ambiguous, we may consider the affirmative one to be so too. And, just as in the case of the negation, one of the transcriptions of the ordinary statement is true, while the other is false . This is how the puzzlement of the ordinary user is reflected by Russell's theory. In a way this involves a slight modification of Russell's theory of description , or perhaps, a clarification of it. The definition of ''4) (1x) ( 0x)' as '( 3.x) [0x & (y) (0y :::J y = x) & '4)X] ' occurs in a clarified or ideal language. One is not offering a 1 25

definition for an English phrase like 'the present king of France' in context. Rather, what should be meant by the misleading claim that the English phrase is defined , or that the phrase in context is an existential statement, is that the English phrase is coordinated to, or transcribed by, a certain expression in PM, which is contextually defined in a certain way. The point can then be made that the English phrase, in use, lends itself to alternative transcriptions; one of which is true, the other false. It is this feature of Russell's theory that corresponds to the fact that puzzlement arises if we raise a question about the truth value of certain sentences with unfulfilled descriptions. In a way closure is achieved when we recall t hat, where the description is fulfilled , the alternative transcriptions are equivalent , in both the affirmative and the negative cases. That is, E ! Cx ) (0x ). =:J : (3x) [0x & (y) (0 y =:J y = x) & ----- '4J X ]

. = . ----- (3x ) [0x & (y) (0y =:J y = x) & '4JX] and

E ! (1x ) (0x ) . =:J : '4' (1x ) (0x ). = . (y) [y = (1x ) (0x) . =:J '\VY ] are theorems of PM. Strawson, too, seeks to deal with the puzzlement that arises in the case of unfulfilled descriptions. He does so by holding that the relevant sentences either do not make assertions or are not used to make assertions that are true or false, since their use presupposes an unfulfilled condition. In short , he introduces the neutral value and the unexplicated notion of presupposes to reflect the ordinary bewilderment. The fundamental point is that to settle the dispute between Russell and Strawson we cannot appeal to the situation they seek to account for: the puzzle that arises in ordinary contexts. One can only hope to settle the issue by appealing to other features of the competing views, such as that St rawson must introduce an unanalyzed notion of presupposes, that he gives up the standard use of the truth functional connectives, etc. One must , in short , seek philosophical reasons for resolving a philosophical dispute, which should not be surprising . There is a further relevant , if peculiar, feature of Strawson's solution. While hold ing that sentences like the 0 is 'V are subject-predicate in form, Strawson holds that sentences like the 0 exists are not. The latter type of sentence is used to make an existential statement and , hence, does not presuppose one. It is obvious that he must hold something like this. If he did not then the assertion that the 0 did not exist would pose an obvious 1 26

problem. Thus, in effect, both Strawson and Russell treat the apparent subject-predicate sentence of ordinary language "the 0 exists' as the existential statement that "there is one and only one �f. Interestingly enough, Strawson holds that this difference between ·the 0 exists' and ·the 0 is tp ' reveals that •exists' is not really a predicate. To ascribe a predicate to a subject is to presuppose, but not assert, that the subject exists. But it is awkward to insist on this condition holding for sentences like •the 0 exists' and ·the 0 does not exist'. Hence, Strawson believes he has found a ground for distinguishing existential statements from subject-predicate ones ( 1 964, p. 1 9 1 ). It seems as if Strawson here ·clarifies' ordinary usage by appeal to his analysis, rather than attempting to j ustify his analysis by appeal to ordinary usage. ( Is •exists' really a predicate in ordinary usage?) This point aside, we may note that Strawson has two uses of descriptions: with predicates and in existential statements. Russell, too , we recall had two axioms about descriptions ; one for the use of a description with predicates and one with '£!'. (But, Russell did not argue, on such feeble ground, that 'exists' is not a predicate. ) In a way Russell's distinction between the two contextual uses of a description also reflects Strawson's notion of presupposition. For Russell, E ! ( 1x) (0x) . =:J . (x)tpx =:J tp('x) (0x) is a theorem of PM. Thus, given that the 0 exists, the description ' ( 1x) (0x)' can, in effect, be treated like a proper name that names. That a name names is, we recall, a 'presupposition' of a Russellian ideal language. Strawson, by insisting that '( 1x) (0x)' is a simple subject sign, like a name, cannot consider a sentence like the 0 is not tp to be ambiguous, as Russell does. Hence, if he were to preserve the claims that every indicative sentence is true or false and that the negation of a sentence takes the opposite truth value of the sentence (in a two valued calculus), he would face the problems Russell starts from. Russell, to solve the familiar problems, gives up the idea that the phrase 'the 0' is a simple sign and, consequently, does not consider a sentence like 'the 0 is not tp' to be unambiguous. Strawson retains the simplicity of the phrase and the claim that the sentence is not ambiguous. Hence, he must reject Russell's views about truth, falsity, and negation ; as well as reintroduce the classical propositional entities in the new dress of 'statements' or 'assertions'. But, again, there is no warrant from ordinary usage that a description is a simple label, like a name, nor that a sentence like the 0 is not tp is not ambiguous. At most, one could claim that descriptive phrases are used, ordinarily, as grammatically proper names, in the specific sense of 'grammatical' that I mentioned in the previous section on names. With respect to ordinary usage, we might consider the following three 1 27

sentences : ( 1 ) The present king of France is bald : (2) The one and only present king of France is bald ; (3) The one and only existent present king of France is bald . One might reasonably hold that (2) merely makes explicit and emphatic what is implicit in ( 1 ) . But does making explicit mean that what is stated in (2) is only 'presupposed' by ( 1 ) or that it is 'implicitly stated' in ( 1 ) ? If one does not appeal to the problems responsible for the introduction of Russell's theory of descriptions and Strawson's theory of presuppositions , how could one hope to make clear what the question is all about? Further , if one does not appeal to the problems of intentionality mentioned above , how could one hope to argue that ( 1 ) only presupposes , and does not implicitly state , (2) on the basis of ordinary usage? Again, what we may take ordinary usage to provide is the fact that (2) does make explicit something implicit in ( 1 ) . But , both Russel l and Strawson have ways of fitting their respective views to this fact or feature of usage . With respect to (3) we may well feel puzzled , since one does not ordinarily talk about the non-existent so and so (forgetting stories and the issues of secondary reference ) . Thus , (3) may be puzzling in that it expresses an even more obvious condition implicit in ( 1 ) . But , again , how is 'implici t' to be taken here? Strawson is claiming that our ordinary usage presupposes ( or entails?) that his sense of 'presuppose' is involved and not Russell's analysis. What I am suggesting is that ordinary usage neither 'presupposes· nor entails either view . There is an instructive corollary to Strawson 's attack on Russell. In his original article , Strawson also castigates 'logicians' for holding that a sentence like 'All mermaids are blond' is true since there are no mermaids . 1 0 For Strawson , the use of such a sentence to make a true assertion also presupposes that there are mermaids . Such an assertion being true presupposes a "subj ect, , j ust as the assertion ""about" the present king of France presupposes a "subject" . Forgetting the issue of assertions , we may note that , for the logicians Strawson castigates , the sentence 'All mermaids are not blond' is also true , since there are no mermaids . Thus , the unfortunate logician claims that '(x)(Fx :) Gx)' and '(x)(Fx :) ----- Gx)' both hold . But for the same logician , the conj unction of '(x)(Fx :) Gx)' and ' (x)(Fx :) ----- Gx)' is logically equivalent to ' (x) -- Fx' , which , in turn , is logically equivalent to · ----- (3.x)Fx'. Hence , all the logician is claiming is that there are no mermaids , when he holds that both universal conditionals are true . And this is not surprising , as he holds them both to be true , in the first place , since there are no mermaids . In short , all that is being asserted by the logician is that · ----- (3.x)(Fx)' is logically equivalent to the conj unction of '(x)(Fx :) Gx)' and ' (x)(Fx :) -­ Gx)'.This equivalence is , of course , based on the logician's treatment of ' :) ' . Thus , what Strawson may really be obj ecting to is the logician's truth 128

table for " if-then" . This would jibe with his introduction of the notion of presupposes, which we saw, forces him to abandon the standard logical connectives, to be content with unexplicated uses of "presuppose' , · negation', ·conceptual truth', etc. and, in effect, to give up a two-valued logic. In reintroducing the classical propositional entities under the name of assertions, Strawson is thus doing more than asking us to recognize additional entities� entities which Russell sought to avoid. He is also asking us to give up the one coherent analysis of logical truth and inference that the tradition has produced . Perhaps there is another point behind Strawson ' s attack on Russell and formal logic. Strawson · s claim that existence is not a predicate leads him to hold that when we say that a class exists it is not clear what we mean if we do not mean that the class has members (1964, pp. 191 ff. ) . He must say something like this, since we cannot transform the apparent subject­ predicate sentence "the class 0 exists' on the model of the transformation of "the 0 exists' into " there is one and only one 0 '. Hence his analysis of existential statements forces him to adopt or, perhaps, reflects his commitment to an extreme form of nominalism. For, on his view, we cannot even say "the class 0 exists' in anything like the Platonist's sense. Strawson 's analysis must then lead to his rejection of the null class, not only as a Platonic entity, but as a legitimate notion. Perhaps this is another strand in his rejection of ' all 0· s are 'V's' as true, if there are no 0 ' s. Be that as it may, we may note that Strawson 's analysis is made to clear up a problematic area of usage and is not presupposed or justified by such usage. By contrast, he implicitly claims that a sentence like 'The present King of France is bald ' is unproblematic and unambiguous in his critique of Russell; for he dismisses Russell's transcription of such a statement as unfaithful to ordinary usage. What he in effect does is deny that there is a problem about descriptive phrases which Russell undertook to solve. I t is then not surprising that Strawson must end with unexplicated notions, like 'presuppose', for he has merely wrapped the problem in ordinary usage and offered it as a solution. Keith Donellan has modified Strawson's critiq ue of Russell by distinguishing between attributive and referential uses of definite descriptions. His idea is simply put. Sometimes when one utters a descriptive phrase one uses it to refer to someone or something irrespective of whether the predicate of the description applies to the object. Thus, imagine a college student sitting down to dinner at home and hearing ' The professor is here' , uttered by a member of the family. As everyone knows , it is the student who is being 'referred to' and not a guest. Yet, the student is not a professor, let alone the only professor in the world. On another occasion one may say ' The man who will be elected will face hard times'. Here one is supposedly claiming that there is some 1 29

man who will be elected without referring to any specific individual.The distinction has generated much pointless debate, since Donellan fails to notice a simple point. If one takes descriptions to be ordinary language expressions of the form 'The (lf , as Russell did, then, of course, one can baptize a child with the name, for example, 'The professor' or, even, 'The Present King of France', assuming this is permitted by law. The expression is, then, a name, while it is also, in form, a definite description. Realizing this, we see that Donellan's distinction is of no philosophical significance. One does not refute Russell by baptizing a child 'The professor'.Rather, one should apply to descriptions a parallel distinction to Russell's distinction between logically proper names and expressions that are grammatically names but not logically proper names. In the case of expressions of the form 'The 0' we may recognize that some of them are used as names or labels. But, even in such cases, Donellan's distinction ignores the descriptive component that is present. For, it is always appropriate to point out, or at least note, that an object does not have the property mentioned in the phrase and, consequently, that something false is stated or implied. There is nothing like that in the case of purely indicating labels - logically proper names. Thus, there is no purely referential use of descriptive phrases. That there are referential uses of expressions that are grammatically descriptions is an obvious sociological fact about linguistic use. It marks another case where grammatical form and logical form do not coincide in ordinary language, and, aside from this, has no philosophical significance. If one took a perspicuous language to be a formal 'mirror' of ordinary linguistic use, one could employ expressions like ' ( 1x )(0x) ' as primitive signs like ' a' , and interpret the former as one interprets the latter. One could also use expressions of the form ' ( 1x) (0x)' as expressions defined in Russell's fashion. But, aside from duplicating an ambiguity of ordinary language there would be no point to doing so. It is worth recalling the similar ambiguities associated with the term 'is'. One point of employing perspicuous schemata is to avoid ambiguities that lead to philosophical perplexities.Here, what we should note is that if one introduces ' (1 x)(0x)' as a primitive term requiring interpretation, then one either ignores the occurrence of the predicate '0' in the expression (and what then is the point of using it?), as well as the imbedded sentential expression '0x' , or one understands it to play a 'descriptive' role. Doing the former , we would be pointlessly using esoteric names: doing the latter we do not merely use the expressions as names. Realizing this we may safely relegate Donellan's distinction to the sociology, rather than the philosophy, of language. Donellan 's distinction is associated with the view , linked with Quine and Searle, that names may be replaced by descriptions or are used with 130

back-up descriptions "in mind'. Thus , some deny the purely referential or labeling use of indicative signs like names . What is interesting is that such a view could only be promulgated in a context where nominalism is taken for granted. For, it is clear that if one takes predicates to stand for properties , then the predicates employed in a definite description of an object will either be taken to refer ( to the relevant properties) without need of further back-up descriptions ( of the properties) or they , too, will be used only in the context of such back-up descriptions. If the former , then we recognize purely referential signs - predicates. If the latter, we face an ominous infinite regress. One is thus forced to take predicates to refer without recourse to their being disguised descriptions or requiring back-up descriptions . And , if this is so in the case of predicates , why is it not so in the case of proper names? It may be that an old philosophical theme is implicitly involved: the idea that, in thought, only universal concepts are "grasped'. Hence, the thought connected to the use of a proper name must involve such universal concepts , which are indicated by the predicates in the correlated description. What is ironic of course is that a nominalistic theme, the denial that predicates refer, joins with a classic theme , the idea that to think of an individual is to think in terms of a set of properties or concepts, to reject names as pure labels. Notes 1

2

3

4

5

Contradictory sentences would be pai rs of sentences of a certain form . i.e. involving the negation sign. Even if one wishes to insist that a notion like 'contradictory' applies properly to assertions . not sentences . there is still a precise sense in which we can talk of contradictory sentences, especially if we avoid the issues that arise from the possibility of interpreting one sign in several ways and hence using ·one' sentence design to say several things on different occasions. To get at several different themes in Strawson ·s views we can and shall consider his view , at times , to claim that a sentence can take one of three values ; rather than that a sentence is not p roperly said to have a truth value, since some sentences are neither true nor false . Strawson speaks of secondary reference i� notes added to P . F . Strawson, 'On Referring , · orginally in Mind, 1950 ; London : reprinted in Essays in Conceptual Analysis ( London: 1960) , pp. 35 , 40. What is involved here is that (S 1 ) is really a meta-linguistic statement since the arrow corresponds to Strawson ·s noti on of 'presup p ose'. Alternatively , the standard use of implies corresponds to a statement . containing ' :=) ' , being a logical truth . For the arrow , there is no corresponding truth-functional notion of 'if-then' , as there is for 'implies' or 'entails ' . Thus if we were to replace ( R 1 ) by ( R' 1 ) '4' ( 'x) ( 0x) � (:!x)[0x & (y ) (0y :=) y = x) ] with • � ' for the metalinguistic 'entails' , the difference would lie in the explication that is available for • � ' but not for •�· . That is, ·�· is neither a truth-functional notion itself, nor is it connected with one as 'entails' is connected with ':=) ·.

131

6

7

8

9

10

Again, in the twofold sense of note 5. Putting the matter as I have should not lead one to think that one sense of 'presuppose' , whereby a sentence being true presupposes that another is true, has replaced a second sense , whereby a sentence being true or false presupposes that another is true. The second is reflected in the set of sentences (S 1 ) - (Sy) along with a corresponding set for '~ '4' ('x) (0x) ' , such as ' ~ '4' ('x)(0x) � (3x)[0x& (y)(0y => y = x)] '. In a way , what Strawson's view amounts to is the rej ection of the construal of'~ '4' ( ,x) (0x) ' as a case of primary scope for the description. In short , Strawson gives up the standard truth table for ' ~ ' . Russell retains the standard negation and allows for both '[('x)(0x) ] . ~ '4J ('x)(0 x) ' and ' ~ [('x) (0x)] . '4' ('x) (0x) ' . That Strawson also has a nontruth functional use o f 'if-then ' , so that 'if p then q ' is neither true nor false in some cases where 'p ' is false or irrelevant to 'q' is not our concern in this paper; though it is a related question. If, instead of (S 1 ) , one employed the rule that for every '0' used to build a definite description in S' there was a theorem ' (3x)[0x & (y)(0y => y = x) ] ' , the difference from PM would come out in that the use of this rule would not be held to make statements with descriptions abbreviations for the corresponding existential statements. For example , in P . F . Strawson, Introduction to Logical Theory (London: 1964) , Strawson says "From the fact that a statement is equivalent to another statement of a certain form , it does not follow that it itself is a statement of that form" (pp. 191-192). Also see his discussion of 'if-then' in P . F . Strawson, Introduction to Logical Theory (London : 1964) , p. 85.

Nominalism , General Terms , and Predication

Platonism , in its most recent and seemingly most cogent form , has rested on (a) the supposed indispensability of descriptive predicate terms in so­ called "improved', or 'clarified', or 'perspicuous' languages ; (b) the distinction between subject and predicate terms based on the asymmetry of the predication relation ; and ( c) the claimed ontological significance of the different categories of terms implied by (a) and (b). Nominalism , in one of its most pervasive recent forms, has involved the denial of the criterion of ontological commitment embedded in (c) by explicitly or implicitly adopting the criterion expressed in Quine's formula 'to be is to be the value of a variable '. To avoid the obvious charge that nominalists merely ignore abstact entities by the arbitrary ploy of changing the rules, i. e. denying that ontological commitments are made by the inclusion of primitive predicates in schemata of certain kinds by simply employing a different criterion of commitment, some nominalists have sought to argue for their criterion by pointing to the distinction between singular and general terms and the radically different roles such terms play. The distinction between singular and general terms becomes the premise for an argument that purportedly supports Quine's criterion of ontological commitment . This criterion, in turn, provides the basis for the nominalist's use of primitive predicates, as general terms, without ontological commitment to abstract entities. In this paper I shall argue that the nominalist's gambit is inadequate in that the distinction between singular and general terms, as employed by a philosopher like Quine, merely provides a way of stating the nominalist's position and does not provide a reason for holding such a position. To put it another way, if we consider the contemporary nominalist to argue from the distinction between singular and general terms to the cogency of nominalism, since the former provides a ground for Quine's criterion of ontological commitment which, in turn, provides the basis for the latter, then the line of thought is question begging. It is so in that the very way the nominalist draws the distinction presupposes a nominalistic view, since a careful statement of that distinction amounts to a restatement of the nominalistic position. Quine 's use of the terms 'singular' and 'general' involves a set of related notions: names, concrete, abstract, individual, entity. A nominalist, according to Quine, is one who restricts the values of the variables of his

133

'language' or 'scheme' to individual, concrete entities. A platonist is one who allows abstract entities as such values. The nominalist, then, only 'refers' to concrete individuals. Thus, if he admits any names at all, such names could only be names of concrete individuals. What we shall see is that these basic notions, consistently used, merely reflect a nominalistic ontology. Quine's conceptual apparatus is not what it appears to be: a set of conceptual tools for the statement and analysis of alternative ontologies and, in terms of which, one may argue for or against various alternatives. Rather, the mere use of such notions amounts to the acceptance of an ontological position. Hence, Quine provides us with terminological stipulations in lieu of an argument or an analysis.

1. General Terms and a Vicious Circle

The distinction between the singular and the general sometimes seems to hinge on the notion of "naming" : The distinction between singular and general terms is more vital from a l ogical point of view . Thus far it h as been drawn only in a very vague way : a term is singul ar if it purports to name an obj ect (one and only one) , and otherwise general . Note the key word 'purports' ; it separates the question off from such questions of fact as the existence of Socrates and Cerberus. Whether a word purports to name one and only one obj ect is a question of language , and is not contingent on facts of existence . 1

The phrase "question of language" is not the most precise of expressions. One thing that could be meant is "question of syntax'". Thus, we might interpret Quine's distinction between naming andpurports to name in the following manner: the former is a semantical relation, naming being a semantical matter ; the latter is a syntactical property, purporting to name being a syntactical matter. The singular terms of a clarified language would be those with a certain syntactical property which they would have irrespective of whether or not they actually named anything. Such a syntactical characterization, whatever it may be , is not sufficient for Quine 's purposes. It must also be understood that the schema will be so interpreted that terms of that kind, when they refer, will refer to one and only one thing. Or, perhaps better, terms of that kind will be terms that may refer to one and only one obj ect. Thus, the notion of "purporting to refer to one and only one entity" involves both a syntactical characterization and an interpretation rule. With such an interpretation rule implicitly understood we may think of terms of a certain syntactical kind as being those which purport to name irrespective of whether they 134

do or do not in fact name . Those which did in fact name might then be thought to have in addition to the syntactical property in question a semantical property expressed by ""being a name''. Non-logical terms of the schema that have the syntactical property will be called the singular terms of the schema. Other non-logical terms may, for the time , be characterized as general terms. It is not clear from the above quotation if Quine is saying that the general terms of the schema will stand in the same relation to objects that singular terms do , but will stand , or ""purport to" stand, in that relation to more than one object. For , he could be claiming that general terms do not stand , or ··purport to" stand , in that relation at all. If he is making the latter claim then the syntactical property which would characterize general terms would correspond to a semantical relation that such terms could stand in to many objects. The semantical relation, which semantically general terms would stand in to obj ects, would thus not be the naming or referring relation which singular terms stand in to obj ects. This seems to be the import of The general term may indeed '•be true of" each of many things , viz . , each red thing , or each man , b ut this kind of reference is not called naming ; "naming'' , at least as I shall use the word , is limited to the case where the named obj ect purports to be unique . 2

It is clear that Quine is not merely claiming that a term is general ifit in fact refers to more than one entity. That is, one could not take him to hold that there is a relation between the non-logical signs of the schema and objects, reference or naming, such that if a sign stands in that relation to one and only one obj ect it is a name, while if it stands in this relation to more than one obj ect it is not a name but a general term. This will not do since it is not simply a case of a term in fact standing for one and only one object. Notice that in the last quotation the named object is said to 'purport to be unique'. What can it mean for an object to 'purport to be unique'? We are obviously dealing with a metaphorical way of stating what I called the implicit interpretation rule. Some signs, it is understood, will be coordinated with one and only one object. Such signs purport to name one and only one object, and the objects they are coordinated with purport to be unique. General terms will either be coordinated to many objects or will stand to objects in a relation that is not based on being coordinated or assigned to objects. Thus, consider a domain (D), of three objects, a, b, c. Assume that the three signs, 'a', 'b' , 'c', are assigned to the objects and that we employ the terms 'F ' , ' G', 'H', in a framework of standard formation rules so that 'Fa', 'Ga' , 'Ha' , etc.are wffs.We may do two things, relevant here, with the 'predicates'. We may assign them to 135

objects so that 'F' is assigned to a, b, c, 'G' to a and b , and 'H' to a. On the basis of such a coordination we may then say that

Fa, Ga, Ha, Gb , etc. are true sentences, while Gc, Hb, Hc are false sentences. In allowing for the formation of 'Hb' along with 'Ha' as wffs we can state what we mean by saying that 'H' does not purport to uniquely refer to a, even though a is the only object coordinated with 'H'. What this means is that a term is general if it is a predicate in the standard sense, while a term is singular if it is a name or zero-level constant (for the time I ignore definite descriptions). But we clearly have not explained the difference between names and predicates in terms of the notions of singular and general. We have merely used the latter pair of terms to sum up the standard differences traditionally associated with the former pair. A second thing we may do is hold that the predicates are not assigned to objects but are'true of them'.Thus, in 1) 'a' refers to a 2) 'H' is true of a 'refers to' in (1) and 'is true of' in (2) do not stand for the same relation.( 1) reflects the trivial fact that the sign ' a ' has been coordinated with the object it then stands for. (2) reflects another, non-trivial, type of fact.Just what it is that (2) reflects, we will consider shortly. For the moment, we note that one apparently can now hold that a term is singular if it may stand to an object in the relation indicated by 'refers to' , whereas it is general if it may not so stand. General terms may stand to objects in the relation is true of, while singular terms may not stand in that relation at all. Thus, we can apparently explain the difference between singular and general terms, and hence the difference between subject and predicate terms, by means of the two relations refers to and is true of. Singular , or subject, terms may stand in the first but not the second relation; generaL or predicate, terms may stand in the second but not the first relation. The use of'may' is crucial.If we think in terms of signs that infact stand in the relations involved, we preclude empty predicates (and, possibly, "unit' predicates).Moreover, there is even an ambiguity in the sense of 'may' , since a contradictory predicate may not stand to any object in the relevant relation , yet it may occur in a wff as a predicate. A name or definite description may not even occur as a predicate in a wffand, hence , may not 1 36

stand to any object as predicates stand. We deal with the 'stronger' sense that is embodied in the formation rules and not that reflected by logical truths and falsehoods. The distinction , while crucial for some issues, need not preoccupy us here. Having noted it, one may suggest that predicate signs be taken in pairs , such as 'H' an d · ----- H' or, perhaps, 'Hx' and · ----- Hx' , and claim that both are general in that at least one ofthem does in fact stand in the is true of relation to at least one object. In a way, this move attempts to explicate one sense of 'possibility', that reflected by the formation rules, in terms of another sense, that connected with the notion of " logical truth·.But, we obviously presume that general terms come in pairs, whereas singular ones do not.However we may put it there are two points revealed by the difference between •a' is true of a and •H' is true of b, where both are false, but for radically different reasons, and the difference between •a' refers to a and ·H' is true of a, where both are true, but, again , for radically different reasons. The two pairs indicate, first, that we deal with two relations in which terms stand to objects and, second, that certain terms can stand and others cannot stand in the relevant relations. Thus, we face two questions. Given that we know how ·refers to' is used, in that we know what it is to assign a term to an object, how are we to understand 'is true of? Why is it that some terms can stand in one relation to objects but not in the other? Before taking up these questions, consider one objection to the above discussion on the ground that I have ignored definite descriptions by speaking of coordinating names.For, a description also 'purports' to uniquely refer, but it is not "coordinated' to one and only one object . Consider '(1 x)8x' and the context 8x&(y) (8y :J y= x). 1 37

Such a'predicate' can apply to only one object. Hence, if it applies at all it applies to only one thing. Yet, the predicate does not purport to refer to one and only one thing . Given this fact we may say that ('x)0x

purports to refer to one and only one thing since'('x)0x' is a subject term and not a predicate. Just as '0x&(y)(0y => y = x)' purports to uniquely apply since if it applies at all it applies to one and only one thing, so'(' x)0x' purports to uniquely refer since if it'refers' at all it refers to one and only one thing.Thus, though grammatically, a description, like'('x)0x', and a name, such as 'a' , are of a kind, in that both may be juxtaposed with predicates to form sentences, there is a difference in the sense in which they may be said to refer. But, clearly, to speak of '('x)0x' as singular since it purports to uniquely refer presupposes the basic distinction between subjects and predicates in a twofold way. First , all one means by asserting that'(1x)8x' refers to one and only one thing is that '0x' applies to one and only one thing, just as all one means by saying that '('x)0x' purports to uniquely refer is that'0x & (y) (Sy => y = x)' purports to uniquely apply. Hence, we make implicit use of the predicative sense of 'apply' as well as the distinction between a predicate applying and a term like '('x)0x' that may not be said to apply. Second, if we treat descriptions in Russell's manner then we take them, in context, to be used as parts of sentences that are elliptical for sentences with existential quantifiers. Such sentences obviously embody the subject-predicate distinction in terms of distinguishing subject and predicate place and predicate terms from quantified variables in subject-place. Thus, one cannot hope to explicate the difference between predicates and 'names' by appeal to the dichotomy'singular-general' , if one is taking a definite description as a prototype of a singular term. We can now return to the questions raised above. Consider a typical platonist's answer to those questions. A general term or predicate is true of an object if the object exemplifies the property that the predicate names or refers to. A name cannot be true of an object since a name does not stand for a property.In short, the crucial difference between singular terms and predicates is based on the different sorts of things the terms stand for , and that difference, in turn, rests on the asymmetry of the exemplification relation or nex us or tie. The nominalist must explain his use of 'is true of' without mentioning properties, or other 'abstract' objects.He can attempt to do so in one of two ways.He can hold that is true of is a basic relation between signs and objects or he can hold that is true of is to be understood in terms of satisfaction and the consequent claim that an object , say a, satisfies a predicate (or open 138

sentence), say 'Hx ' , if and only if a is H. The first gambit has the absurd consequence that a sentence , say "Ha', will be true in virtue of a relation between a sign and a thing. The second gambit avoid s the issue by using the predicate (or its metalinguistic correlate) in a predicative way to explain what it is for a predicate to be true of an object. Thus, the second gambit uses the familiar redundancy of the predicate 'tru e' to avoid talking about properties (or classes) . Such objections suffice, I believe, to rebut the nominalist who appeals to general terms, but I do not wish to pursue them here. Here, I will argue that granting a nominalist his response , his explanation is patently circular. The nominalist we are discussing holds that a general term is one which can be true of something, while a singular term is one which can refer to something. If he is asked to explain this distinction, he may do one of three things, depending on which of the two ways he construes is true of. First, he may hold that there is really nothing to be explained, in that we just recognize two basic relations between signs and things which neither require nor are capable of further explication. Second, he may hold that singular terms just differ from general terms in that the former, but not the latter, may stand for entities. Third , he may claim that there are two relations but one, is true of, may be explicated along the lines indicated above by appealing to the Hredundancy" of 'is true' . The second reply is obviously question-begging in that one outrightly assumes what he supposedly is arguing for. The first reply forces the nominalist to recognize that he does something the platonist does not have to do, for the latter recognize s only one basic relation between simple descriptive signs (names and primitive predicates) and obj ects- reference or naming. For the platonist such predicates refer to attribu tes as names refer to objects. This reveals that in one sense the nominalist does not offer a simpler gambit than the platonist, since the nominalist acknowledges two fundamental relations between signs and things, while the platonist recognizes two fundamental kinds of ' thing' but only one connection between signs and things. Yet, the platonist must also recognize a relation between his two kinds of entities, particulars and properties. But, even this feature of the platonist's view is matched by the nominalist, for the latter does not merely assign predicates to particulars. Nor does he really take is true of to be basic. If he does he cannot specify the truth condition for '8' being true of a that such nominalists characteristically appeal to. What such nominalists generally do is hold that '8' is true of a, since a is 8 . Thus, the first reply merges with the third. Were they to really take is true of as a basic relation, they would either (A) absurdly assert that we have two different connections between signs and the objects they are assigned to and one such assigned connection is the ground of truth, or (B) hold that predicates stand in a relation to objects that

139

accounts for the truth of atomic sentences irrespective of our assignment of predicates to objects. The nominalist avoids both such absurdities by echoing Tarski's Convention T. He holds that a predicate '8' is true of a particular on the condition that the particular is 8 . Thus, the nominalist assumes that he may use the predicate ascriptively without assigning it to any entity. This, of course, is precisely what is at issue. Hence to counter the platonist's obj ection , the nominalist must assume the view he purports to establish. By recognizing attributes, the platonist is not forced to use the predicate in his explanation in the way the nominalist must. If we ask the platonist why'Sa' is true, he responds that the obj ect which 'a' refers to has the property which '8' refers to. The nominalist holds that 'Sa' is true because'8' is true of a, which is so since a is 8. He thus uses'0' in his explanation and, in so doing, he not only reveals that his gambit is no simpler than the platonist's but that to give the explanation he does give he must assume what he purports to prove: that general terms may be used without commitment to attributes.3 The nominalist's circular exposition is disguised by being presented in a different manner. Singular terms are those which can refer, but they are also those which can replace variables or are at a'level' of language where we have quantifiers and variables.Thus, Quine's criterion of ontological commitment connects with his discussion of singular terms and reference. We have two senses of 'singular term' : (S 1 ) A term is singular if and only if it can refer to something (S2) A term is singular if and only if it can replace a variable or if it may occur in a position where a variable may occur. Quine's conception of ontological commitment involves the assumption of (S 1 ) and (S2 ) . 4 But, what this means is that if we have a logical schema limited to lower functional logic , the constant predicates included in such a schema are not taken as referring to "abstract' entities.But, they are not so taken since we assume that only singular terms may make ontological commitments.A specific singular term will in fact involve an ontological commitment if existential generalization is permitted (or made) with respect to it. Thus, from (S 1 ) and (S2 ) we get : (S3) A term can refer to something if and only if it can replace a variable. Moreover, understanding that (S3) implies that a singular term is in fact taken to stand for something if and only if it is generalizedfrom, we arrive at the relevant theme in Quines's criterion of ontological commitment. We then conclude that general terms do not, by their use, involve us in ontological commitments.But, we arrive at such a criterion by assuming 1 40

the 'criteria' of (S 1 ) and (S:J Hence, the distinction between singular and general terms reflects, rather than supports, the nominalist's position. Strawson has also criticized Quine's attempt to draw the distinction between singular and general terms the way he does.5 He has held that the distinction is to be based on the 'identifying' referential function of singular terms and of the application of the term in predicative position to what is thereby identified.This difference is further held to presuppose the type distinction between the two kinds of terms. Strawson thus acknowledges the fundamental asymmetry of the predication relation and recognizes the two basic notions of reference and application. Yet, a puzzling inconsistency prevents his resolving our problem in a straightforward manner. On the one hand he speaks of predicates signifying attributes.6 On the other hand, after criticizing Quine, he paradoxically accepts the latter' s basic claim regarding general terms and ontological commitment. For, Strawson holds that we recognize attributes in that we transfer the pattern of identifyingly referring to them while applying higher type predicates to them.7 As it is this 'analogy' with particulars that is behind the recognition of attributes, it is clear that we recognize attributes associated with general terms only in so far as we employ such terms as "subject' terms in sentences with higher type predicates. This is the theme central to Quine's nominalism. That the nominalist's view is merely reflected, rather than supported, by the distinction Quine draws is also suggested by the fact that only in the lower functional calculus does the distinction separate general and singular terms into exclusive classes.In an extended functional calculus predicates will satisfy (S2 ) and, hence, be singular.What will distinguish a singular term like 'a' from a term like '8' , which is both singular and general, is simply that the latter is general while the former is not.Thus, one may remark that there is nothing to fuss about ; the nominalist of Quine 's type simply recognizes that in platonistic schemes some terms are both singular and general, i. e. , the predicates. This merely means that our criteria for singular and general must be such as not to prevent a term from satisfying both.Thus, to (S 1 )-(S3 ) we add (S4 ) A term is general if and only if it may be true of things. and, by so doing , we do not preclude a term from being both singular and general.Yet, while we do not wish to preclude predicates from being both singular and general in extended functional calculi, we do wish to deny that the singular terms of a lower functional calculus may also be general. But this clearly reveals that Quine , like Strawson, simply repeats the traditional distinction between singular and general terms. Quine proceeds to stipulate that the distinction is the ground for taking (S3 ) to be 14 1

the criterion of ontological commitment in a lower functional calculus. Aside from such a stipulation all the nominalist has done is to reiterate , in a complex and unclear manner, the basic asymmetry of the predicative relation and the theme that only predicates may be coherently predicated. By so doing, the nominalist fails to provide the argument that he implicitly promised. Quine ' s discussion thus becomes a declaration, rather than a defense, of nominalism. But this is not the only problem with Quine's'defense' of nominalism . The taking of predicate terms to be both singular and general reveals further defects in his gambit.

2. On Being Both Singular and General

Quine has written: I attach much importance to the traditional distinction between general terms and abstract singular terms , 'square' versus 'squareness ' , because of the ontological point: use of the general term does not of itself commit us to the admission of a corresponding abstract entity into our ontology ; on the other hand the use of an abstract singular term , subj ect to the standard behavior of singular terms such as the l aw of putting equals for equals , flatly commits us to an abstract entity named by the term . 8 . . . Though we do not recognize the general terms 'dog' and 'white' as names of dogkind and the cl ass of white things , genuine names of those abstract entities are not far to seek , namely, the singular terms 'dogkind' and 'the class of white things' . Singular terms naming entities are quite properly substituted for v ariables which admit those entities as values . . . 9

He apparently holds that we should have, in a perspicuous platonistic schema, as in natural languages, two terms, one distinctively singular, but not general, and one general. If 'green' is the English transcription of a descriptive constant predicate in a schema with predicate quantifiers and variables (and if we, in fact , existentially quantify from it) , our 'discourse' commits 'us' to an abstract entity. Therefore , we should have another term, transcribed by "greenness' , which would be a distinctively singular (and not general) term and which would refer to or name that entity. It is because the predicate or general term is in the range of the quantifier, and hence 'singular' as well as general , that we are committed to that entity. Why, then, aside from the clues of natural languages, introduce a new term ? The concern goes back to Frege. Frege held that a term could not function predicatively and stand for an 'object', while a term in subject place did not function predicatively and did stand for (or purport to stand 142

, for) an obj ect . Thus , ·green in · Green is a color' stood for an object which it did not stand for in This is green· . In the latter sentence it stood for a function . 1 0 Thus , the term stood for different "things' in the two cases , and , hence , is ambiguous as a term . To avoid the ambiguity one could employ two terms , one to be used as a predicate , say ·Gx' , and one to be used as a subj ect term , say " Gx' . "Gx' refers to a function ( or concept) while " Gx' refers to a concept correlate which is, in Frege's sense , an object . There has always been a mystery about Frege 's concept correlates. There need not be when we recall a familiar Fregean theme . For Frege , the sentence (S 1 ) ·a is green , is taken to be 'equivalent' to (S 2 ) 'a , falls under th e concept green , where the phrase ' the concept green' must be taken to stand for an obj ect . Thus , we can transcribe (S 1 ) and (S 2 ) , respectively , as

and

with "F for "falls under' . In short , when , and only when , (the function) Gx maps the obj ect a onto The True, (the function) F(x,y) maps the objects a and Gx onto The True. Or, to put it in a distinctively non-Fregean way , the obj ect a is green if and only if the obj ect a stands in the exemplification relation to the obj ect Gx . The exemplification relation is asymmetrical . Hence , even with Gx taken to be an obj ect or individual , as a is , there is a fundamental difference between obj ects like a and obj ects like Gx . The function F(x ,y) separates the obj ects into two distinct classes that are quite familiar: those obj ects which may and those which may not have other objects fall under them . The Fregean pattern is adopted by the nominalist of Quine's type to characterize the platonist . A platonist is held to take 'Gx ' as a singular term , which refers to an entity and , consequen tly, to adopt 'Gx' in place of or in addition to "Gx' . 1 1 Quine's discussion reveals that he takes existents and individuals to be one and the same. Greenness is an existent in that it is an individual, albeit an abstract one . It differs from a, not in that one is an individual while the other is not , but in that one is a concrete individual while the other is an abstract object. Both 'names' - 'a' and ' Gx ' - are zero-level constants in a Principia style schema and both can thus stand as subject terms for first-level predicates . It is as if the nominalist only recognizes the platonist 's entities if the latter are construed , in a fundamental way , as the nominalist 's individuals. This is indeed peculiar when we recall that the original issue was whether we need recognize 143

entities other than individuals. The platonist held we did since we require primitive predicates. The nominalist held we did not since we mere ly need use 'dummy' predicates. What we see is that the nominalist rej ects the platonist's criterion of ontological commitment since the nominalist insists that he need only recognize attributes if they are taken to be individuals. But, the issue is not whether there are abstract individuals as well as concrete individuals: the issue, to repeat, is whether there are universals in addition to individuals. The nominalist avoids universals by transforming the question in a question-begging manner. His concern with singular terms, 'greenness' and 'Gx', is symptomatic of this weakness. The nominalist runs together the claims that

(N 1 ) Only individuals exist, hence attributes do not exist and

(N 2) Only concrete individuals exist, hence abstract individuals do not exist.

He thereby equates an argument for (N 1 ) with an argument for (N 2) by adopting Quine's criterion of ontological commitment. For, he insists that the platonist must claim that 'a is green' is to be transcribed, in a perspicuous schema , by ' F(a, Gx)' and not by ' Ga' . By holding that 'F(a, Gx) ' is not needed, since 'Ga' will do, the nominalist 'establishes' (N 2) and, as he takes it, (N 1 ). There is a simple theme that aids the nominalist's line. Frege and Quine emphasize the role of the law of identity: satisfying it is indicative of 'being an obj ect', for Frege , and of 'being an entity' for Quine. Identity is generally represented by a predicate, say " = '. He nce, an implicit argument emerges. If 'Gx' stands for an entity (obj ect) then that e ntity (obj ect) must be self-identical. If so, to express that by Gx = Gx is problematic, for we have an 'incomplete' expression, so we must use Gx = Gx. Thus, we recognize the use of 'Gx' as a subj ect term standing for the appropriate entity (obj ect) . It is worth recal ling that Wittgenstein questioned the force of representing identity and difference by true sentences, employing such a predicate or its contradictory, and thought of such matters being shown by an appropriate symbolism. Such a 1 44

symbolism would have one primitive sign for one thing and, hence, represent different things by different signs. Along this line we might examine an 'argument' similar to the nominalist's implicit one appealing to identity. Suppose a nominalist argued that the platonist must recognize an attribute of being an attribute, common to all attributes, and an attribute of being an individual, common to all individuals. Then, he must allow for self- predication , or , least , he must treat predicates as terms capable of occupying subject place in sentences. If one does not so treat predicates, he is not a platonist and does not , therefore, acknowledge attributes. We might observe, along with Wittgenstein, that being an attribute is not itself an attribute among attributes. Consequently , such a 'formal property' is perspicuously 'represented' by features of the schema, and not a predicate in the schema. In this case the relevant features are the division of terms into predicates and those which can only occupy subject place along with the appropriate formation rules for wffs. Similar remarks are relevant to the suggestion that a relational predicate transcribing ·exemplifies' is required and , hence, forces us to acknowledge a Bradleyian regress if we hold that things exemplify attributes. The nominalist might suggest that talking of a singular term, in the sense of one that will only occupy subject place and that will refer to the abstract entity (the attribute) which the predicate commits a schema to (with predicate quantification), is only a way of reflecting the distinction acknowledged in ordinary language by the use of "greenness' and 'green'. But , there is more to it. The very way the nominalist puts matters reveals that he is thinking of singular terms as differing from general terms in that the former are only used as subject signs and the latter are only used as predicates. The basic distinction between such terms, which supposedly provides a basis for Quine 's form of nominalism, then clearly amounts to no more than a way of stating the nominalist's position , if the latter adopts Quine's criterion of ontological commitment. Moreover , two further points become clarified if we seriously consider the use of distinctively singular terms like 'greenness' . Suppose we add a correlate of such a term, say 'Gx', to a schema with predicate quantification. We may then assert '(3f) (f= Gx)' or, perhaps, '(3x) (x = Gx) ' , as well as 'Ga :::J (3f)fa' , and, hence, make an ontological commitment to the attribute. The first point to note is that we no longer require predicates like 'G '. To say that a is green we may now assert, in Fregean fashion , that a stands to Gx in a relation - falling under or exemplification. In short , we may use (T2 ) in place of (T 1 ) , and this we may do for every standard predicate. Just as Quine once suggested that we may introduce a predicate, say 'Pegasizes', by holding it to be understood in terms of 'being identical with Pegasus',

145

and then make use of the description 'the x such that x Pegasizes' in place of the name 'Pegasus', so we may replace the standard uses of the predicate 'G' by employing the distinctively singular term 'Gx:'. Thus, the platonist, on the nominalist's view, would have only one distinctively general predicate term, the term for the exemplification relation, say 'F'. Even ordinary relations would be construed as individuals so that 'a is to the left of b', for example, would be expressed by F(a, b, Lxy) with 'F' understood to be 'multigrade' , as Quine and Goodman have put it. (Alternatively, we could recognize many 'exemplification' relations, each for a different number of terms. ) If one then sought to treat exemplification itself as an individual, we can see a simple way of expressing Bradley's famous 'paradox' . We could not transform 'a exemplifies greenness' into F 1 (a, Gx:, Fxy) without introducing a new general term, 'F 1 ' , or allowing for the retention of 'F' along with 'Fxy'. If we then understand platonism as the claim that we have two kinds of individuals, both represented by 'singular terms' and distinguished by the asymmetry of the exemplification relation, Bradley's 'paradox' reveals that platonism is self-refuting. Exemplification cannot be solely represented by such a term and, hence, cannot be taken as an 'individual' or 'existent' . Platonism can only be maintained if we recognize that there are things other than individuals, in the nominalist's broad sense of individual. One may then hold that the platonist, following a cue from Quine, may say that he need not recognize a relation of exemplification, and merely use a general term to express (but not refer to) such a relation. But this blatantly violates the platonist's pattern. Alternatively, we may recognize that the platonisf s view cannot even be cogently expressed in terms of the nominalist's notions of singular and general term and his criterion of ontological commitment. Thus, in yet another way, the nominalist begs the question. There is a second point. Quine makes much of the distinction between concrete and abstract individuals. His distinction reflects the nominalist's special use of "individual'. One might take 'individual' simply to indicate any entity in that a thing is what it is and , hence , individual. For Quine an individual, in effect, is what may be referred to by a subject term . Thus, a and Gx: are individuals, but the former is concrete while the latter is abstract. The distinction is sometimes put in terms of concrete individuals being in space and time. We may, for our 146

purposes , forget the temporal aspect and concentrate on the spatial question. (To aid in that we may , as some put it, consider a "time-slice'.) What is it to be in space? One simple and standard response is in terms of having spatial properties or standing in spatial relations . Just as a is green, it is to the left . of b etc. With a set of first-level predicates , such as 'Lxy' for "is to the left of . we may then note that the names •a' and "b' may replace the variables in " Lxy' , but that predicates like "G' and "B' , for 'blue' , may not. Thus . properties like being green and being blue, not being capable of standing in relations like Lxy , are not in space or spatial "things'. Even if we quantify over predicates and hence take such attributes to be entities , via the nominalist's criterion , this difference is preserved. But suppose we make use of the singular abstract names , such as 'Gx' , 'Bx', 'Lxy', etc. We might then use F(a, b, Lxy) F(a, Gx) F(b, Bx) to express that a is to the left of b, that a i s green , and that b is blue. May we also have

as meaningful and true? There is presumably no syntactical prohibition against (T3) as there usually is agai nst 'L( G ,B)' , since not all the terms in (T3 ) are of the same 'type'. One can point out that if we allow for (T3 ) , then standard spatial 'laws' will not hold, since we could and would have an object between itselfand itself. Thi s reflects the fact that we recognize two radically different sorts of objects or individuals, not that one is in space and the other is not. Of course, one may stipulate that to be in space is to obey certain spatial 'laws '. But this is hardly illuminating. Moreover , suppose in fact , that no green object was to the left of another or between two other green objects. We would still hold that one could be and hence that greenness could be to the left of greenness , etc. , whi le no green object could be to the left of itself, etc. The reason is that greenness is an attribute (or property or universal) and not a particular . We may explain why certain spatial relationships are possible , and others not , in terms of the difference between particulars and properties. Paradoxically , if we treat properties as particulars , along the nominalist's lines , the distinction between the concrete and the abstract, in terms of which the nominal ist seeks to explai n the difference between properties and particulars , i s lost. It i s lost since the nominalist is forced to hold that both an abstract object l ike Gx and a concrete (?bject like a are in space. Moreover , they do not 147

differ in that certain spatial laws hold for the concrete, but not for the abstract objects. For, such generalizations do not suffice to reflect 'the fact' that an abstract object like Gx may stand in spatial relations that concrete objects like a may not stand in. The difference between concrete and abstract obj ects must be put in terms of what may and what may not occur. (Recall the earlier discussion of Quine' s use of 'purports' . ) It thus reduces to the familiar categorial, or 'logical' as some would put it, distinction between particulars and attributes that is usually reflected by the syntactical categories of a Principia type schema. The nominalist does not then explain the latter in terms of the former; he does the exact opposite. If he seeks to rebut the objection by pointing to the difference between the signs 'a' and 'Gx', and the formation of the latter from a predicate, then he clearly reveals that he has merely adopted a different terminology for a familiar distinction: that which may and that which may not be attributed . The only difference is that the nominalist, by forcing the platonist to use 'Gx' and 'a' as singular terms, does not permit the perspicuous representation of the difference by the type distinction. 1 2 The point can be put simply when we consider 'F(Gx,Gx)' and 'F(Gx ,Bx) '. The latter must be held to be either false or meaningless; the former 'can' be taken to be true, or false, or meaningless. In any case one must settle the matter by stipulation. The stipulation is based on the fact that terms like 'Gx', and 'Bx' stand for attributes. Consequently, it is clear that one who insists on introducing such terms takes the distinction between concrete particulars, like a, and abstract objects, like Gx and Bx, to be basic. He does not explain the difference in terms of a being in space while Bx and Gx are not, since attempting to do so would require that he explain why 'F(Gx, Gx, Lxy)', etc. , are false or true or meaningless. Whatever answer he gives will be based on Gx being an attribute- a kind of abstract object. 1 3

Notes 1

2 3

W. V. Quine , Methods of Logic (London : Routledge & Kegan Paul , 1952), p. 205 . Ibid. , p. 205 Quine later appears to take the grammatical distinction between subj ect and predicate terms and the asymmetry of the predication relation as basic linguistic categories. If we understand the related on tological question to be a query as to what such lingui stic categories reflect , Quine 's answer clearly involves an appeal to is true of and the use of the relevant predicate in the purported explanation. "Such talk of purport is only a picturesq ue way of alluding to distinctive grammatical roles that singular and general terms play in sentences . It is by grammatical role that general and singular terms are properly to be distinguished.

148

4

0 7

8

9

10

11

12

13

The basic combination in which general and singular terms find their contrasting roles is that of predication . . . Predication j oins a general term and a singular term to form a sentence that is true or false according as the general term is true or false of the object , if any , to which the singular term refers. " W.V. Quine . Word & Object, (Cambridge: MIT Press, 1960), p. 96. In fact Quine sometimes explicates (S 1 ) and his use of ·purports' in terms of (S, ) . "In terms of logical structure , what it means to say that the singular term 'purports to nime one and only one obj ect' is just this : The singular term belongs in positions of the kind in which it would also be coherent to use variables ·x·. •y ', etc . " - Methods of Logic, p. 205 . P. F. Strawson , 'Singular Terms and Predication' , Journal of Philosophy, 58 ( 1961 ) : 393 -412, reprinted in Philosophical Logic, ed. P. F . Strawson (Oxford : Oxford University Press, 196 7) . pp. 69-88. Ibid. , p. 88. Thus, Strawson writes: "But to say of things of other types that they also answer to this description is simply to say that we have occasion to bring such things under higher principles of grouping , principles which serve to group them in ways analogous to the ways in which expressions signifying properties ( or kinds) of particulars serve to group particulars". Ibid. , p. 88. W. V. Quine . 'Identity . Ostension , and Hypostasis' , in From a Logical PointofView, 2d ed. (Cambridge : Harvard University Press , 1964) , p . 76 . W. V . Quine . ' Reification of Universals' , in From a Logical Point of View, pp. 1 13- 14. Frege 's discussion of concepts and functions involves a highly problematic ambiguity that is related, but not directly relevant, to the present discussion. See my 'Frege on Concepts as Functions: a Fundamental Ambiguity ', in this volume , pp. 48 f. In a series of interesting papers Nino Cocchiarella has examined features of calculi that include such ·nominalized' predicates. See his 'Whither Russell's Paradox of Predication'? in Logic and Ontology, edited by M.K. Munitz, (New York: NYU Press , 1973) , pp . 133-58; 'Properties as Individuals in Formal Ontology' , Nous, 6 ( 1972) , pp. 165-87 ; 'Fregean Semantics for a Realist Ontology' , Notre Dame Journal ofFormal Logic, 15 ( 1974) : 552-68. Talk of abstract obj ects , such as numbers, which need not be construed as attributes ( or classes) does not affect the issues at stake here . The point here concerns the way in wh ich one distinguishes objects like a from attributes like Gx . If one introduces numerals in the same syntactical category as 'a' and 'Gx: ' , he will have to make additional stipulations about the meaningfulness and truth values of 'F(8 , Bx)' , 'F(7 ,8, Lx:y)' , '- F(8, Gx) ' , etc. lfone is a logicist , as Quine is, there need be no significantdifference between numbers and attributes (or classes). For a discussion of numbers as attributes, rather than classes, see my 'Properties, Abstracts, and the Axiom of Infinity' , in this volume, pp. 339 f. The emptiness of the distinction between singular and general terms for purposes of ontological analysis becomes quite clear when we recall that 'Gx:' will only be formed when 'G' occurs in an extended functional calculus. Thus, we would have properties of properties, and, hence , abstract individuals of different 'types'. We would thus require the reflection of such differences by ordering the different types of abstract objects to avoid further absurdities on the order of 'F(Bx:, Gx:) ' . One will also face a host of problems and 'stipulations' with respect to complex predicates: for example, are '-F(a, Gx:)' and 'F(a, -Gx)' logically equivalent? For a discussion of some related questions in a 'logic' of properties, see 'Properties, Abstracts, and the Axiom of Infinity'.

Nominalism , Platonism and Being True of

One attempt to avoid a platonistic ontology involves using the notion 'is true of'. 1 Using this notion, one claims that a predicate, say 'white', is true of a particular object whereas a sign such as a proper name refers to the object. This is done to avoid holding (1) that a predicate refers to a universal property while the proper name refers to an individual and (2) that a sentence used to predicate the one sign of the other indicates the fact that the individual exemplifies the property. Thus, an attempt is made to offer a coherent ontology recognizing only individuals. If cogent, this gambit provides a more parsimonious ontology than platonistic alternatives. Here I shall argue that the gambit fails. A key point will be that an ontology is reflected in our use of a language not only by the signs we explicitly hold to be connected with objects but also by the ways we connect signs to things. Let 'a' and 'W' be a proper name and a predicate respectively.Let "R 1 stand for 'refers to' . On what I shall call the platonistic account the signs 'a' and 'W' both stand or are put in the relation R 1 to an individual and a color respectively.They are, as some say, interpreted or paired with such objects. On the alternative account the name 'a' is also thought of as referring to the particular but the predicate is not taken to refer to anything.To take the predicate as a mere label of several objects just as the name is a label of one is to propose a different version of nominal ism, one that is sometimes referred to as the doctrine of "common names'.On the present nominalistic gambit the predicate is taken to stand in a different relation to the particular than the name stands in, the relation of being true of.2 Let 'R2 ' stand for this latter relation. On the platonist's view we have the following two sentences expressing the interpretation of the signs 'a' and ·w'

while

may (be used to) assert that a exemplifies W.That is, on the platonist's analysis the sentence S3 may be taken to assert that a particular 1 50

exemplifies a property, since "a' and ·w' are taken to refer (be tied), respectively, to a particular and a universal property. 3 Thus, while one may observe that "Wa' is used to assert simply that a has W or a is W, the point is that on the platonist' s analysi s the fact that a is W is construed in terms of a relation between a particular and a universal. The nominalist, by contrast, seeks to construe what may be called the·neutral sense' of "a is W' in terms of an alternative analysis. 4 Before considering that, we may also introduce, with ·o· for ' designates' and T' for 'true',

to express the fact that the sentence is taken to be true when what it designates, a fact, is the case. In place of (St ) and (S2 ) the would-be nominalist has

The difference in form be tween (S 5 ) and (S 6 ) as well as that between (S 6 ) and ( S., ) reveals in what sense the nominalistic gambit is no more parsimonious than the platonist's. The platonist, if I may so put it, recognizes two types of entity, particulars and universals, and needs only one relation, Rt, between signs and things. The nominalist recognizes only one kind of entity, particulars, but requires two relations between signs and things, Rt and R2 . Moreover, his second relation is radically different from the first. There is an obvious triviality or emptiness about (S t ), (S 2 ) and (S 5 ) . Given that we interpret a sign as a mere label (S 1 ) , (S 2 ) and (S5 ) reflect the red undancy that a label labels what it labels. They can, with appropriate rules, be seen to be true by inspection. But (S6) is quite different. It holds when the individual has the property, and hence does not merely reflect our having interpreted certain signs. Thus, if we consider an ontology to be reflected not only by what entities are held to exist but also by how we connect signs to those entities, the nominalist's scheme is as rich as the platonist's. One might obj ect that the nominalist still has a more parsi monious ontology since he does not require the tie of exemplification which holds between a universal an d a particular. A platonic ontology recognizes two types of entities, one tie between signs and obj ects and a tie bet ween the obj ects, which is the basis for the fact that the particular has the property. Thus, four basic things are involved in his account of correct ascriptions of predicates. The nomin alist, by contrast, requires only three: one type of entity and two ties connecting signs to that type of entity. But the obj ection can be answered by noting that either R2 must be taken as a relation with a two-fold function, in contrast to R 1 , or it cannot fulfill the

15 1

predicative role assigned to it. R 1 serves only to tie signs to objects so that sentences may be used to assert facts.But R2 not only ties predicates to objects, hence fulfilling the interpretive function, it also fulfills the predicative function by tying them to objects in a second way.This we see when we note that (S6 ) does the jobs of both (S2 ) and (S3 ) ; to assert 'Wa' is to assert that 'W' is true of a . The platonist asserts (S 3 ) to state that the individual has the property; the nominalist asserts (S7 )'W' is true of a . But (S 7 ) is (S6) , whereas (S 3 ) is not (S 2 ) . That something has gone wrong is marked by the occurrence in (S 7 ) of the name of the predicate, and not the predicate itself, to ascribe a property to an object. Actually (S7 ) plays a third role in addition to its interpretative and its assertive roles.In view of the use of the term'true' and the occurrence of the name of the predicate in (S 7 ) that statement takes on the role of (S4 ) and the notion of 'designates' as it occurs in the latter.As (S 7 ) is the nominalist's version of (S 3) , it is also his version of (S8 ) 'Wa' is true. The ordinary equivalence of'"W' is true of a" . ·a is W', and "'a is W' is true" covers up the variety of roles of (S 7 ) and the mixing of questions of interpretation, assertion and designation.This mixture is revealed when one asks about the basis or ground of truth of (S 7 ) . To inquire about a basis or ground of truth is to ask what entity (object, fact, feature) in the nominalist's ontology corresponds to the truth of S7 in the sense in which one may say that the fact that a exemplifies W provides a ground of truth for the platonist. To put it slightly differently one may ask what feature it is, in the nominalist's analysis , the existence of which guarantees the truth of S7 . What must there be or what relations must hold among what there is for S7 to be true? The nominalist cannot answer that a has W. For if he takes that answer seriously , i.e., as indicating a fact to be analyzed, he gives up his game by acknowledging a component of the fact other than a . (On the other hand, if he gives that answer without admitting that anything further need be said he gives up the metaphysical game altogether.) He must reply that a is the basis, since particulars are all that he acknowledges in his ontology. But then a will also be the basis for another truth abouth a, say " Ca' , where 'C' indicates the shape of the object. Thus, one and the same thing is the ground for two sentences , which are not deductively related , being true.This may lead us to hold that we have come across a sense in which the particulars of the nominalist are complex.Let me explain. 1 52

On the platonist's view what grounds the truth of the sentence "Wa' is a fact which is a complex whose constituents are the referents of "W' and of •a' in a relation. A different fact grounds the truth of "Ca' , yet both facts contain a common constituent. On the nominalist's view the particular grounds the truth of both assertions, i.e. , its existence is what makes the sentences true. But, for the nominalist , there are no facts as on the platonist's view. (If there were the nominalist would have to distinguish the facts from the particular and explain their structure, without recourse to universal properties or relations. That this cannot be done I will merely assert, but not argue , since the nominalistic gambit we are considering seeks to avoid facts.) Hence, the particular takes over the roles of two distinct facts, each of which is complex, on the platonist's view. Playing the role they do in the nominalist' s ontology, his particulars may be called complex by contrast with the platonist's particulars. Even without contrasting the nominalist to the platonist, one may note that in one sense it is simply the particular's being what it is that grounds the truth of 'Wa', but in another sense it is the particular's being white as distinct from its being circular. Yet if he recognizes an entity as complex the nominalist must say something about its structure. To do so, however, would be to recognize constituents of his particulars. But this he cannot consistently do. For, then it would be either a constituent or an 'arrangement' of constituents which would ground the truth of a sentence like (S7 ) . The former alternative would amount to introducing something other than the ordinary particular and the relations R 1 and R 2 which the nominalist explicitly recognizes. (S 7 ) would then be true in virtue of the fact that a contains such a constituent or in virtue of that constituent being combined with others to constitute the complex object a . Whatever this constituent is one has departed from the gambit and adopted either a different form of nominalism, if the constituent is an individual (a particularized property perhaps - this whiteness), or even platonism, if the constituent is a universal. The latter alternative , grounding the truth of (S7 ) in an arrangement of constituents, differs significantly from the first only if the constituents, as such, make no difference to the truth of (S7 ) , i. e. , if the arrangement alone is relevant. This would amount to a view similar to some interpretations of Wittgenstein's Tractatus which claim that an atomic fact (or even a property like red) is a set of particular objects in an arrangement (configuration) . 5 But such a view is only apparently nominalistic since we then have particular objects being in arrangements (i. e., relations) instead of particulars exemplifying universals. The problem of universals and exemplification breaks out again with respect to such 'arrangements' and their exemplification. To argue this in detail is not necessary here since, in any case, on this alternative the particular a is turned into a sort of fact composed of

153

simpler particulars in an arrangement. Hence, the point of the 'is true of' gambit, which is to avoid the introduction of universals by means of the relations R 1 and R2 and not by the introduction of further entities and arrangements among them, is abandoned. Thus to consider a as complex is to give up this version of nominalism. In a way the nominalist's own formula makes the point in question. The platonist takes 'Wa' to be true since a is W while 'Ca' is true since a is C. (S4 ) would also reflect that different facts ground the truth of the different sentences. By contrast the nominalist simply has (S7 ) 'W' is true of a and

(S9 ) 'C' is true of a where the only sign for a non-linguistic entity is "a', since the predicates 'W' and 'C' are not used but only mentioned. To the question what accounts for the truth of (S 7 ) and (S 9 ) the only possible answer is "a' . 6 But to say that a accounts for the truth of (S 7 ) and (S 9 ) is not to give an account at all. For a could be such that 'W' would not be true of it. Thus it is not a as such that grounds the truth of (S 7) , but a's being white. Notice that for the platonist a also could be such that 'Wa' is false. But since it is the fact that a is W which grounds the truth of 'Wa' , it could not be the case that the ground exists without the sentence being true. For the nominalist, as we have just seen, the ground of a sentence's truth, the particular , can exist without the sentence being true. Thus the nominalist's ground is clearly no ground at all in contrast with the platonist's. In order to take a as the ground of truth for (S7 ) the nominalist must then deny that a could be such that 'W' would not be true of it. He must turn (S 7 ) into a necessary truth. This could be done in two ways , neither of which is consistent with his view. First, 'W' could be defined by extension as a class including a. This is reminiscent of an extreme nominalistic gambit: the attempt to define all predicates in terms of proper names. To do so, aside from all the problematic consequences, is to give up the whole point of using "is true of', since this latter notion is introduced to avoid the futility of the extreme form of nominalism. Second, one could consider 'W' as indicating a constituent of a and hold, in idealistic fashion , that such a constituent of a particular is a necessary constituent since no whole could be what it is without having just the parts it does have. This also gives up the gambit since 'W' then refers, in the sense of R 1 , to something while being true of the particular a. Moreover, what type of entity could 'W' refer to but a property? 154

Aside from the ontological concern with universals there may be another motive behind the introduction of 'is true of'. We may consider that 'refer' is more accurately taken to indicate a relation between signs, sign users and sign referents. As such it may be construed behavioristically. This is just to say that our using signs to refer to objects is a form of behavior. Asserting sentences is also behaving. But whether a sentence is true or false is not a matter of behavior. Yet the assimilation of 'is true of' and 'refer' might aid the further assimilation of both the asserting and the truth of sentences to the behavioristics of language usage. The 'is true of gambit may thus obscure the fact that truth is one thing and human behavior another. This perhaps indicates an affinity for the pragmatic-idealistic conception of truth . Thus there may be more behind or implicit in the gambit than the attempt to avoid a platonistic ontology. Be that as it may, we may conclude, first, that the use of 'is true of fails as a cogent nominalistic alternative and, second, that ontological commitments made through the use of a language are a matter not only of what entities one recognizes the language to be tied to but of what means one employs to tie the language to objects so that sentences may be 'about' them and true. Notes 1

3

4

Quine is one philosopher who adopts the 'is true of' gambit ; W. V. Quine , Methods of Logic (Holt, Rineh art and Winston : 1959) : 65-67. Even though Quine sometimes speaks as if the relation between the predicate and the many particulars is the same as that between a name and the particular it names , it can be shown that he cannot consistently do so . To demonstrate this would require an analysis of Quine 's notions of 'singular term· and 'general term· and take us beyond the scope of this paper . There are, of course, other variants of platoni sm. Thus a platonist may consider a p articular to be a composite of universal properties and hence not interpret or analyze 'a is white' in terms of a relation of exemplification between a particular and a universal property. Rather, it is W being a part of a or a's being a complex obj ect with W as a constituent that then serves in the analysis of 'a is W' ; for example, see my ' Universals, Particulars and Predication', in this volume, pp. 263 f. One may speak of a neutral sense of 'Wa' in that the sign sequence (S 3 ) may be thought to be interpreted, in one sense, into the Engli sh sentence 'a is white· wit hout any concern with an ontological analysis. In this sense 'Wa' merely says 'a is white ' and the latter, in turn , merely says that a is white. But this is not sufficient when one takes up ontological issues. If one does take them up the questi on is then one of interpreting or analyzing the sentence, whether 'Wa' or 'a is white' makes no difference, in terms of an ontological position. Thus, on the platonist's view, the sign 'W' stands for an entity W just as the sign 'a' stands for the object a and the sentence 'Wa ' , if true, is taken to refer to a fact analyzed in terms of a particular, a universal and the tie of exemplification. It is then apparent that while 'Wa' merely says that a is white, in one sense, it may be taken to involve much more, in another sense.

1 55

5 6

One example is G . E . M. Anscombe, A n Introduction to Wittgenstein's Tractatus (London) , pp. 99 , 1 1 1 . Since the predicates 'W' and 'C' are referred to in (S7) and (S9) it is possible to hold that the connection between the predicates and the obj ect a grounds the truth of (S7 ) and (S9) ; but this would be absurd.

Mapping , Meaning, and Metaphysics

The positivistic attack on the meaningfulness of philosophical or metaphysical propositions led to a preoccupation with so-called clarified or perspicuous language schemata. Yet , this aspect of what Gustav Bergmann has aptly called the 'li nguistic turn' in philosophy provided a ground , somewhat paradoxically, for the rejection of the philosophical nihilism that was part of the positivistic legacy. Instead of dismissing metaphysical issues and claims as empty verbiage or mere nonsense , some phi losophers in the linguistic tradition attempted to employ clarified languages , or ideal or perspicuous languages as they were sometimes called , to both restate and resolve the classical problems as questions and insights about the structure and interpretation of such schemata. Unquestionably, one underlyi ng motive for the interest in ideal languages as a key to the restatement and resolution of philosophical problems was the ancient connection between thought and language. In getting at features of clarified languages some linguistic philosophers, explicitly or implicitly , thought they were getting at features of mental processes. Ideal languages thus became a means for studying the connection between thought and the ordinary world of experience , as well as for examining the logical structure of thought processes. Wittgenstein did , after all , 'identify' thought with language in the Tractatus. This identification was also to become a means for materialists to deny the mental by speaking of thought processes in terms of overt linguistic behavior and dispositions to such behavior. Other phi losophers simply took such perspicuous language schemata to afford unproblematic ways of putting the traditional issues and proposed solutions to them. Still others spoke in terms of an articulation of our conceptual scheme or of an elaboration of a conceptual framework that would be adequate for the inclusion of science, mathematics , and so forth. It is not surprising that with such varied offshoots of the earlier concern with the meaning of metaphysical pronouncements we have not been supplied with a clear and unproblematic articulation of the role of such ideal schemata in the philosophical enterprise. Nor is it surprising that it sometimes appears as if all a philosopher means by a perspicuous or clarified language is one which enables us to produce formalized transcriptions of statements of a natural language , excluding or including metaphysical assertions depending upon one's philosophical inclination ,

1 57

while banning those involving the familiar paradoxes. Supposedly, expressing matters in the symbolism of the predicate calculus, or set theory, or of a modal calculus is synonymous with clarification and · analysis for some philosophers. In this paper I propose to offer a partial remedy by specifying a presupposition for the cogent employment of formal schemata in the clarification and solution of the classical metaphysical or ontological problems and to consider the proper role of the background natural language (involving a conceptual scheme or background theory , as some would put it) . The presupposition involved will be discussed by way of considering a misguided use of clarified schemata that is both common and crucial to some leading contemporary versions of nominalism, a recent and popular conception of ontological reduction, and a current and fashionable view about 'theories' of truth. I will suggest that the views considered in this paper amount to a denial of some of the very problems they propose to resolve . While they may appear to use clarified schemata to consider and answer traditional philosophical problems, the philosophers we shall consider actually reject the issues they propose to resolve. They are thus enabled to get apparent solutions with far too few acknowledged entities. What I shall argue is that they can do so only because they run counter to the presupposition in question. However, I shall not offer any arguments for the principle at issue. It is too basic and too simple for one to defend. What I shall attempt to do is show how several philosophers, by denying it, illustrate the pointlessness of using clarified schemata for some purported philosophical purposes. I. Nominalism and Reference All philosophers recognize the difference between the quite ordinary claim that one who asserts that there is a white piece of paper on a desk is mistaken since what he took to be paper is plastic or because what appeared to be white is actual ly pink and the quite extraordinary claim that such a person is mistaken because there are neither physical objects , nor colors, nor relations or because physical objects are not really colored. Recognizing this, we recognize also that we may un­ problematically say, in an appropriate context and sense , that there are objects , properties of objects, and relations in which objects stand. In the same way we clearly recognize that some claims about objects having certain properties or standing in certain relations are true and others are false. When a philosopher raises questions about whether there really are common properties , in addition to objects, and about what corresponds to a true sentence (or makes a sentence true), so that it is true if there is

158

such a correspondent and false if not, he is raising a question that is not to be answered in the way in which one answers questions as to whether an obj ect is paper or plastic, white or pink . It should also be clear that what we take to be an appropriate response will depen d on how we construe the question. ' That there is some need to construe or ' reconstruct' the question can be taken to be the point of the positivistic revolution, at least for those who believe that while there is merit in the positivistically inspi red concern with the meaning of metaphysical assertions there is a trace of madness in the wholesale rej ection of such assertions based on that concern. Construing or reconstructing an issue obligates one who attempts to do so to restate the question without using the term or phrase that appears to be problematic or, at least, to provide a context of explication to remove the problem about its meaning or use. In so doing, he must not transform the question into one that is irrelevant to the historical issue. Thus , one who transforms the problem of universals into a question as to whether or not all or most natural languages have so­ called "general' terms obvi ously loses the thread of the historical problem. A more promising beginning is to suggest that the philosophical problem about properties and universals can be gotten at by raising questi ons about the reference and ascription of some predicates . Do some predi cates refer as proper names or so-called 'singular terms' do and, if so, to what do they refer? If they do not refer as proper names do, do they refer at all, and, if so, what is the difference in the mode of reference? Some may profess to find such questions as puzzling and inarticulate as the original problem. Whether to respon d to such critics or si mply to achieve further clarification, we can suggest getting at these questi ons, and related ones , by considering a simple model situation. We will talk about three obj ects and only note their colors, shapes, and some spatial relations . Con sider the domai n of the three obj ects pictured below. (I)

D

•

D

We have, then, two white squares and a black circle in our miniature 'universe' or domain. We may also consider a series of statements that are unproblematically true in the sense that it was unproblematic, in my earlier example, that a pink piece of plastic, rather than a white sheet of paper, was on my desk . (II) The square to the left of the circle is white. The circle is black. The circle is between the squares . etc. 1 59

It is both clear and unproblematic that we can speak of the word 'black' standing for one color or color property and the word 'white' standing for another. We may also say, unproblematically, that the sentences of (II) may be used to ascribe properties to the objects of the domain (I). At this point a philosopher seeking to establish the existence of universal properties may argue that the use of a predicate to ascribe one and the same property to two of the objects presupposes that we have recognized universals. But this is not an argument ; it is merely a prologue to one. What such a philosopher wants to hold is that the only account that would fit the unproblematic facts expressed by the list of (II) is one which would acknowledge universal properties. But what is it to give an 'account' in such circumstances and what is it to acknowledge universal properties? One may attempt to deal with such questions by introducing a schemat ic language. Let us now imagine ourselves to construct a sign system using a set of signs - 'a' , 'b' , 'c', - to refer to the objects of (I) ; signs that we later intend to take on the role of monadic predicates- 'W1 ' , 'S1 ' , 'B 1 ' , 'C 1 ' , and signs that will become relational predicates - 'L2 ' , "R/ , 'B3 ' . We can also specify formation rules in the standard way so that patterns like ' W 1 (a) ', 'L i (a, b) ' , 'Bia, b, c)', etc. become sentential patterns. We may now consider ourselves to have, in an extended sense, three 'domains': the miniature world of objects (I), the set of true English sentences (II) , and the artificial linguistic schema (III), which is as yet uninterpreted . It is clear that we can coordinate the signs 'a' , 'b' , and "c' to the left hand square, the circle, and the right hand square, respectively. In a way, I just did so. In so connecting the signs of (III) to the objects of ( I) I made use of written sentences of English or, as some might say, I made use of a 'framework'. Of course I need not have made use of written sentences of any natural language. If I was not concerned with communication, I might have said things to myself or employed gestures or perhaps, if one is not committed to the thesis that thought is linguistic, had appropriate thoughts that did not involve the use of language. But let us forget such possibilities and the issues they involve and consider the coordination to have been made using that same language to which the sentences of (1 1 ) belong. The problems I am concerned with have t o d o with the use of such a background language or " framework' and its proper role in the explication and resolution of ontological issues. I am not concerned with whether such statements of the background language or framework establish or merely record the coordination in question. A coordination has been made and it is revealed by the statements of the background language or framework . We can consider such a background language or framework to be a further domain (IV), which includes the sentences of (II) and further sentences to the effect that the sentences of (II) are true. Having coordinated the signs "a' , "b' , 'c' to their referents in the domain

1 60

(I) we may say that we have acknowledged such objects by the coordination. We may reflect this by a series of statements in our background language , (IV), making use of the standard quoting convention for referring to signs and sign patterns. (1 ) ·a· refers to the left hand square. , (2) "b refers to the circle. (3) ·c· refers to the right hand square. The recognition of the obj ects a, b, and c, reflected in (1 ) , (2) , and (3) , may be unproblematically called ontological. I say ·unproblematically' and "ontological' for two reasons. First, we must recognize such obj ects to exist , as we recognized a pink piece of plastic on the desk, to have a coordination and , second , the recognition of such obj ects is an obvious condition for our use of the sentence patterns of (III) to make statements about such obj ects. A n ontological committment with respect to a domain like (III) and a coordination is clearly reflected by the pairing of something of (III) with something of (I). This ought not to be controversial since it does not enable us to prej udge the issues surrounding the controversy between nominalists and those who rej ect nominali sm. One may hold that common properties or universals need not be recognized in an ontology since we need not map predicates like ·w 1 ' , " L2 ' , etc. onto colors, spatial relations, etc. as we mapped the signs •a' , "b', an d 'c' onto obj ects in order to have sentence patterns like 'W 1 (a)' reflect sentences of the background language such as "The square to the , left of the circle is white . Nominalism may take several forms. In one form, which I associate with Nelson Goodman , 2 the predicates are coordinated to the particulars a, b, and c in the same sense that the signs 'a' , 'b' , and 'c' are coordinated to them , except that a predicate may be coordinated to more than one particular - j ust as in natural languages proper names are often given to more than one individual. Thus, in addition to ( 1 ) , (2) , and (3), reflecting the coordination of names to obj ects, we would have (4) (5) (6) (7)

'W 1 ' refers to a an d 'W/ refers to c. ' Si ' refers to a and 'S i ' refers to c. 'B1 ' refers to b. 'C 1 ' refers to b.

In all of ( 1 ) through (7) the expression ' refers to' is used uni vocally. An anti nominalist or 'platonist' might advocate the use of (4 ) ' W / refers to the color white. (5: ) 'S 1 ' refers to the shape square. 1 61

(6 ) 'B 1 ' refers to the color black. (7 : ) 'C1 ' refers to the shape circle. Such a philosopher could then hold that the objects of (I) exemplify properties such as white, square, etc. , which are further constituents of the domain (I). The exemplification of a property by an object constitutes a fact which is then the ground of truth for a sentence like 'W 1 (a) ' . Such a platonist also uses 'refers to' univocally in (1) and, say, (4P ). Goodman, not recognizing properties of the objects in (I), has none for the objects of (I) to exemplify. Thus, sometimes he expresses his view by stating that objects exemplify predicates - terms or linguistic items. That is , the objects of (I) exemplify items of the domain (II I) and not further items of (I). Though Goodman uses 'exemplification' in a more complicated manner than the above remarks suggest, they are in keeping with his views as confined to the context of our problem. He is concerned with the problem of universals only in passing, as his interest is focused on the complexities of symbolism in attempting to explicate concepts like expression, representation, and metaphor and to develop an aesthetic theory. He speaks of exemplification in cases where predicates denote objects and, in turn, are taken to be denoted by the objects . For example, just as we let 'W i ' denote a, we might take a to be a sample of the predicate and, hence, to denote it. In this vein he says things like: Is exemplification , then , more intri nsic , less arbitrary , than denotation? The difference amounts to this : for a word , say , to denote red things requires nothing more than letting it refer to them ; but for my green sweater to exemplify a predicate , letting the sweater refer to that predicate is not enough . The sweater must be denoted by that predicate : that is , I must also let the predicate refer to the sweater.

and . . . but a singular label may equally well be exemplified by what it denotes . . . . The 'difference in domain' discussed earlier thus reduces to this : while anything may be denoted , only labels may be exemplified . 3

The import for our problem is clear enough. If one spoke of 'W 1 • being possessed by a, or being true of a, or being exemplified by a, these locutions are to be understood in terms of 'W 1 ' referring to or denoting the object a and not in terms of the object possessing, or having, or exemplifying a property referred to by the predicate. That is , such locutions are to be explicated in terms of statements like (4) and do not involve anything like ( 4 ) . Complications involving our "letting' or 'taking' the converse to h�ld , to get at certain of Goodman's concerns with the notion of exemplification , are not germane for our present 1 62

concerns . Given the connection of possessing ( or having , or exemplifying) with denoting , one could also then say that 'a' is true of a or that a exemplifie � •a' as well as ·w 1 ' . Moreover , since ·w 1 ' and ·s 1 ' have the same denotation one can hold that 'W 1 = S 1 ' is true , if one allows for identity statements of that kind . One thus incorporates a so-called extensionalist view with respect to predicates . In effect predicates are class terms given in extension, and , consequently , 'W 1 (x) ' , or "x£W 1 ' , is elliptical for ·x = a or x = c' . There are a number of traditional problems associated with such a view. A typical one is that an ordinary statement of ( II) , such as This is white' , used when one ascribes a color to a, correlates with ""'W i ' refers to a" or with 'a = a or a = c' . But, whereas the natural l anguage statement is neither one whose truth follows from statements reflecting coordinations of signs with things (semantical rules , if you will) as set forth in ( 1 )-(7) nor a statement which is of the form of a tautological identity , it is coordinated , on the view in question , with such a rule or tautology. Such a standard dispute I do not wish to pursue here . Rather , I am concerned with an attempt to avoid the obj ections and the problem by holding that 'W 1 ' is true of or applies to a since a is white , and not in virtue of a semantical rule or tautological identity. Such a move is involved in Goodman's concern with 'possession' , 'sample ' , and 'exemplification' . An obj ect is gray , or is an instance of or possesses grayness , if and only if 'gray' applies to the obj ect . Thus , while a picture denotes what it describes , what properties the picture or the predicate possess depends rather upon what predicates denote i t . I can l e t anything denote red things , b u t I cannot l e t anything that i s not red be a sample of redness . Let us, then , take exemplification of predicates and other labels as elementary . In so speaking , say of a chip as a sample of 'red' rather than of redness , we must remember that what exemplifies here is something denoted by, rather than an inscription of, the predicate . What a symbol exemplifies must apply to it . A man , but not an inscription , may exemplify ( every inscription of) 'man' . . . 4

Goodman , it is clear , wants to treat predicates like 'W 1 ' as denoting obj ects like a, as the name 'a' denotes a, but he also seeks to avoid the obvious obj ections based on the difference between assigning 'a' to a, as we did e arlier , and ascri bing 'W/ to a, as we may correctly do if 'W i (a)' is to express what the English 'This is white ' or , if one insists, 'a is white ' expresses . In short , what is the condition for veridically applying 'W 1 ' to a, a condition over and above the assignment of a sign to an obj ect , but one which avoids acknowledging properties? The answer is supplied by the obvious truth that 'gray' applies to an obj ect if and only if the obj ect is gray . The condition , in our case , is supplied by the English statement that a is white . [At times it appears as if I have summarily introduced the sign 'a' into ( IV) . It is worth reflecting on the difference between my 1 63

unproblematically doing that and the problem that a similar introduction of 'W 1 ' into (IV) would raise, in the context of the issues at stake in this paper. ] Such a defense of the nominalist view involves the two-fold claim that (4) holds because both a and c are white and that (4) , nevertheless, makes use of the same denoting relation as (1). The nominalist justifies his claims by appealing to the biconditional: (a) 'W1 ' is true of a = a is white. He further holds that that is all one need do to eliminate the problem of universals. The predicate 'W 1 ' holds of an object not because it refers to a property which the object has or exemplifies, but because the object is white. This claim focuses our attention on the right hand side of the biconditional (a) and the use of the term 'white' in that sentence . The term 'white' is a term of the background or natural language , and that is the crucial point in considering the cogency of the nominalist's response. Suppose that instead of (a) one invoked (a') 'W i ' is true of a = W i (a). An appeal to (a') would immediately lead to a question about the use of 'W/ on the right hand side. Since (a ') is merely a restatement of (a") 'W i ' refers to a = W 1 (a) it is obviously problematic to hold that (4) reflects or establishes a correspondence between a sign and a thing as (1) does. We do not simply coordinate a sign from the domain (III) to something from the domain (I) ; we coordinate a sign, 'W/, to an object , a, on the basis of the object's having a property or being of a kind. But , it is precisely such a condition that gives rise to the philosophical puzzle . To appeal to the condition that a is W 1 obviously amounts to a circular avoidance of the issue, not to a solution of the puzzle. It would obviously be problematic to use (1 ') 'a' refers to a as an expression of the coordination rule for 'a' , without a prior coordination of 'a' to some object. The nominalist of Goodman's stripe seeks to avoid the similar problem about 'W 1 ' in (a') and (a") by employing (a) and the use of 'white' as a term of the natural or background language. While it is obviously circular and inadequate to explain or coordinate 'W 1 ' by using that sign, as in (a') and (a"), the 1 64

nominalist apparently thinks it is cogent to employ the background term 'white' as in (a). Such a use of the background term becomes the key to the nominalist' s gambit. Before considering such a problematic use of a background language, we may note the similar use of formalized or perspicuous schemata by two other prominent nominalists. Willard Van Orman Quine 's version of nominalism sometimes depends on holding that there are two denoting relations so that instead of (4) - (7) we would have (4' ) 'W 1 • is true of a and ·w 1 • is true of c. etc. where "is true of' is not elliptical for ·refers to' but expresses another basic relation between words and objects. 5 This would be like taking "refers' in (4) in the sense of ·applies to' rather than in the sense of "is assigned to' . This leads to obvious questions as to the basis for statements like (4'). Sometimes there is no answer from Quine, but at places it appears as if the response would be

·w 1 ' is true of a = "W 1 (a)' is true = a is white. Thus , as in Goodman's case , he reverts to a term of the natural language and the unproblematic assertion that the object in question 'is white' . That is, in the one case "is true of' is elliptical for 'refers to' or 'denotes' ; in the other case it functions more like 'is satisfied by' or 'applies to·. But the difference comes to naught for both uses lead to the same appeal to a's being white. Some things Wilfrid Sellars has written would lead, and in fact have led, his readers to believe that the nominalistic position can be established by eliminating predicates and relation terms from a schema like (111).6 Thus, instead of having predicates like 'W 1 ' , etc. in (III), we would adopt the convention that the signs for the objects a, b, and c be considered to encompass different type fonts. Instead of having a formation rule putting a predicate next to a subject term, as in 'W 1 (a)', to form a sentence we would have , for example, the following signs a, A , a , A

to express, respectively, 'W. (a) ' , 'S/a) ' , 'C 1 (a)', and 'B 1 (a)' . In a like vein we could replace ' R2 (a, b )' by, say, ' ab' , and 'Lia, b )' by something like T . 7 In this manner we do not make use of either monadic or relational predicates and, hence, we perspicuously show that we do not take predicates to denote entities but only take'singular terms' to do so. Were 165

the possibility of so using type fonts and spatial relations among sign tokens taken as an argument for Sellars' nominalism it would be a poor argument indeed . Esoteric penmanship would simply replace philosophical argument . Obviously what would now stand for a property would be a letter' s being in a type font of a certain kind, rather than a predicate letter, and , hence, something is now relevant for the perspicuous schematic language to contain correlates of natural language sentences that previously was not relevant , that is, the type font of the singular term or proper name. However, I think that what Sellars intends is to merely show, in a perspicuous manner, the rejection of properties as entities, while assuming that there are other grounds on which to base the rejection. These grounds are to be found in Sellars' arguments that the proper formulation of coordinating statements like (4) through (7) for a schema that included predicates, and hence one which would not be as perspicuous a schema as one employing only names in a variety of type fonts, would be in terms of something like (4) (5) (6) (7s )

'W1 ' is a · white· 'Si ' is a · square· 'C 1 ' is a · circle· 'B i ' is a · black·

where the dots around the English predicates indicate that the predicates 'W 1 ' , etc. play the role in (III) that their English counterparts play in (II) and (IV) . What this amounts to is a complicated way of stating that ' W1 ' correlates with or is interpreted by or, even, means the same as ' white' does in English. 8 By contrast , in (1) through (3) we retain the referential connection between the terms 'a' , 'b' , and 'c' and the objects they denote. Sellars, like Good man and Quine, thus also makes use of the English term 'white' , but in a more complicated way. 9 The co mplication supposedly provides the defense of the view. There is a feature of this difference that some will take to be impo rtant , but which I take to be irrelevant. Alfred Tarski's convention T, whi ch we shall discuss in section II, makes much of the absence of a semantical device, like the quoting operator, on the right hand side of the biconditional: 'a is white' is true = a is white. Sellars retains a semantical device by the use of (4 ) to avoid properties like white, whereas one may put the Goodman-Qui�e gambit in terms of: 'W 1 ' is true of or denotes a = a is white,

1 66

and , hence, in view of the right hand side of the biconditional , claim that one has stated the condition for the application of ·w 1 ' without use of a semantical device. This, as we shall see , is the way in which Donald Davidson will try to talk about truth without bothering with facts and properties. The difference makes no difference for purposes of my argument, since all the gambits make crucial use of the English statement or predicate to block the acknowledgment of properties. What difference there is that is reflected by Sellars' device of the dot quotes reflects his recognition that the simple appeal to the English statement is inadequate. But the recognition goes for naught. For , in spite of the complexity introduced by his gambit and the systematic use he puts it to in dealing with problems in the philosophy of the mind , Sellars ultimately does what the others do, only he does it in a more direct , yet less obvious manner. He connects a predicate of (III) with a predicate of a natural language as in (4J Goodman , Quine , and Davidson (as we shall see in the latter's case) get the same result using the natural language predicate in a sentence that states the condition for the predicate of (III) applying to , or being true of, or being satisfied by (in the form of a sentence or an open sentence) , or denoting the object a . Goodman, Quine , and Sellars make a common move when they use the terms of the natural or background language to avoid a coordination of the predicates of (111) to properties of the objects of (I). Contrast what they do with the philosopher who makes use of (4 ) through (7 P) . According to the latter , we are as much commit�ed, by such a coordination , to the properties as we are to the objects a, b, c. He then goes on, having recognized properties , to consider whether one will be taking properties to be universals , or classes , or quality instances, etc. In so doing , he explains the meaning of the term ·universal' as it occurs in philosophical discourse , in a philosopher's talk about the domains (I) , (II) , and (I II) and about the coordinations among them. Such discourse takes place in (IV). A philosopher who maintains that our natural language reference to properties is properly analyzed in terms of particularized quality instances , rather than in terms of universals, will hold that while 'W/ may be coordinated to the color of a , we must introduce a further term, say 'W2 ' , to refer to the color of c. He will probably also hold that we must recognize a rather special relation that will hold between the quality instances W 1 and W2 , and , hence, introduce a further term into (III) to refer to such a relation. A philosopher advocating the recognition of universal characteristics will hold , among other things, that 'W 1 ' stands for one and the same entity with respect to both a and c. Such arguments as may then ensue between proponents of these and other alternatives are not my concern here. We may note , however, that in spelling out the various alternatives one recaptures,

167

rather than ignores, much of the traditional dialectic, and, in so doing, one expounds and explicates more of the contextual meaning of philosophical usage of terms like 'universal'. One uses the model domain (I) and the schematic language (III) to elucidate the traditional vocabulary. We also make use of the list (II) and the background language (IV) . The latter is crucial in our coordinating the philosophical term 'universal' , a problematic term of the background language, to the unproblematic background term 'property', in order to propound and explicate the alternative metaphysical views. Such coordinations of metaphysical terms to ordinary ones is quite different from the coordination of the terms of (III) to the entities of (I). But such comments invite an objection. Does the platonist not make use of the background term 'white' in (4 ) and did I not express myself by saying we could take 'W 1 ' to stand for the property indicated by 'white' in English? Is this not to do exactly what Goodman and Sellars do, except for my speaking of 'standing for a property' where Sellars might use 'plays the same role as' ? Or, to put it another way, is the only difference that the platonist uses the term 'white' without quotation marks in ( 4 ) , whereas the term occurs within the special dot quotes in (4) ? Is n6t Sellars' line of argument cogent and my way of speaking verbose and redundant? The appearance is deceptive, for I have not made use of the background natural language in the way in which the trio of nominalists do. Their line of argument negates the point of employing schematic languages to resolve the issues at hand. With respect to the problem of universals, the question is whether to take colors, shapes, etc. as being of that ontological kind. We are asking whether the correlate of a term of the background language , say 'white' as it occurs in sentences of (II) , can be used in sentences of (I II) so that the latter can be taken to express what the former does without the term 'W 1 ' being correlated to a common property. (In my previous sentence the term 'property' is used in an unproblematic sense.) One thing our three nominalists do is refuse to ask the question we raise , since they get the sentence of (I II), 'W 1 (a) ', to express what a sentence of (II) does by ruling that 'W 1 ' correlates to the term "white'. That they do this by talk of 'true of' or the use of the dot quotes, rather than by talk of meaning or correlating, is beside the point . Another thing they do , knowingly or not , is refuse to acknowledge that terms like •fact' and "property' are used in an unproblematic sense on some occasions when we speak of the property white and the fact that an object is white. Once one recognizes that there is an unproblematic sense in which we may speak of properties and facts it becomes difficult to avoid the issue in the way in which the trio of nominalists do . Recognizing that there are properties in an unproblematic sense, we must take the relevant ontological question to involve a request about the nature or status of properties. By doing what

168

they d o, they bypass the problem of universals by a sort of semantical detour. Once such a problem is raised, rather than avoided , we can obviously not be content with the claim that "W . (a)' holds because a is white or with the assertion that ·w 1 ' corresponds to or plays the same role as "white' as solutions to the problem of universals. In short, using a schematic formal language to explicate and resolve ontological issues involves our coordinating the primitive terms of the schema to 'things' of the domain (I). To express or state what is coordinated to what by use of a background schema or language or framework is one thing ; to coordinate a term of the schema to a term of the background language is quite another thing. One simple aspect of the situation may blind the nominalist to this obvious difference. Suppose the nominalist is asked why he does not hold that we need not recognize the object a since we could coordinate the sign •a' of the schema to the background language phrase ' the left hand square' or "this' , in an appropriate context, rather than to the relevant object of (I). Thus, we could avoid recognizing a particular object like a as an entity. This, of course, suggests a reductio ad absurdum of the nominalist's position. He could reply that while the sign ·a' of the schema (III) refers to the same thing as an appropriate phrase of the background language, it does not mean the same thing as such a phrase. In the case of singular terms it is inappropriate to hold that different terms or phrases mean the same thing when they refer to the same thing . In the case of predicates or so-called 'general' terms, however , the notions of meaning and reference are readily run together , as one speaks , of ·w 1 ' meaning the same thing as 'white or referring to the same color as the English term. Since the natural language expressions 'means the same' and 'refers to the same' are somewhat interchangeable for predicates in such contexts, one can easily be led to think that the question of what a term like 'W 1 ' is coordinated with can be replaced by a question as to what natural language term it represents. Once the replacement is made the ontological question is lost. Moreover, replacement on such grounds reveals that, on the one hand , the nominalist depends too much on an apparent difference in natural languages between 'singular' and 'general' terms and , on the other hand, he begs the question by emphasizing one of the apparently interchangeable natural language expressions at the expense of the other. One may object that I am merely stipulating a problem and the condition for its solution by fiat. Hence, the so-called ontological issue is an artificial one. There comes a point in all such disputes where argument must cease. As I see it, what is offered by the approach taken here is a way of construing the classical ontological issues (or , at least, some of them) so that they are not buried by esoteric linguistic devices. When it is held

169

that the problem is to be formulated in terms of acknowledging the coordination of the primitive signs of (III) to things, not words, I am clearly making a more restricted use of the background natural language than the nominalists we have considered. That one must make such a restricted use of the background language may be taken as a principle guiding one's conception of the philosophical enterprise. The acceptance or rejection of such a principle leads to two sets of rules for explicating and resolving metaphysical questions. In effect, one who deals with the ontological questions according to such a principle accepts more stringent conditions for their solution; thus, it is no surprise that he ends up with more entities. We may also note that a historical question arises. Which use of perspicuous language schemata fits best with the elucidation of the historical tradition and provides us with insights about it? What I have said is more in the nature of specifying what a particular philosophical issue is, rather than offering an argument purporting to establish how we must construe such issues. But, as in all such cases, we can judge the respective viewpoints by their fruits. A paradigm case is provided by Quine's use of a background schema or language. Quine's approach to four further philosophical issues involves a remarkably similar use of a background language to that employed by all three nominalists to avoid the acknowledgment of universal properties. Quine has used the approach to eliminate names, talk of truth values, the appeal to classes in validity theory for lower functional logic, and reference to mental entities. The approach is so similar in all four cases that one is tempted to speak of'Quine's way out' as a general rule for disposing of things and issues. In the case of names, Quine 's move is astoundingly simple. Suppose that the linguistic apparatus of (III) has been enlarged to include a system of quantificational logic and identity along with a standard approach to definite descriptions along Russellian lines. Assume we also have signs in the same syntactical category as 'a' , 'b' and 'c' which are not coordinated with objects, that is names that do not name. Let'Pegasus' be such a sign. According to Quine we can define a predicate 'pegasizes' so that (D 1 ) pegasizes (x) = df. x = Pegasus. Given (D 1 ) , we can now eliminate 'Pegasus' from (III) and use, instead, 'the x such that x pegasizes' or, in Russell's symbolism, '('x) pegasizes (x)'. The procedure is taken to be legitimate since (D 1 ) really belongs to the background schema which is used in introducing terms into (III). This supposedly answers the obvious question about the cogency of eliminating a term used to define another term by replacing the former with the latter. 10

1 70

The same technique is used in the case of propositional logic where we have a so-called 'internal' criterion for a tautology , being a theorem in a system , and an 'external' criterion , having the value true in all lines of a truth table . The external criterion, as Quine views it, commits us to truth values as entities. To avoid such a commitment we can eliminate the reference to truth values by , first, noting that the external and the internal criteria yield the same results and , second, dispensing with the external criterion by specifying that a tautology is to be characterized only in terms of the internal criterion. One is tempted to make the obvious protest that the internal criterion is what it is , namely a criterion for a form being tautologous, only because it correlates with the external criterion. Once again the background language is thought to provide a reply. Since the crucial statements of the coordination of the two criteria are made in a background language, one does not need to refer to truth values in a schema employing the system of propositional logic and specifying a proposition to be a tautology in terms of that system. The use of such a schema, not employing terms standing for truth values, does not involve the user in ontological commitments associated with such terms. Essentially the same move is made to eliminate reference to classes and hence the recognition of classes as entities in the case of the elementary predicate calculus. We have an external criterion in terms of standard validity theory making reference to classes and an internal criterion specifying a propositional form to be a logical truth in terms of being a theorem of a certain system. We eliminate reference to what is talked about in the terms of the external criterion by employing only the internal criterion. Quine 's most dramatic and philosophically interesting use of the technique in question occurs in his offhand dismissal of mental entities. Consider a language system that employs mentalistic terms and another that employs only physicalistic terminology. Many have felt that they can eliminate a commitment to mental entities by avoiding the use of mentalistic terminology. One way of avoiding such use, according to some philosophers, is by means of discovered correlations between mental states and physical ones. Quine gives the attempted reduction a new twist. Suppose we have not discovered such correlations as yet for some mental states. Let M 1 be such a kind of mental state. All we need do, according to Quine, is introduce the term 'P 1 -state' so that we stipulate that a P 1 -state is a physical state that holds of a subject if and only if there is a corresponding M 1 -state holding. Having done this, we now eliminate the original expression for the mental state, M l from the language and ' use the expression 'P 1 -state' to characterize subjects in that state. By the use of such a linguistic schema we thus eliminate an ontological commitment to the mental states, since we have eliminated the 171

mentalistic terms from the vocabulary of the schema. 1 1 Again, the coordination is stated in a background schema containing reference to both M and the 'corresponding' physical state. But here, as in all the above c�ses, since we employ the schema with the terms for the 'entities' we seek to eliminate, how have we succeeded in removing them from our ontology? One response, which I do not take to be explicitly Quine's, is to point out that the ontological questions are resolved in terms of the commitments of a formal schema, such as (III), and not in terms of our background natural language. Such a response would presuppose the use of the background language I have been criticizing and reveal its triviality . That was one reason for bringing Quine's four 'reductions' into the discussion. Quine, himself, has an explicit response that is just as inadequate. Our dependence upon a background theory becomes especially evident when we reduce our universe U to another V by appeal to a proxy function . For it is only in a theory with an inclusive universe , embracing U and V, that we can make sense of the proxy function . The function maps U into V and hence needs all the old obj ects of U as well as their new proxies in V . . . . If the new obj ects h appen to be among the old , so that V is a subclass of U , then the old theory with universe U can itself sometimes qualify as the background theory in which to describe its own ontological reduction . But we cannot do better than that ; we cannot declare our new ontological economies without having recourse to the uneconomical old ontology . This sounds , perhaps , l ike a predicament : as if no ontological economy is j ustifiable unless it is a false economy and the repudiated obj ects really exist after all . But actually this is wrong ; there is no more cause to worry here than there is in reduction ad absurdum, where we assume a falsehood that we are out to disprove . If what we want to show is that the unive rse U is excessive and that only a part exists, or need exist , then we are quite within our rights to assume all of U for the space of the argument . We show thereby that if all of U were needed then not all of U would be needed ; and so our ontological reduction is sealed by reductio ad absurdum. 1 2

Thus, if we take 'square of' to be a proxy function (or, perhaps more accurately, the mapping whereby each statement of the one 'theory' about the elements of U is coordinated to a statement in the theory about the elements of V) and consider a universe U of the natural numbers and a subclass of U consisting of the squares of the members of U, we can consider U to be 'reduced' to this subclass, V. In short, 2 and 3 will have been reduced to 4 and 9 , respectively, and, hence, we can do without 2 and 3, as they are not members of V. 9, as a member of U , will have been reduced to 8 1, as a member of V, but will nevertheless persist as a member of V, since it is the square of 3, but no longer, so to speak, will it be 9. Of course , it is the role 9 pl ays as the ninth member of one progression and as

172

the third of another that is crucial . So, of course, there is no puzzle in one sense . What is puzzling is talk of 'reduction' in such cases. Such talk is based on the essential presupposition that a coordination or mapping by means of a proxy function constitutes a reduction of some things to other things. For such a presupposition there neither is nor can be an argument. One merely stipulates a use of the term "reduction'. Be that as it may , such a stipulation about reduction does not help in the cases of the elimination of the purported name 'Pegasus' and the mental state M 1 . In those cases we are concerned with the circularity of introducing or specifying the meaning of an expression, ·pegasizes' in the one case and 'P 1 -state' in the other, in terms of another expression and then eliminating or reducing the latter. Quine may overlook the difference between the two types of cases since, in some vague sense, a background language operates in both. That is, when we reduce 9 to 8 1 the statements of the coordination , on which the reduction is based, must be made in a background language employing the terms "8 1 ' and "9'. This, as Quine views matters , does not interfere with the reduction of the one to the other. Therefore , he might think that when we eliminate 'Pegasus' in favor of 'pegasizes' , the statement (D 1 ) occurs in the background language and hence does not preclude our eliminating one term in favor of the other , even though we use both in the background language. But this makes no literal sense. It is one thing , specious as it is, to reduce 9 to 8 1 , where the expressions '9' and " 8 1' are taken as terms of arithmetical systems in some ordinary mathematical context . It is quite another thing to speak of such a reduction where the expression '81' is defined , for example , as 'the square of 9'. Quine apparently sees no difference or takes it to be insignificant. Quine's approach to ontological reduction illuminates the attempt by Goodman and Sellars to avoid a commitment to universal properties by a specious use of the background natural language. There is, as I see it , no difference between what they do with ·w 1 ' and ·white' and what Quine does with 'pegasizes' and 'Pegasus'. This suggests that the reductio that is involved concerns the nominalist gambit and Quine's views on reduction, and not the denial of ontological reduction on the basis of a mapping. To be succinct , connecting a line of philosophical argument with Quine's '"way out" constitutes , if not a reductio ad absurdum of such a line of thought, at least guilt by association.

II. Truth and Triviality Davidson has more recently made use of the same approach as Sellars, Goodman , and Quine in his consideration of the concept of 'truth'. He 1 73

has held that a clear and sufficient theory of truth is offered if we have, in a schema, a predicate'T' which satisfies Tarski's Convention T. Of the problem of truth and its solution, he writes: Tarski taught us to appreciate the problem , an d he gave an ingenious solution . The solution depends on fi rst characterizing a relation called satisfaction and then defining truth by means of it. The entities that are satisfied are sentences both open and closed ; the satisfiers are functions that map the variables of the obj ect language onto the entities over which they range - almost everything , if the language is English . . . The semantic concept of truth as developed by Tarski deserves to be called a correspondence theory because of the part played by the concept of satisfaction ; for clearly what h as been done is that the property of being true has been explained , and non trivially , in terms of a relation between language and something else . The relation , satisfaction is not , it must be allowed, exactly what intuition expected of correspondence ; and the functions or sequences that satisfy may not seem much like facts . . . If we thought of proper names i nstead , satisfiers could be more nearly the ordinary obj ects of our talk - namely , ordered n-tuples of such . Thus 'Dolores loves Dagmar' would be satisfied by Dolores and Dagmar (in that order) provided Dolores loved Dagmar. 1 3

In appealing to the expression 'provided Dolores loved Dagmar' Davidson avoids the question about facts, as grounds of truth, just as Sellars, Goodman, and Quine avoid the question about universals. Suppose we include a truth predicate in (III) and satisfy the Tarski-type condition so that, for example,

holds. What this means is that the sentence'W i (a)' is true if and only if a is white. But the question that philosophers have traditionally pondered involves the connection of sentences like 'W 1 (a)' with something of (I). What Davidson provides from the domain (I) is the object a, since a is white. That is, one gets the English rendition of 'W 1 (a)' and the object as the ground of truth, or explanation. Thus, we get an individual from (I) and a statement from (IV), only it is put in the roundabout way of saying that a is a satisfier, provided a is white. We might ask in what sense we have a theory of truth when we have a condition for a truth predicate in a schema? We are told that one question replaces another, and this is supposedly justified since the one question is clear and the other is not. One can almost hear the reply being made in a Viennese coffee house of the 1930's. The central merit of Convention T is that it substitutes for an important but murky problem a task whose aim is clear. After the substitution one appreciates better

174

what was wanted in the first place , and gains insight into the aetiology of confusion . The origi nal question is not confused , only vague . It is : what is it for a sentence ( or utterance or statement) to be true ? Confusion threatens when this question is reformulated as , what makes a sentence true? The real trouble comes when this in turn is taken to suggest that truth must be explained in terms of a relation between a sentence as a whole and some entity , perhaps a fact , or state of affairs . Convention T shows how to ask the original question without inviting these subsequent formulations . 14

Here we clearly have the replacement of a phi losophical question by one that is of i nterest to those i nterested in certain formal or logical properties of linguistic schemata. It is clear that Tarski was interested i n a purely technical question: Eve n a superficial analysis . . . shows that in general composite sente nces are in no way compounds of simple sentences. Sentential functions do in fact arise in this way from elementary functions . that is from inclusions ; sentences on the contrary are certain speci al cases of sentential function s . In view of this fact , no method can be given which would enable us to define the required concept directly by recursive means . The possibility suggests itself, however, of introducing a more ge neral concept which is applicable to any sentential function, can be recursively defined , and . when applied to sentences, leads us directly to the concept of truth . These requirements are met by the notion of the satisfaction of a given sentential function by given objects, and in the present case by given classes of individuals . 1 5

The incompleteness of Tarski's Convention T as a criterion for a theory of truth and hi s concern with a formal questi on is apparent from his statement about languages with a finite number of sentences. If the language investigated only contained a finite numbe r of sentences fixed from the beginning, and if we could enumerate all these sentences, then the problem of the construction of a correct definition of truth would present no difficulties. For this purpose it would suffice to complete the following scheme: x E Tr if and only if either x = x 1 and p 1 , or x = x 2 and p 2 , . . . or x = x 0 and p 0 , the symbols 'x 1 ' , 'x 2 ' , . . . , 'x 0 ' being replaced by structural descriptive names of all the sentences of the language investigated and 'p 1 ' , 'p 2 ' , • • • , 'p 0 ' by the corresponding translation of these sentences into the metalanguage . 1 6

In short, we give a 'correct definiti on of truth' or a ' theory of truth' by listing the true sentences, in such a finite case. The inadequacy of such a 'theory' for concerns philosophers have traditi onally had appears obvious. It i s interesting to recall that Sellars replaces a question about the coordination of predicates with the domain (I) by a question about the coordination of predicates of (IV) with predicates of (III) . Davidson duplicates Sellars' move in a two-fold way. He replaces an ontological or philosophical question with one that is not, and he must literally, but 1 75

covertly, make Sellars' move about predicates. 1 7 The latter feature comes out when we note that a satisfies 'Wi (a) ' because a is white. That ends the matter, just as a list of true sentences would end the matter for Tarski, in the case of a language with a finite number of sentences which we could enumerate, as in (III) . By making the response 'because a is white' , Davidson makes implicit use of the background schema or language to avoid a philosophical issue in the same way Sellars does. Properties and facts are avoided by the same kind of move. There is another interesting feature of Davidson's appeal to Convention T. In avoiding relational facts like the fact that Dolores loves Dagmar, Davidson speaks of ordered 'n-tuples' of ordinary objects such as the pair Dolores and Dagmar in that order. He thus appeals to a further kind of entity, an ordered pair. Some philosophers might question the ontological parsimony gained by speaking of an ordered pair which satisfies 'Dolores loved Dagmar' , while disdaining to talk of such metaphysical and vague entities as facts. But we can understand how one could come to make such a move. Logicians and mathematicians deal with ordered n-tuples in a systematic manner. In this sense, much is known about such 'things'. Thus, the philosophical problems about truth and facts disappear, in part, in favor of the textbook treatment of non­ philosophical aspects of logic and mathematics. It is worth noting how the relational case and the ordering involved suggests a plausibility that is totally absent from the monadic case of 'W1 ' . In the latter case one has to say, in the traditional terminology, that a is the ground of truth of " W 1 (a) ' . This is what directly leads to thejustification of such a claim in terms of a's being white, which returns us to the specious use of the background language to avoid the issue. A similar problem arises for those who seek to suggest that the fact that an object , say a, stand s in a relation to another object, b , is best represented by a juxtaposition of the signs for a and b, and not by a sentence employing a relational predicate. In the monadic case, they must make a different move; hence Sellars' use of type font . A nonlinguistic version of the problem arises when some speak of a fact such as that a stands in L2 to b as a configuration of objects, without a relation being involved. To accommodate facts like that a is white, they must construe the object a as some sort of esoteric compound to avoid introd ucing properties as constituents of facts. For Davidson, no such complications arise; he is content to note that a is white. The way in which the classical problems surrounding the notion of truth are avoided by the use of convention T as a sufficient criterion for a "theory' can most easily be seen when the question Davidson skirts as 'murky' is rephrased. Davidson finds it unclear what is asked when a philosopher raises the question 'What makes a sentence true'? In a similar way, one may purport to be puzzled by the antinominalist's query

176

as to what is the ground for the veridical ascription of one predicate to two obj ects . In effect , purported puzzlement about the terms ·makes' and ' ground' inj ect the approach , if not the style , of the later Wittgenstein into the discussion . But such puzzlement about usage can be dispelled . When one asks what makes a sentence like 'W 1 (a) ' true , he is asking what is it , in the domain (I) , that is such that its existence is a sufficient condition for the truth of the sentence . The existence of the obj ect a obviously does not constitute a sufficient condition , since its existence is compatible with the sentence being false . Of course one could develop a view whe reby the existence of the object would be a sufficient condition for the veridical ascription of the predicate 'W 1 ' . For example , one could hold that obj ects , like a, were to be construed as classes of properties like W 1 • One then analyzes the notion of an obj ect so that predicative ascriptions , if veridical , are necessarily so , in some sense of 'necessary' . But this would hardly be in the style of the gambits I am criticizing. The existence of the object a might be taken as a sufficient condition for the truth of ·a exists' or ' ( 3x) (x = a)'. It cannot be taken as such a condition for the truth of 'W 1 (a) ' , so long as we hold that the obj ect a need not have been white even though it is white . By contrast , the philosopher who introduces facts does furnish a sufficient condition for the truth of ·w 1 (a) ' . It is inconsistent , in the terms of the view acknowledging facts , to hold that the fact exists but that the sentence is not true . Given that the fact exists , it follows , in the context of such a view, that the sentence is true . 1 8 One may seek to furnish a sufficient condition via talk about satisfaction . Such a gambit would hold that we need not appeal to an entity like a fact since it is the relation of satisfaction between the obj ect a and the sentence 'W 1 (a) ' , or perhaps between the obj ect and the 'open' sentence 'W i (x) ' , if (III) were expanded to permit such expressions, that provides the ground or condition of truth . Such a response has a peculiar consequence . Let us forget for the moment that a satisfies 'W 1 (a) ' (or ' W 1 (x) ' ) - a is white

might provide a basis for saying that talk of satisfaction here is empty , since we are merely repeating that 'W 1 (a) ' is true on the ground that a is white and , once again , not answering the question . That is , let us concentrate on the left hand side of the above biconditional as a statement of the truth condition in the sense that it offers a sufficient condition for 'W 1 (a) ' being true . One appeals to the satisfaction relation as obtaining between the obj ect a and a linguistic expression as the basis or ground of truth , in lieu of appealing to an entity such as the fact that a is white . This means that the truth ground necessarily involves language . 177

The ground of truth, not the bearer of truth or what is true, is , therefore , i n part linguistic. Such a line not only j ars with a fundamental piece of common sense - the way things are is the basis for our statements about them being true - but reveals an implicit idealist theme in the talk of satisfaction and convention T. If sentences are the sort of things that are true or false , then we cannot have truth, in one obvious sense , without having language and , hence , the originators and users oflanguage . This is one thing. It is not to say that the conditions or the grounds for sentences being true involve language and its users . That there were mountains on earth before sentence tokens I take to be true . Taken literally , the gambit presently under discussion holds that a relation between the obj ects which were mountains and certain expressions (whether tokens or types matters not here) constitutes the sufficient condition we were concerned with . This means that the condition or ground of truth - that in virtue of which the claim is true - had to await the advent of language and its users . This is a way of saying that mountains are the sorts of things that they are in virtue of the use of language . If we now recall the attempts to explicate 'thinking' in terms of 'speaking' the link with idealism is even more obvious . The point I am concerned with can be put another way. There is an ambiguity in the phrase 'truth condition' that is exploited by the response I am rejecting as idealistic. Since the obj ect a satisfies a certain expression of (III) if and only if a is white , one can think of the truth condi tion alternatively as a's being white or as a's satisfying the expression . If one concentrates on the l atter as a response to my query about a sufficient condition , he is forced to the absurdity of the idealist's view . If he shifts back to repeating that a is white , he abandons the attempt to supply a sufficient condition and simply ignores the question . I suspect one reason a philosopher might appeal to the satisfaction rel ation in order to specify a sufficient condition is a failure to distinguish between the obvious truth that if sentences are the bearers of truth one must have language for ascriptions of truth and the far from obvious claim that it is a connection between language and obj ects that constitutes the ground or basis for sentences being true . There is another point that is skirted by those who disdain talk of facts or 'grounds' of truth by means of the appeal to convention T. If asked what it is that grounds the truth of any subj ect - predicate sente nce of (III) , or any other schema of the same kind , one who acknowledges properties and facts can reply : (T' ) A subj ect - predicate sentence is true if and only if the obj ect denoted by the name exemplifies the property denoted by the predicate . 1 78

He can thus furnish a general condition of truth for subject - predicate sentences (and correspondingly for sentences with relational terms). B y contrast , one who appeals only t o objects, sequences, and satisfaction can not provide a corresponding general criterion. He requires a specific sentence like 'W 1 (a)' so that he can hold that

·w i (a) ' is true if and only if Wi (a) (or a is white). He cannot employ variables so that something like A subject - predicate sentence 'F(x) ' is true if and only if F(x), becomes an alternative to (T' ). What he can do is give the T- sentence for each subject - predicate sentence of (III), or similar schemata, but this is not to do what t he general statement (T' ) does. What this brings out is that it is one thing to provide a condition for attaching a so- called truth predicate to a subject - predicate sentence ; it is another thing to state what it is, in general, for such a sentence to be true. 1 9 One could retort that (T") A subject - predicate sentence is true if and only if the object denoted by the name satisfies the sentence (or a corresponding open sentence containing the predicate) . does exactly what (T' ) does without talk of properties and exemplification. B ut , with (T'') we embrace the idealistic absurdity I discussed above. Moreover, (T") goes against the basic theme of the gambit employing satisfaction and does not accomplish what (T' ) does. For the idea of characterizing satisfaction in the case of a and 'W 1 (a) ' is that the object satisfies the sentence provided that a is white. This is precisely what cannot be characterized in the general case, for one cannot characterize the condition in general. The point is simply that given a specific sentence l ike 'W 1 (a)' one can state the condition for a satisfying it (or ' W i (x) ') , that is, the condition for the sentence being true: W i (a). Without the specific sentence being furnished, one must use the term 'satisfies' as in (T") rather then eliminate it as in the specific case of ·w i (a)'. 20 This is a reflection of the point that the list of true sentences constitutes a ' theory of truth' in the case of a schema like (III) . Where several distinguished philosophers employ a common pattern which one believes to be misguided , one feels compelled to seek a reason. No doubt the old positivistic disdain for the traditional metaphysical 179

issues and the wish to exorcise them is partly responsible, but other possibilities seem relevant. Some philosophers have held that the approach I have advocated mistakenly presupposes that we need not make use of a background language. Such an approach supposedly assumes that the coordination of (I) with (III) takes place in a linguistic vacuum. Properly conceived , such a coordination cannot be divorced from the context of our natural language. Recognizing this, we might just as well use the background schema as Sellars does, to avoid properties, facts, etc. Interestingly enough, the positi vistic rejection of metaphysical issues involved in the use of perspicuous language schemata along the lines that I have been rejecting joins with a traditional argument of pragmatic-idealistic-holistic flavor that has long been directed against the empiricist-atomist tradition. One can almost detect the old phrase 'vicious abstraction' echoing in the charge that the context supplied by the background language is overlooked in the approach taken here. It is no accident that the pattern I have been presenting stems from the views of Russell and the early Wittgenstein. Nor is it an accident that the holistic-pragmatic pattern influences the work of Quine, Sellars, and Goodman. B e the historical connections as they may, the point at issue is not that we need make use of a background framework or language or that we cannot abstract the philosophical questions from it . The issue is whether we can make distinctions with respect to the different ways of using such a background schema. We need not deny that our talk about (I) and (III) takes place in a natural language to deny that it is legitimate to use 'W 1 ' plays the same role as " white' plays in English and 'W 1 ' is true of a (or 'W 1 (a)' is satisfied by a because a is white) to avoid the recognition of universals and facts. Facts and properties pose a number of well known problems for Goodman, Quine, and Davidson: problems that center around so-called identity criteria and questions of extensionality. Such concerns could also lead to an attempt to avoid the 'problematic entities· at virtually any cost . This can help us understand why they do what they do. Sellars, I suspect, has a further motive. M ore sympathetic to the historical tradition, Sellars fears that wi th properties come facts, and with facts we face the issues surrounding the notions of negation and possibility. Thus, he seeks to bl ock the threat of an expanding ontology at the outset. Perhaps the same motive led Wittgenstein to abandon the themes of the Tractatus and seek to turn phil osophy away from the classical problems of ontology. 2 1

1 80

Whatever their motives, it is clear, I trust , that the philosophers we have considered avoid ontological issues and commitments by a broader use of the background language than I wish to allow. I am not , in this paper, denying t hat one can avoid these issues as they do. I am merely noting that they do so by playing the philosophical game by different rules. The most obvious point of dispute centers on the purpose of playing the philosophical game as this paper suggests it is to be played . I recall a philosopher once asking , �what good are properties' ? and , in a similar vein, another claiming that facts were not useful for what he was about . Properties, facts, or, for that matter, Aristotelian substances and natures, may not be good for much, but so long as they suffice as part of a framework that may cogently deal with an admittedly extraordinary set of puzzles, they are as good as they need to be. One issue is whether such puzzles can be handled without appealing to such things. I have argued t hat one cannot dispense with properties and facts as some philosophers have sought to do. Another question is whether one should neglect the puzzles that lead philosophers to talk of such things. Perhaps that is the crux of the matter. One thing I have been implicitly suggesting is that in so far as they make use of the patterns criticized in this paper, the philosophers we have considered accept the rejection of the classical metaphysical issues along lines associated with the later Wittgenstein. The use of the background language that I have obj ected to amounts to a Wittgensteinian dism issal of philosophical puzzles, only it is hidden by a more formalistic style of philosophizing. It is interesting to note some remarks of Tarski's in this connection: I have heard it remarked that the formal definition of truth has nothing to do with 'the philosophical problem of truth' . However, nobody has ever pointed out to me in an intelligible way j ust what this problem is . . . . In general, I do not believe that there is such a thing as 'the philosophical problem of truth ' . I do believe that there are various intelligible and interesting (but not necessarily philosophical) problems conce rning the notion of truth , but I also believe that they can be exactly formulated and possibly solved only on the basis of a precise conception of this notion . . . . In fact , the semantic defin ition of truth implies nothing regarding the conditions under which a sentence like ( 1 ) : ( 1 ) snow is white can be asserted . It implies only that , whenever we assert or rej ect this sentence , we must be ready to assert or rej ect the correlated sentence (2) : (2) the sentence 'snow is white ' is true. 22

This paper is not an attempt to meet Wittgenstein's attack, or, perhaps more accurately, the attack that is commonly associated with his name, by offering an argument for the legitimacy of ontological questions as herein construed. For my purposes in this paper it suffices if certain

181

differences have been highlighted and clarified, and if the suggestion of the connection between the philosophers discussed above and the later Wittgenstein is seen to be plausible . This requires a modification of the use of the term 'nominalism' throughout the paper . In so far as one denies the restriction on the use of the background language that I have advocated, he refuses to acknowledge the question to which nominalism is one answer. He thus avoids properties , not by adopting a nominalistic gambit, but by refusing to play the philosophical game. A classical nominalist would recognize the need to link the predicates like 'W 1 ' of (III) with 'something' of (I) , and not some term or statement of (IV) . He would deny the need to link such a term to a certain kind of entity. At times, some of the philosophers we have considered appear to be classical nominalists, while at other times they adopt the gambit we have discussed in order to defend their views. They thus waver between classical nominal ism and, as it were , nominalism by default. Notes 1

2

3

4

5

For some related comments see H. Hochberg , 'Metaphysical Explanation· , in Metaphilosophy l ( 1970) : 139- 165. Goodman's views are presented in his book Languages of Art (Indianapolis: 1968) , especially pages 50-68. However, my presentation depends in part upon arguments he presented in talks and responses to questions at the University of Minnesota in the spring of 1973. Languages of Art, pp. 56--57 , 59 . One familiar with Goodman's views knows th at they involve many more and complex themes than the line of argument considered here. The consideration of predication and qualia in The Structure of Appearance and the role of the notion of exemplification in his attempt to deal with questions of symbolism and metaphor in art mentioned above are examples. (For critical analyses of some of these themes see G. Bergmann, 'Particularity and the New Nominalism', Methodos 6 ( 1954) : 13 1-147 and Realism (Madison: 1967), pp. 17-59 ; A. Hausman and F. Wilson, Carnap and Goodman : Two Formalists (The Hague : 1967) . In spite of all the complexities that would arise in an attempt to do expository j ustice to Goodman's views. the simple line of argument I am considering is a crucial part of one way Goodman seeks to block the attempt to force the acknowledgment of attributes in an adequate ontology. In so far as this line of argument is inadequate, so is any nominalistic view that depends on it. Ibid. , pp. 51, 54-55, 58-59. This use of 'is true of' is thus connected with the talk of truth and satisfaction that we shall deal with in part I I . Sometimes, however, Quine seems to express a view more like that I have attributed to Goodman. For a detailed consideration of Quine's view and its problematic consequences see 'Nominalism, Platonism, and Being True of in this volume, pp. 150 f. Since quibbles might arise as to whether Quine is a nominalist due to his views about classes and numbers, it should be stressed that I am talking about nominalism in terms of how one handles the relevant issues with respect to the simple domain (I) and the truth that a and c are both white. In a similar vein, I am not concerned with the Quinean theme that there are no ontological questions or commitments in such a simple case. since we do not require ( and in fact do not have) quantifiers.

1 82

Wilfrid Sellars, 'Naming and Saying', in Science, Perception and Reality (London and New York : 1963) pp . 225-246 . For a commentary along such lines see Robert Ackermann 'Perspicuous Languages', in The Ontological Turn , eds. M. Gram and E. D. Klemke (Iowa City: 1974) . 7 The need for a neutral name form to be used in relational patterns is a feature of Sellars' view that I am not sure he fully appreciates. It not only involves a radical distinction between relational and non-relational predication, but it can be taken to embody a recognition of the difference between the obj ect as such and the obj ect as exemplifying a property : the neutral form of the name as opposed to the name's occurring in a font standing for a property ( as I would put it). See Naming and Saying, pp . 232-236. 8 This is not to say that Sellars explicates ( 4s) by such a use of 'means'. The reverse would be closer to the truth. As in the case of Goodman, an exposition that would do Sellars justice would require the consideration of a complex of themes. In this case the themes have to do with the problems surrounding notions of intentionality, proposition, meaning, and mind. (For related discussions by Sellars see ' Notes on Intentionality' , The Journal of Philosophy 6 1 ( 1964): 655-065, and "Truth and ' Correspondence "' , reprinted in Science, Perception, and Reality, pp . 197-224). Nevertheless, what I have said does simply put what Sellars' views involve that is relevant for the issues at hand . As in the case of Goodman, I do not think that any distortion is involved in that the theme I am concerned with is an essential part of Sellars· gambit . Moreover, I am convinced , but cannot argue here, that Sellars' analysis of intentionality, while insightful and cogent in many details , ultimately fails due to his basic nominalism. 9 My presentat ion of Sellars' view relies on his presentation of his viewpoint , as I understood it , in a series of lecture s , responses to questions , and arguments he gave at the University of Minnesota in the fall of 1973. 1 ° For a detailed consideration of some other aspects of Quines's views see 'On Pegasizing' , pp. 10 1 f. and 'Professor Quine , Pegasus, and Dr. Cartwright' , pp. 86 f. in this volume . 1 1 Quine's materialism has been criticized in detail in 'Of Mind and Myth', pp. 353 f. in this volume. 12 W . V . Quine , Ontological Relativity (New York and London : 1969) , pp . 57-58. 1 3 D. Davidson , 'True to the Facts' , The Journal of Philosophy 66 ( 1969): 757-758. I will follow Davidson in speaking of an obj ect , a , satisfying a sentence 'W 1 a' as well as an 'open' sentence . It is worth noting that any obj ect may be said to satisfy 'Wa' , provided that a is white . 14 D . Davidson, 'In Defense of Convention T' , in Truth, Syntax, and Modality, ed. H. Leblanc (Amsterdam : 1973) , p. 80. 15 A. Tarski , 'The Concept of Truth in Formalized Languages', Logic, Semantics, Metamathematics (Oxford : 1956) , p . 189. 1 6 Ibid. , p . 188. 17 To hold that Davidson is unfairly linked with the nominalistic gambit , since he acknowledges abstract entities like sequences, would be as pointless as suggesting that Quine , by recognizing classes for mathematical purposes, could not be taken to play the nominalist's game. Th e most that could be said is that Davidson , unlike Goodman and Sellars, does not knowingly and explicitly accept the nominalist's gambit . 1 8 Such a claim invites a charge of vacuity against the appeal to facts. That is another matter. For a related discussion of such questions see 'Facts and Truth' pp . 279 f. in this volume. 19 I am indebted to discussion with Per Lindstrom for this point. 20 Tarski, in effect, pointed this out in 'The Semantic Conception of Truth ' , reprinted in Semantics and the Philosophy ofLanguage, ed. L . Linsky , (Illinois : 1972) : 30--3 1 . 2 1 Strongly influenced by the later Wittgenstein , E.B . Allaire has adopted the line of Sellars and Davidson with respect to the issue of facts. Where Sellars seeks to prevent the 6

183

22

ontological issues from arising in the case of predicates and properties, Allaire in his article 'Truth' , Metaphilosophy 6 ( 1975) : 26 1-275 , seeks to draw the line in the case of sentences and facts. In effect his line of argument is a less technical but more roundabout duplicate of Sellars' and Davidson 's, involving the same specious use of a background natural language . The issue seems to reduce to the point at which a philosopher seeks to avoid raising the question of reference between language and what language is about . The nominalists we have considered seek to avoid the question about predicates, while recognizing it for so­ called singular terms . One may look at the later Wittgenstein as refusing to raise the question at all and thus obliterating ontological questions completely. Tarski , 'The Seman tic Conception of Truth ' , p . 33.

Sellars and Goodman on Predicates , Properties and Truth

Goodman offers one concise . and characteristic, argument and makes one concise, but puzzling . claim to support nominalism. The argument is that it is mistaken to hold that ""a predicate applies initially to a property as its name, and then only derivatively to the things that have that property'' . 1 This is so since the nominalist legitimately ''cancels out the property and treats the predicate as bearing a one-many relationship directly to the several things it applies to or denotes". 2 By holding that the nominalisfs move is acceptable, Goodman mislocates the problem. Consider the predicate 'white' and the set of objects to which it correctly applies. Given the predicate and the set, we have a set of pairs, one element of each pair being the predicate, the other being a member of the set of white things. That we have such a set of pairs, given the predicate and the original set , is obvious, trivial, and irrelevant to the issue. The issue is about the set of white things and the basis for an object being a member of it. Goodman responds to this query by claiming that the "'English language makes them white just by applying the term 'white' to them ; application of the term 'white' is not dictated by their somehow being antecedently white, whatever that might mean" . 3 The problem with Goodman's claim is that it is not clear what he asserts. No one would accuse him of failing to distinguish between applying predicates in speech and applying coats of paint in a studio. Yet, what does it mean to claim that the English language makes things white? Surely he is not merely claiming that what is called 'white' in English is called 'white' in English. Nor can he be pointing out that we must have the term 'white' in order to classify objects as 'white', that is by application of the predicate. The problem with Goodman's view is that it seems either to purchase truth at the price of triviality or to be significant, even startling , only when it borders on the absurd. The problem is highlighted by a peculiarity in his latest restatement of his thesis. Consider the sentence: (s 1 ) The English language makes them white by applying the term 'white' to them. Keeping Church's "translation problem' in mind, let us translate the sentence into Swedish. (s2 ) Det engelska sprdket gor dem vita genom att ge dem namnet 'white'. By Goodman's own view, we have changed a true sentence, (s 1 ) , into a false sentence , (s2 ) . For, he tells us: (s 3 ) A language that applies the term 'blanc' to them makes them blanc. 4 Thus, (s2 ) is false, for the English language does not make them vita, if I, like

185

Goodman, may use literal nonsense to make a point. 5 Yet, for Goodman , it is embarassing, or ought to be, for (s2 ) is false and (s3 ) is not a sentence of English.6 Consider, instead, (s4) The English language applies the term 'white' to white things; and (s5 ) The English language makes them 'white' by applying the term 'white' to them. (s4 ) does translate into a true Swedish sentence. (s5 ) may be taken to state Goodman's view that objects exemplify predicates not properties.7 So understood, the Swedish translation of (s5 ) will have the same truth value as (s5 ) . But this leads us to suspect two things. First, (s 1 ) is simply a dramatic way of putting (s4 ) . The drama provides the philosophical content obviously lacking in (s4 ) . Yet, to get that content Goodman must make a claim that is, at best, only true in one language as well as a claim that is not literally sensible in any language. Second, if (s 1 ) is taken as a forceful way of putting (s5 ) , then Goodman argues in circles since he can no longer justify the nominalistic claim that objects exemplify predicates, rather than properties, by appealing to the truth of (s 1 ) . For, (s 1 ) , being a restatement merely reiterates, rather than justifies, the nominalistic view expressed by (s,J Sellars makes two claims. The first is that I mistakenly take -the background language uncritically, at its face or surface value. The second is that the nominalistic gambit I attributed to him, along with Goodman, Quine, and, derivatively, Davidson, does not apply in his case, since I ignored three major aspects of his view. (A) A predicate, such as "'W i"' , of a schema as well as the English '"'white'" and, more generally, ' · white · ' are ·metalinguistic sortals ·. (B) 'The color white (whiteness) is a property' is construed as 'The (or a ) · white · is a predicate'. And, · a exemplifies whiteness' is construed as 'The (or a) · white · is true of a'.Moreover, the latter holds if · · a is white · is true·, which is to say that · · a is white · s are semantically assertible' . (C) The connection of the predicate ·white' with extra-linguistic reality essentially involves the connection of subject-predicate statemen ts of which the predicate is "white', and the subject term is a name (or a demonstrative) , with white objects. 8 As the dispute centers on (A), (B), and (C), I will comment very briefly on the first claim . What is it to use the background schema or language uncritically? Sellars agrees that we all recognize properties, such as colors, and facts in some sense that does not give comfort to a Platonist or a Logical Atomist of Russell's type. 9 Thus, we may speak, as we do speak , of the color white as the color of an object. The question is how to construe such locutions in the framework of a philosophical position so that we may acknowledge the truth that an object (along with other 1 86

objects) is white. Moreover, that they are so is not merely a matter of our use of language and of the existence of the objects.There is a clear sense in which my naming an object •a' reflects (I) a stipulative or •linguistic' connection between the sign and the object , (II) the acknowledged existence of the object, and (III) the framework regarding the grammatical and inference rules concerning names. The question is whether the application of predicates to objects involves something more and, if so, what. To aid those who pretend to be puzzled about the problem of universals, I considered the issue in terms of the interpretation of a simple schema that would contain transcriptions of singular, factual claims, such as This is white' . By doing so, I no more took the background language at its face value than does Sellars, or anyone else who recognizes that we speak of objects having colors in contexts that are philosophically unproblematic. 1 0 I would have done what Sellars accuses me of doing if I had held that the way we use the background language settles the issue , since we recognize and speak of colors, shapes, relations, etc. that are common to different objects and pairs of objects. This I did not do. Rather, by constrasting different ways of talking about linking such a schema to ·extra-linguistic' reality, I sought to explicate some strands of the issue between nominalists and realists. This involved pointing to a way in which some nominalists avoid, rather than resolve, the classical issue. Sellars claims that he does not avoid the issue. This is supposedly shown by (A), (B), and (C) . The crucial claim is (B), as it was clear from my discussion that Sellars's dotted predicates are ·metalinguistic sortals' and ( C) figures in the unpacking of (B) . I had tried to focus on the skirting of the issue by nominalists who use the truth

to resolve the question about the interpretation of ·w 1 ' . Sellars replies that he does not make the move from the left-hand sentence to the right­ hand sentence of the 'biconditional'. He does not because he uses: (T2 ) 'W1 ' is true of a assertible.

=

· W 1 a · is true - · W 1 a · s are semantically

Thus, Sellars, unlike Davidson, does not offer the trivial '··w 1 a' is true = W 1 a" as an answer to the philosophical problem of truth. As the problem of abstract entities is reduced to the problem of the explication of the concept of truth, via (B) and (T2 ) , Sellars has a point.It was unfair to link

187

him, and Goodman , with Davidson , without further discussion . For, both Goodman and Sellars proceed to give further accounts . Goodman's gambit drives him into 'relativism' and talk of m aking worlds with words . Sellars' move involves the explication of 'semantically assertible' and suggestive overtones of a coherence theory of truth . 1 1 Yet, in the end, Sellars' gambit invokes the basic move I had questioned. This we can see by tracing his pattern in his own words . He recognizes the obvious question about 'semantic (S) assertibility' and replies: . . . S-assertible by us. For truth in the 'absolute' sense is, i n its o wn way, language relative , relative to our language . 1 2

And this means that a statement is ' . . . S-assertible in our conceptual structure ' since ; . . . the fundamental form of 'true' is true in conceptual structure CS i . The 'unqualified' sense of 'true ' pertains to the special case where CS i is our conceptual structure . . . . 13

But, our conceptual structure , and S-assertibility , must come to terms with matters of fact and the truth about the world as we are 'haunted by the ideal of the truth about the world' . 14 This accommodation takes place via the role of statements as 'pictures' : The fundamental j ob of singular first-level matter-of-factual state ments is to picture , and hence the fundamental j ob of referring expressions is to be correl ated as simple linguistic objects by matter-of-fact ual relations with single nonlinguistic objects. 15

For , pictures can be more or less adequate or ·correct' : Truth , as we have seen, is not a relation . Picturing , on the other hand , is a relation , indeed a relation between two relational structures. And pictures , like maps , can be more or less adequate . The adequacy concerns the · method of projection' . A picture (candidate) subj ect to the rules of a given method of proj ection (conceptual framework) , which is a correct picture (successful candidate) , is S­ asse rtible with respect to that method of proj ection . Thus the S-assertibility of a matter-of-factual proposition . . . is a matter of · fa · s being elements of correct pictures of the world in accordance with the semantic rules of CS 1 • • • in thinking of pictures as correct or incorrect we are thinking of the uniformities involved as directly or indirectly subj ect to rules of criticism . . . . Linguistic picture making is not the performance of asserting matter-of-factual propositions. The criterion of

188

the correctness of the performance of asserting a basic matter-of-factual proposition is the correctness of the proposition qua picture , i . e . the fact that it coincides with the picture the world-cum-language would generate in accordance with the u niformities controlled by the semantical rules of the language . 1 6

And how are we to understand the role of such uniformities and semantical rules with respect to the issue at hand ? First, we must understand that for referring expressions it will not do to say that ' their job i s to refer to certain objects'. Rather, we must note that (non­ demonstrative) referring expressions belong to the 'natural order' and are connected ' with objects in a way which involves language entry transitions, intralinguistic moves ( consequence uni formities) and language departure transiti ons (wi llings-out-loud)' . 1 7 In its way, what Sellars says is true enough. Understanding, in one sense, how names function requires informati on gathered from a wi de range of disciplines covering sociological, psychological, and ( empirical) linguistic factors. It will also involve questions of logical inference and 'formal semantics'. What is problematic is whether noting all this enables one to resolve the problem of universals. B ut, let us see how the argument proceeds. N oting what we must note about referring expressi ons involves recogni­ zi ng that there must be a 'stable framework' of propositions which contai n such expressions. These must describe the spatio-temporal lo­ cation of the referents and of the language user. 1 8 This, in turn, will in­ volve the use of "characterizi ng expressions' and "in particular, those characterizing expressions which stand for spatial and temporal rela­ tions'. 1 9 All this will involve both the axiomatics of spatio-temporal di scourse and the participation of the language user in the course of events ' which pragmatism has stressed from its inception'. 20 More­ over, . . . in order for 'a', ' b ' , etc . to be correlated with obj ects , the spatio-tem­ poral story tellings in which they occur , however schematic, must be depictings . This means that certain matter-of-factual relations , satisfied by 'a's , 'b's, etc . , as elements in the language , must be counterparts of relations satisfied by the obj ects which they represent in the pictures. 2 1

We may then naturally wonder about the status of the relations which the objects satisfy and, hence, about the original question concerning the linking of predicates with extra-linguistic reality.We are told: We saw . . . that for a predicate to stand for an attribute or a relation is for it to be of a certain kind . Thus, to stand for tri angul arity is to be a · triangular· . What is it to be a • triangular · ? It is to be an i tem which does the j ob done in the base language by

189

'triangular' s . Specifically, it is to give a singular term concatenated with it a counterpart character, T' . It is T' individual constants which correctly picture triangular obj ects, provided that the individual constants are correlated, as above , with the obj ects. But although T' individual constants are correlated with T obj ects, the concept of this correlation is not the analysis ofwhat it is for T' individual constants to standfor triangularity, nor does it explain what it is for T' individual constants to denote triangular obj ects . The correlation between obj ects and their linguistic pictures must not be confused with the pseudo-relations standing for and denoting. Thus, that 'triangular' s stand for triangularity essentially involves the intralinguistic consequence uniformities governed by the consequence rules (axiomatics) of geometrical pre dica tes . 22

Here, the careful reader of Sellars may well obj ect that we did not see what Sellars tells us that we saw. Rather, we were merely told what Sellars tells us once again. There is, however, a difference. We are now in the middle of an argument that is designed to substantiate what we were earlier told. Such an argument thus purports to be an argument for Sellars' brand of nominalism. But, the argument has suddenly lapsed into a reiteration of the thesis, and nothing of substance is added by asserting that to take a predicate like 'triangular' tostandfor triangularity involves the geometrical truths connecti ng that predicate with others . Nor does it matter if we put things as Sellars does in terms of · triangular · s and the 'natural correlation' of such constants with the obj ects they 'picture' ; for, once again, one merely reiterates a theme without presenting an argument. Of course, to change the example, the predicates 'white' and 'black' would not 'stand for' the colors that they do if, as the world is , we did not recognize that '(x) (x is white = x is not black)' is true. This is not to say that the claim that "white' stands for the color that it does asserts that truth. It is also not to provide a ground for holding that such truths sum up what is meant by saying that "white' stands for the color. Aside from sidetracking the issue with the interj ection of points about the natural correlations between uses of tokens and observations of obj ects , as well as other natural connections of the same sort , Sellars unfurls the old holistic battle flag. Without using the term ' meaning' , he invokes the argument that predicates get thei r meaning from the holistic context, including such correlations and connections . Hence, we need not appeal to the realist's attributes to provi de a referent, as meaning, since we have the context to supply such 'meani ng'. One may then understand the claim that a predicate 'stands for' an attribute in a harmless way that does not involve the 'pseudo­ relation' of standing for, si nce all that can reasonably be meant is that the pred icate plays the role it does in the language. Thus, abstract singular terms are not part of the 'illustrative' or ' depicting' component of a

1 90

sentence : "" . . . ultimately, abstract singular terms must relate to the ·truth move' in which the non-illustrating component falls away , as in That snow is white is true Snow is white". 23 Yet , all this is to no avail until Sellars spells out just what the role of a predicate is in those basic matter-of-factual statements. These , recall, allow us to get at the ·truth about the world', as more or less correct pictures (or as elements of correct pictures). This role finally is spelled out for us: •fa's (in L) correctly picture O as 0 must be carefully distinguished from 'fa's (in L) stand for that 00 , and that 00 is true The former tells us that (in L) utterances consisting of an 'f' concatenated with an •a' are correlated with 0 , which is 0 , i n accordance with the semantic uniformities which correlate utterances of lower-case letters of the alphabet with obj ects such as O , and which correlate utterances of lower-case letters of the alphabet which are concatenated with an T with obj ects which are 0 . 24

Here we have the crucial theme . For all the detail and wealth of description that the story involves, the role of predicates in correct pictures (or elements of correct pictures) comes down, as far as the problem of universals is concerned, to the use of '0' in the above quotation. It is buttressed by something Sellars says about the realist's move (the 'contrary' position) in connection with the 'language entry role' of predicates : The crudest form of the contrary position consists in taking the language entry role of a perceptual predicate , the fact that statements i nvolving the predicate are correct responses to obj ects which exemplify the perceptual character for which it stands, to constitute the fact that it stands for this character . 25

For Sellars, the fact that the response is correct is the key to understanding what could mislead such a realist into claiming that the predicate stands for the character. Yet, the response being correct involves its being a token of a (relatively) correct picture (or element of a correct picture). And, it is so, in our case , since a is white. We can see why this will not do when we recall the pattern of Sellars' moves.

191

The original question was about the linking of 'W 1 ' to extralinguistic reality.The realist employs: (1) 'W 1 ' stands for the color white. (2) 'a' stands for the object 0 . (3) 'W 1 a' states that the object O has the color (property) white. Sellars supposedly avoids ( 1 ) and (3), and hence realism, by reducing the problem of abstract entities to the problem of truth.Thus, in place of (1) and (3) we get: (4) 'W 1 a' is true = 'W 1 a' s are semantically assertible. And this, in turn, was to be understood in terms of: 'W 1 a' s are semantically assertible = 'W 1 a' s belong to our conceptual structure. The condition, recall, for a singular matter-of-factual statement belonging to our conceptual structure is that it be a correct picture of extra-linguistic reality. And, recall further, that'W 1 a ' is such a correct picture if two conditions are fulfilled: (I) 'a' s are correlated with O� and (II) 'W 1 a' s are correlated with O which is white. (Such correlations are in accordance with the'semantic uniformities' and'rules' . ) But, it appeared as if we would avoid (1) and (3) by treating the sentence as a whole in terms of (4).Yet, the treatment of (4), in terms of (I) and (II), involves our separating the subject term from the predicate and appealing to the different linguistic roles predicates and singular terms have in ordinary use. [It is really not relevant that, instead of putting it this way, Sellars gives a sketch of the complexities of the psychological, sociological, and 'linguistic' context for the utterances of tokens of predicates, of the laws connecting the predicates with other terms, and of the "logical' apparatus that applies to predicates.] We thus end up with (I) and (II) and the claim that whereas 'a' is correlated with 0 , 'W 1 a' is true (or "W 1 ' is correctly ascribed to 0) since ('a' is so correlated and) 0 is white ! In the end, Sellars thus does what his fellow nominalists do.They use the predicate ' white', predicatively, to avoid recognizing properties. To see the significance of this look back at (3) and ( 1 ) . 26 In evaluating Sellars' arguments we must not be misled by the emphasis he, like Quine, places on the role of abstract singular terms as opposed to predicates. It is as if Sellars seeks to convince us that one blocks the road to realism by construing (1) as: (1') · W 1 • s are · white · s.But this is no more effective than construing (1) as: ( 1 '') All objects to which 'W 1 ' 192

applies are objects which are white. If one thinks ( I " ) helps avoid abstract entities, since one uses "white' as a predicate, rather than as (part of) an abstract singular term, as in ( 1), he clearly argues in circles. For, the original problem concerned that use. The point here is that Sellars, like the other nominalists I considered, ultimately does argue that way. They all merely reiterate the ordinary use of predicates, as characterizing expressions, to avoid the realist's problem. That Sellars does this comes out quite clearly in a passage towards the end of his discussion of picturing. Speaking about: ' R(a, b ) · (in L) correctly pictures 0 1 and 0 2 Sellars tells us that : . . . if we proceed to j usti fy such a statement we must say . . . because R * ( 'a' , ' b ' ) and R(0 1 , 0 2 )

rather than . . . because Concat( ' R · , 'a , b ' ) and Exempl( R-ness(0 1 , 0 2 ) ) The ' R · of ' R( 0 1 , 0 2 ) ' stands for a complex matte r-of-factual relation and not the pseudo-relation of exemplification . 2 7

The passage reveals Sellars' ultimate appeal to 0 1 standing in R to 0 2 . This, of course , is the sort of statement that gives rise to the problem of universals . Moreover, Sellars does not merely appeal to it as an ordinary truth from which we start , about which problems arise , and which we must analyze. He invokes it at the final and crucial stage of his analysis and, hence, of his argument for nominalism. That is why his argument is not an argument. Sellars asserts that we need not or must not express the fact in question in terms of 0 1 , 0 2 , and R(-ness) standing in the exemplification tie or nexus. We need not do that since we can merely hold that 0 1 stands in R to 0 2 • But this blurs a crucial distinction. There is, first, the realist's claim that a cogent analysis (of the sentence of the schema and, hence, of the original natural language sentence), which is presented in our interpretation of the schema and our comments about it, must acknowledge two particulars, a relational property, and a tie or combinatorial relation of exemplification. There is, second, the question as to whether the realist, in giving the analysis he gives , is forced to (I) replace predicates of the schema by abstract singular terms, (II) add a unique relational predicate, such as 'Exempl', to the schema, and (II I) replace sentences like 'R(a, b )' by sentences like "Exempl (R- ness, a, b )' or "Exempl (R-ness(a, b) ). ' The realist is claiming that 'R(a, b )' is true on the condition that the objects exemplify the relation (in the proper 1 93

'direction' or structure). Sellars rebuts the realist by asserting that we need not and must not make such a claim, since we need only hold that 0 1 stands in R to 0 2 and not that 0 1 and 0 2 exemplify the relation R . Taken one way, all he does is reiterate that we may be content with noting that R(0 1 , 0 2 ) , as, earlier, we noted that a was white. Taken a second way, he points out that we need not add 'R-ness' and 'Exempl' to the schema in place of (or in addition to) the predicate 'R' (or the unique juxtapositional pattern of Jumbelese).Taken a third way, he claims that realism falls to Bradley's paradox. The first 'way' begs the question. The second way requires an argument that the realist must so alter the schema. But Sellars, like Quine, merely assumes, without argument, that the realist must do that. The third way does provide a challenge and a problem that the realist must meet. That I cannot do in this paper. 28 Here, I have been calling attention to that side of Sellars' nominalism which is summed up in his conclusion that " 'R' stands for a complex matter-of­ factual relation". If we now question such an appeal to the relation R, we may guess what Sellars will tell us : 'R' is not being used as an abstract singular term and, hence, is not to be taken as connected with a "basic entity". To understand the significance of that, we must begin at the beginning with : 'R(a, b )' is true = · R(a, b) · s are semantically assertible. 29 Notes 1 N. Goodman , 'Predicates Without Properties· , Midwest Studies in Philosophy , vol. I I , 1977 , pp. 212-213. 2 lbid . , p . 2 1 3 . 3 Ibid. , p. 2 12. 4 Ibid. , p. 21 2. 5 It wi ll not do to obj ect that there is no 'translation problem· since no statement of one language is translatable into a statement of another language . Here. I am using ·translation' as a teacher of a language course uses it. To obj ect to that use , in this context , is pointless. 6 One might obj ect by holding that (s_J is a sentence of a ·metalanguage· which includes English and French words. Aside from the obvious artificial ity of such a move , the problem persists for some sentence until we take a 'universal metalanguage·. One would then have to endlessly extend it for 'newly constructed languages· . It is worth noting that the situation thus created is analogous to the one Russell exploited in his classic argument for relational universals in The Problems of Philosophy . (Oxford : 1912) and ·on the Relation of Universals to Particulars' , in Proceedings of the A ristotelian Societv , l 9 1 1 - 19 1 2 . 7 O n this claim see m y ' Mapping , Meaning, and Metaphys ics· , in this volume , pp. 1 5 7 f. 11 W . Sellars , 'Hochberg on Mapping , Meaning , and Metaphysics' , Midwest Studies in Philosophy . vol . I I . 1977 , pp. 2 1 9-220 . " They are not taken as 'basic entities' . as he puts it . Ibid. , p. 214. 10 Sellars accuses me of making an ordinary language style move. That is a pointed riposte to

1 94

my similar comment about his gambit. The real issue is about the role that the obvious truth

that a is white plays in one's analysi s. As I hope to show , it plays too crucial a role in the

analysis Sellars offers . Sellars attempts to reconcile features of a coherence account with a correspondence theory's notion of the picturing role of atomi c propositions. The latter aspect is more relevant for the present discussion . See his Truth and Correspondence', reprinted in Science, Perception, and Reality , (New York: 1963) . especi ally pp . 197-200 , 203 , 207-2 15. 12 W. Sellars , Science and Metaphysics (New York: 1968), p. 132 . 1 3 Ibid. , p . 133 . 14 Ibid. , p. 135 . 15 Ibid. , p. 124 . Something Sellars says about arithmetical truth helps us to understand what is going on : In the case of arithmetic, for example , the concept of truth (S-assertibility) coincides with that of provability. It follows, of course , from Goedel's results that, with respect to the conceptual structure (in the sense of axiomatics) to which it belongs, not every arithmetical proposition is either true or false . It also follows that not every arithmetical proposition which is in some sense true is true in the absolute sense , i . e. with respect to our current conceptual structure, if this is taken to be an axiomatics. (Ibid. , p. 1 35 ) . 16 Ibid. , pp . 135-136. 17 Ibid. , pp . 125-126 . I S Ibid. , p . 126. 19 Ibid. , p . 126. 20 Ibid. , p. 1 26. 2 1 Ibid. , p . 1 27. ,, Ibid. , pp . 1 27-128 . The notion o f a � counterpart character .. should not cause t h e reader a problem. It is merely a character of an individual constant. And , we understand that constants so characterized are correlated with objects characterized by a character ( of which the counterpart character is the counterpart). This is a familiar theme connected with Sellars' ideal language Jumbelese . 23 Ibid. , p . 130. 2 4 Ibid. , p. 136. 25 Ibid. , p . 1 28. 26 The point to note is that an abstract singular term , as commonly (or unproblematically) understood, occurs in the 'interpretation rules' and not in the schema. Thus, one does not use the ordinary language transcription ('white'), of the predicate ('W 1 ' ) of the schema , predicatively - as Sellars ultimately must and does . Who , then, is the ordinary language philosopher? 2 7 Science and Metaphysics, p. 1 37 . The use of 'R *' merely indicates Sellars' basic move in Jumbelese. Thus , its use signifies a relation that obtains among names. 28 A fundamental theme of Sellars' gambit is a systematic attempt to demonstrate that the realist cannot resolve Bradley's paradox . I have di scussed this aspect of Sellars' attack in my book Thought, Fact, and Reference, 1 978 , University of Minnesota Press. As I see it, this is the most cogent aspect of Sellars' campaign against realism and sets him apart from Quine, who merely assumes that the distinction between singular and general terms provides a basis for nominalism. 29 Note how Sellars ends his reply in Midwest Studies, vol, II, p. 222 . 11

Russell's Proof of Realism Reproved

In 'On the relations of universals and particulars', 1 Russell offered what has become a classic proof for the existence of universals. In a well-known paper, Alan Donagan defended Russell's and Moore's realistic position by rebutting some standard attacks on it. 2 He did not attempt to show that Russell's proof was viable. In this paper I shall reexamine Russell's purported proof and attempt to show, first why it is not satisfactory , a nd, second, that properly amended and explicated it does establish that universals exist. Russell's argument starts from a case of two objects resembling each other in a respect or quality. He then asks whether the resemblance is a particular or a universal. One presupposition i nvolved is that we can construe the facts that each object has 'the' quality in terms of thefact that one object resembles (i n the appropriate way) the other. This I do not wish to question here , for, while I do not believe it to be a viable claim, it does not really affect the basic issue since that can be argued either from the assumption of resemblance or of two objects being qualitatively alike. Assume, then , that 'S(a , b)' is true , where 'S' transcribes 'resembles' or 'is similar to' and 'a' and 'b' name the objects. The first step that Russell makes is to claim that S is a relation. The second step is the 'proof' that S, being a relation , is a universal . For the time, let us not question the first step. Assume, then, that S is a relation. Russell argues as follows. Suppose the relation S is a particular, rather than a universal. Consider a second case of two similar objects. The relation , in that case, would also be a particular. We then have two relations that are particulars, S and, say, S 1 • But, S and S 1 are, themselves, similar (to each other). Hence , there is a third relation , S 2 , that obtains between S and S 1 and which is also a particular. Yet , S2 , being a similarity relation, will be similar to S and to S 1 • Thus , we have two further relations, SJ and S4 , relating S 2 to S and to S 1 • These , too , are particulars. We thus initiate an infinite series. This means that we either recognize some relations to be universals or we have not really solved the problem , since there must be new facts and relations appealed to at every step . The regress is vicious in that at any stage we have not specified sufficient conditions for the truth of the sentences involved. Thus, we start with the sentences

1 96

(1) ( 2)

S(a , b) S 1 (c, d) .

Taking them to be true forces us to recognize facts composed of the members of the sets ( lJ { a, b , S } (2) {c, d, S 1 } i n an appropriate arrange ment . Call such facts ' ( l f)' and '(2f)' , respectively. On the view Russell see ks to refute , we must also recognize

to be true . Hence , we must recognize a fact composed of the members of the set

Call the fact ' (3f) ' . But, clearly ( lf) and (2f) do not suffice , as truth conditions , for what Russell's opponent must take to be true , (3) and the rest of the 'chain ' . Given the view Russell seeks to refute , (3) is a ·consequence ' , since it follows from ( 1 ) and (2) and the view in question . 3 Thus , the regress is not harmless . In the case of the familiar , but harmless , infinite chain : (C 1 ) S (a, b) , 'S(a, b)' is true , "" S(a , b)" is true' is true , . . . there is no regress . Given the apparatus for employing the predicate 'is true' and the quoting device , it does follow from the existence of the fact that grounds the truth of 'S(a, b ) ' , i . e . ( lf) , that all the sentences of (C 1 ) are tru e . B y contrast , ( l f) and (2f) do not suffice to ground the truth of the members of the chain :

Thus , one must appeal to a new truth condition , or fact , at e ach step . The account , at any stage , reg uires a further account or ground . If one takes S to be a universal , and thus identifies S with S l ' no further account is required , since the chain (C2 ) then stops with ( 1 ) and (2) . Hence , the facts ( l f) and (2f) seem to suffice to ground the truth of ( 1 ) and (2) , as well as of the sentences logically implied by ( 1 ) and (2) . I say 'seems' since one may argue , in the fashion of Bradley , that on the realist 's account problems 197

arise regarding ( I) the connection of the particulars and the relations in the facts and (II) the 'fact' that S is a universal . Such questions are not my concern in this paper, though it is worth observing that variants of such questions also arise from the 'nominalistic' view that Russell disputes. As I have presented Russell's argument, it is clear that he does not argue that ( lf) and (2f) do not suffice to ground ( 1) and (2). 4 Thus, it is irrelevant to attempt to refute him by arguing that they do suffice for that. Yet, the present way of construing his argument shows that Russell mixes two claims. One claim is:

The other claim is:

(/3) if S = S1 , then S is a universal and not a particular. Russell's argument, as reconstructed here, does prove (a) . He believes that he has proven (/3) as well. His notion of 'universal' involves being common to two facts. Yet, trivially, particulars can be common to two facts. Hence, he must mean that something is a universal if it is common to two facts and is a relation, as opposed to the terms of the relation in the facts. He thus appeals to the asymmetry of exemplification or to the structure of facts like ( l f) and (2f). As he uses "universal", (/3) is, then, more explicitly put as

(/3 ' ) if S=S 1 and if S is exemplified by a and b and S 1 by c and d, then S is a universal and not a particular .

Then, given (a) , (/3' ) , and: (y) (lf) and (2f) are such that a and b exemplify S , while c and d exemplify S 1 , but a, b, c and d cannot be exemplified by anything � it follows that S is a universal and a, b, c and d (as assumed) are particulars. With the argument so amended , an implicit theme becomes explicit. We so understand the structure of facts like ( 1 f) and (2f) that we claim that S can be exemplified in more than one fact, while a and b cannot be exemplified in any fact. This is a basic idea involved in the distinction between a rel ation ( or quality) and a particular. The asymmetry of exemplification is thus appealed to. Russell assumes, and does not prove , that facts have a structure that is asymmetrical in this sense. Thus , he assumes, and does not argue for, (y) , while (/3' ) merely explicates the use 1 98

of "universal'. We shall return to this point shortly. Before doing so I wish to consider a simple objection that one may raise to the argument and which Russell did not face. The objection concerns the assumption that there is a relation , which he then argues is a universal. Some would challenge this initial move and hold that to assert that a resembles b is not to make an existential claim, explicitly or implicitly, about resemblance as a relation. By assuming at the outset that there is a relation, Russell can be accused of begging a crucial question. Yet, he does not beg the question. For, he need only assume that there must be a condition fo r the truth of ( 1 ) which cannot be characterized solely by reference to a and b. One must specify , at least, that-S(a, b) . One who then refuses to recognize the question Russell proceeds to raise about S merely ignores the issue. To hold that ( 1 ) is true since a resembles b is to refuse to specify what a truth condition is. Russell implicitly does stat e what he takes such a condition to be: a fact. Of course, Russell must answer some obvious questions about 'facts' an d their 'structure' . This he attempted, in different ways at different times, but he never assumed that the matter ended with the claim that a resembles b. Russell's existential assumption about the relat ion S can thus be put in a less problematic way. The point is that if one takes the truth conditions for atomic sentences to be atomic facts, then he must specify what he takes to be involved in such a fact. Russell' s proof is linked to a specific 'theory' of truth and depends on the viability of such a correspo ndence theory. This, however, is not to say that he begs any questions. I t is merely to note that the matter is more com plex than his argument, taken in isolation, suggests. After all, Russell consistently argued for a correspondence theory of truth at many periods of his writing. I t is the heart of the philosophy of logical atomism, which was to be formulated shortly after the paper on universals and particulars. We may get at a problem , involved in the implicit appeal to the structure of facts in Russell' s proof, by noting that his proof explicitly appeals to the two facts ( l f) and (2f) . Assume a domain with two objects and 'one common property' . The realist would hold that we must recognize the property as a universal in such a case. Yet, Russell' s argument would not seem to apply. However, an extension of it does. The fact in question is, assume, that-S(a, b) . We, then, also have 'S(b, a)' as true. I f one holds that there is an additional fact, that-S(b, a), we can directly apply Russell's pattern. Moreover, we should acknowledge the fact that-S(b, a) for two reasons. Both 'S(a, b)' and 'S(b, a)' are atomic sentences. Hence, neither is a logical consequence of the other. We cannot then hold that the fact which is the truth conditio n for the one suffices as the truth condition for the other. Given that S is symmetrical, i. e. that '(x) (y) (S (x, y) :=J S(y, x))' is true, we can infer either 'S(a, b)' or 199

'S(b, a) ' from the other. But the fact that S is symmetrical is not a matter of logic.Moreover, consider 'S(a, b) � S(b, a) ' . If we take only one fact to be involved, then, speaking somewhat loosely, we can take that conditional to express the claim that if there is the fact that-S (a, b) then there is that fact. This will not do.We can only sensibly speak about what is involved, assuming we recognize facts as truth conditions, by holding that if there is the fact that-S(a, b) then there is also the fact that-S(b, a) . To talk of facts requires one to talk of their structure. The appeal to, or recognition of , S as a relation must involve the recognition of its direction or ordering, whether the relation is symmetrical or asymmetrical. Clearly the structure or ordering of the elements of the fact that-S(a, b) is different from that of S(b, a) . 5 Hence, the facts cannot be the same. Thus, Russell's use of a symmetrical relation of similarity or resemblance does not preclude the use of his argument pattern in our restricted domain. Given that S is symmetrical one might object to the claim that 'S(a, b) ' and 'S(b, a) ' represent different facts.There are two reasons to discount the objection. First, neither 'S(a, b) ' nor 'S(b, a) ' follows from the other. We require the additional assumption that S is symmetrical to derive one from the other. Thus, at best one can claim that given the facts that, say, S(a, b) and that S is symmetrical, we need not acknowledge a third fact, that-S(b, a) . Thus, if one holds that it is a fact that S is symmetrical , we require that fact in addition to the fact that-S (a, b) to ground the truth of 'S(b, a) ' . If one holds that S being symmetrical is not a matter of fact but of logic, he can hold that 'S(a, b) ' and 'S(b, a) ' represent the same fact without appeal to an additional fact. But such a claim is problematic. Alternatively, one may hold that the fact that S is symmetrical can be avoided in that statements like " (x) (y) (S(x, y) � S(y, x) ) ' are true in virtue of pairs of facts: [ S(a, b) , S(b, a) ] , etc. Thus, if one avoids ·generar facts in this way, reminiscent of Russell's ·metalinguistic' treatment of the quantifiers, he should acknowledge both the fact that-S ( a, b) and the fact that-S(b, a) . 6 But , there is a second, more conclusive, reason which does not rely on a way of treating generalizations or on the denial that S being symmetrical is a matter of 'logic' . Let B be a three-term relation where we have: B(a, b, c) B(b, c, a) B (c, b, a) ---- B (a, c, b)

as true. Clearly, we must recognize the difference in the ·sense' (or direction or order) of the relation obtaining in the first three facts represented. But , if so, it becomes obvious that order must be recognized in two-term facts as well.7 200

I n recognizing the distinction between S, as a relation , and a and b as terms we make use of the asymmetry of exemplification. Thus, we might take two notions involved in the concept of a universal to be: (U 1 ) (U z )

x is a predicable if and and only if it can be exemplified. x is a universal if and only if ( a 1 ) it is a predicable and (a2 ) it can be a constituent of two atomic facts , neither of which contains another predicable.

By implicitly using ( U 1 ) and (U J , Russell makes use of a particular explication of predication. Here , too, he does not beg the question, for it must be recalled that he argues , in his paper and elsewhere , that there are particulars. This argument purports to show that there are two sorts of entities , particulars and universals , and an asymmetrical connection presupposed by a viable account of the true ascription of predicates to objects. 8 If one denies the asymmetry between S, on the one hand , and a and b, on the other , then in the case of 'S(a, b) ' , given that facts have structure, we "obtain , five additional 'facts' :

(L1)

a ( b , S) , S(b , a), b (a, S), a(S , b ), b (S, a) .

But , then , we must recognize that we have given two roles to each element and hence also allow for : (L2)

a (b, b ), b (b , a) , etc.

since there is no difference between a particular and a relation with respect to exemplification. To allow a and b to have the same logical role as S in the items represented on (L 1 ) forces one to recognize the items listed on (LJ , unless one claims that a particular can only play the role of a predicable when another constituent of the relevant fact is a relation. This gives up the game, while the acceptance of (L2 ) is patently absurd. We must, then, admit to the asymmetry of structure of the fact that­ S(a, b ) and, hence , to the difference between the relation and its terms. This, taken together with the earlier argument that the relation S is identical with the relation S 1 , constitutes the proof that universals exist. One cannot viably object that in assuming that something can be a constituent of two atomic facts as a predicable I assume too much. This is a minimal condition in that a particular can also be a constituent of two atomic facts, though not as a predicable. Clearly, to argue for or against the existence of universals, one must allow for the possibility that two facts of the form R (a, b) have a common predicable. 9 20 1

Notes 1

3

4

5

6 7

8

9

B . A .W. Russell, 'On the relations of universals and particulars ' , Proceedings of the Aristotelian Society, 1 9 1 1-12, reprinted in R . C. Marsh (ed . ) , Logic and Kno wledge, London: Allen & Unwin , 1956) , pp. 1 05-24 . Alan Donagan , 'Universals and metaphysical realism' , The Monist 47 ( 1 963) , pp. 21 1 -46 . I presuppose the appeal to logical implication as part of a criterion for deciding when facts are to be acknowledged . Thus , if one recognizes a fact grounding the truth of 'p' one need not acknowledge an additional fact grounding the truth of 'p v q' since 'p f-p v q ' is a valid pattern . For a discussion of this see my 'Negation and Generality' , in this volume , pp. 296 f. The relevant point here is that ( l f) and (2f) do not suffice to ground the truth of (3) , yet (3) is true, on the view Russell disputes, if (1) and (2) are true . However, one can argue that ( I f) and (2f) do not suffice as truth conditions for ( 1 ) and (2) since they do not suffice for the whole chain (C2 ) . For, it is reasonable to hold that a purported truth condition does not suffice if its existence , on a given analysis. involves the existence of another condition . It is, then, the existence of the two conditions that constitute the sufficient condition . But, if an infinite chain of conditions is involved clearly no finite set of conditions suffices. Whether one must recognize both a structure of a fact and the ordering of the terms that exemplify the relation need not concern us. On this question see Chapter XI of m y book , Thought, Fact, and Reference, ( Minneapolis: University of Minnesota Press , 1 978) and 'Logical Form , Existence , and Relational Predication ' , in this volume , pp . 204 f. On this treatment of quantification see Chapter XV of: Thought, Fact, and Reference. In a recent paper, 'Relations : Recreational remarks' , Philosophical Studies 34 ( 1 978) . pp. 8 1 -9, E . B . Allaire has argued that we need not recognize such order since , with any list of facts depicting a world (actual) , we can hold that both members of the pair of true sentences stand for the same fact , if the relation is symmetrical . If the relation is asymmetrical then the absence of one of the sentences of the pair would ' represent' the fact that the relation is asymmetrical . He thus avoids 'forms' of facts , but his argument suffers from two defects . First , three-term relations cannot be handled in this way. Second , his ' representation ' must contain duplicate representations of the same fact . Thus, the 'form· or ·structure· is relegated to the sentences. If one takes a more Spartan sense of representation , then either 'R(a, b) ' and 'R(b,a) ' represent different facts or only one sentence appears on the 'list ' of facts representing the world . But , if the latter, then asymmetrical relations must be represented by the occurrence of negative sentences ( or some si milar device) , and not by the 'absence' of the other member of the pair of sentences. Thus , we implicitly acknowledge both negative facts and forms of facts . Russell held to such a re lation in a number of works prior to 'Th e phi losophy of logical atomism ·. In the latter work he adopted a more Fregean treatment of relations holding that the only constituents (components) of facts were particulars and relations (including qualities as 'monadic' relations) . But . even there , he distinguished between relations and particulars in te rms of the basic asymmetry I have employed , since particulars were terms of relations and the simplest facts 'consist in the possession of a quality by some particular thing' , (in : Logic and Knowledge. p . 198, ital ics added) . What he later rej ected was the grounding of the asymmetry in a basic relation (or tie. or nexus) that was a constituent of facts. One must not confuse the use of 'can ' in ( U 1 ) and (U 2) with its use in the assumption that something can fulfill the conditions stated in ( U 2 ) . One who does so may think the use of (U ., ) begs the question . (U 1 ) and (U 2 ) preclude contradictory . logically unary , and compl;x universals . Questions about such 'things' are irrelevant here . In fact , we can consider

202

Russell's argument in terms of a predicable being wh at is exemplified and a universal being what is exemplified in more than one fact. For related comments on the phrase 'can be exemplified' see my article, 'Nominalism, General Terms , and Predication', in this volume, pp . 133 f.

Logical Form, Existence , and Relational Predication

A nominalist is one who claims that universals do not exist ; a realist claims that there are exemplified universals ; and a platonist claims that universals exist whether or not they are exemplified. As I shall use the term, realism is thus a moderate realism, in that the realist's universals exist only as constituents of singular facts. A singular fact is the exemplification of a monadic universal by a particular or two or more particulars standing in a relation . The platonist's universals need not be constituents of facts in which they are connected to particulars. Due to the influence of Quine, many philosophers now associate realism and platonism with claims about abstract entities, whether these be universals or particulars. In fact, one can viably argue that on Quine's view a realist-platonist is one who holds that abstract particulars exist whereas a nominalist holds that all particulars are concrete. One need not argue, on Quine's view, that only individuals exist, since the notion of ontological commitment is so construed that only individuals can exist. 1 A class, for example, is taken to be an abstract individual to which a concrete individual may be related by the membership relation. Similarly, a property and a particular, in that they are subjects of an exemplification relation , may be thought of as two kinds of particulars, concrete and abstract , that are terms of a relation. Here, I shall consider a property to differ from a particular in virtue of the asymmetry of exemplification. Thus, a property may, and a particular may not, sensibly be said to be exemplified. The notion of possibility involved in the previous claim differs from the notion of logical possibility involved in a claim that something can be 0 even though it is in fact not-0. It is a notion reflected in the categories recognized and not by the logical truths and falsehoods. Thus, while an object cannot be both 0 and not-0, one may sensibly predicate "0 and not-0' of an object. This does not mean that ·0 and not-0' need be taken as standing for a property. One might well hold, first, that only primitive-simple predicates need be taken to stand for properties, and, second, that only exemplified properties need be acknowledged. If one does so, then the two notions of possibility will coincide in that noproperty will be such that it cannot be exemplified since it is contradictory. Given that we recognize properties and particulars, we recognize that one kind is such that it is one term of a basic asymmetrical connection , exemplification. Elements of that kind do,

204

and of course may, exemplify, but they may not be exemplified.Elements of the other kind, by contrast, may be exemplified and may also exemplify other properties. To be a realist is to hold that properties are universals.2 Thi s i s to claim that one and the same property is (or may be) exemplified by more than one particular. An extreme nominalist denies that there are properties at all, whether they be universals or not. . He thereby blocks any argument purporting to prove that some properties must be taken as universals. Such a view is embodied in Quine's use of the notion of ·general term· . Goodman ·s talk of the exemplification of predicates , and Sellars's claims about "semantic assertibility'. This issue I shall not pursue in detail here. Rather, there is another facet of the problem of universals that was focused on by Russell long ago and has recently been revived. Russell recognized properties (concepts , functions). He also held that sentences expressed propositions. He was concerned with the conditions for understanding a proposition. One condition was that a person be acquainted with the constituent entities (of a proposition) that the terms in a sentence, expressing the proposition , stood for. In the case of 'The 1V is 0' we would be dealing with the concept 0 and, depending on Russell's views at a period (1 903 , 1 905, 1 9 12), the denoting concept the 'V or the property (concept) being uniquely 'l!). 3 In the case of 'a is 0' , we would supposedly be acquainted with the object a as well as with the concept 0, assuming that 'a' is a genuine proper name . But, in the case of 'aR 1 b', acquaintance with a, b, and the concept (relation) R 1 does not suffice. 4 For, we understand the proposition expressed as contrasted with the proposition expressed by 'bR i a'. Hence , we must grasp a direction or sense of a relation. Thus, we must also be acquainted with such an additional entity if we are to understand the proposition expressed by 'aR i b' . This, in turn, points to our implicit recognition of the form of a relational proposition as well as that of a monadic proposition. Russell thus felt compelled to recognize logical forms and directions or senses of relations as objects of knowledge by acquaintance.5 . We need recognize such objects and our acquaintance with them as conditions for our understanding propositions. Understanding the propositions a is 0 and a has R I to b requires that we understand the difference between an object's having a property and an object's standing in a relation to an(other) obj ect. This difference i·s a matter of logical form, and grasping it involves knowledge of the forms x-is-F 1 and x-and­ y-are-in R2 . Likewise, understanding the difference between a has R 1 to b and b has R I to a involves understanding the difference between the directions a-has-R 2 -to-b and b-has-R2 -to-a. This requires knowledge of such directions. One can see how Russell arrives at the two notions of form and 205

direction by recalling a familiar device of his. Consider the two sentences 'aR 1 b' and 'bR i a' as related to two propositions [aR 1 b] and [ b R1 a] . Consider next the following lists: 6

I

( 1 ) aR i b (2) aR S (3) xR 1 b ( 4) xR S (5) aR2 b (6) xR2y

II ( 1 ' ) bR 1 a (2' ) bR S (3 ' ) xRi a ( 4' ) xR S (5' ) bR 2a " (6' ) X" R.2y. In ( 1 ) and ( l ' ) we have sentences which stand for the propositions. By replacing constant signs with a variable abstract, we get a series of different signs standing for different "things'. 7 Thus, in (2) , ( 3), (2' ), and (3' ) we have monadic predicate abstracts for having a in R 1 to, being in R I to b, having b in R 1 to, being in R I to a, respectively. Such relational properties need not concern us. In ( 4) and ( 4' ), however, we get tokens of the same pattern, the pattern taken by Russell to stand for the relation (function) R 1 • In (6) and ( 6' ), we also get, by abstracting from all constants, tokens of the same pattern. This pattern is taken as representing the two-term relational form. In (5) and (5' ) we have different patterns indicating the different directions ( or ordered pairs) of a relation . One thus has in the above lists a representation of propositions, relational properties, a relation, a form, and two directions. While recognizing logical forms and directions, Russell held that they were not constituents of propositions, since he thought that taking them as constituents led to a Bradley-type regress. Thus, he held that the form connected the constituents into a complex but, in playing such a role, a form was not a further constituent . In taking such forms not to be

206

constituents of. propositions, while claiming that they were objects of knowledge and that knowledge of them was presupposed by our understanding of propositions, Russell was a platonist with respect to logical forms in a twofold sense. To get at what is involved , it will be useful to consider a modified version of a Fregean ontology. 8 Let us assume that 'aR 1 b' is true and that 'bR 1 a' is false, and let us continue to use ' [ aR l b and ' [ bR l a as signs referring to the propositions expressed by the sentences. Let us further assume that the propositions contain, as constituents , senses of the names 'a' and 'b' as well as the , relational concept associated with the sign 'R l . I will use ' [ a b] ' , and ' [ xR l y as signs for such things.Let us also recognize a fact in virtue of which [aR 1 b] is a true proposition. The fact has, as constituents, the objects a and b and the relation xR S . The relational concept [xR 1 y] correlates to the relation xR S as the sense [a] correlates to the object a. The proposition [aR 1 b] is correlated with the fact aR 1 b in virtue of (1) a correlation betwe en constituents, and (2) a correlation between the form of and ordering in the proposition [aR 1 b] and those of the fact that-aR 1 b. Recognizing propositions and facts, we can then distinguish two sorts of entities associated with realism: concepts and universal properties. Both sorts of things play a predicative role , albeit a different kind, since, in recognizing propositions and facts, we recognize two kinds of predicative connections. Both concepts and properties are also universals in that they are taken to play such a predicative role in different propositions and facts. If one holds that all concepts are predicative constituents of propositions, then he is not a platonist with respect to such concepts. But to deny such a claim, and in that sense be a platonist, would seem to be absurd for one who recognizes concepts and propositions . If one also holds that some properties are not predicative constituents of any facts, then he is a platonist with respect to such properties. Such a view is not prima facie absurd, as is its counterpart with respect to concepts. Concepts and properties have traditionally been fused and confused. One can identify them and still recognize that the predicative connection in propositions is different from the predicative tie in facts, assuming that facts are not taken to be propositions of a certain kind . Even if one identifies concepts with properties, we may still distinguish the two kinds of platonism in terms of concept-properties being required to be constituents of facts, propositions, or both . And , whether one identifies concepts with properties or not, he may or may not take the forms and directions involved in propositions to be the same as those involved in facts. We may then distinguish types of pla ton ism, as opposed to realism, with respect to forms and directions, in terms of such things being independent of facts and propositions. At one time Russell held that properties (identified with concepts) and

r,

r

r

r' '[

207

objects combined into complexes. Such complexes, which were, in effect, propositions, were held to be facts or fictions. 9 Since there were complexes for sentences like 'aR 1 b' and 'bR 1 a' , irrespective of one being true and the other false, he held that there was a complex for every appropriate set of constituents. Facts corresponded to true sentences ; fictions to false ones. He then not only identified properties with concepts, but held that objects (rather than senses) were constituents of propositions. He also, in effect, identified facts with true propositions and fictions withfalse propositions. In the context of the above discussion, Russell's view between 1 9 1 0 and 1 9 1 3 is platonistic in two basic senses. He holds that we must be acquainted with logical objects, such as forms and directions, as a condition for our understanding propositions in which they are involved. In this vein he seems to believe that they exist independently of such propositions (and facts). This belief is supported by his concern that a Bradley-type regress is initiated if such logical objects are taken as constituents of propositions. 1 0 And, his view does imply that forms and directions (as well as concepts) may be involved only in fictions or false propositions, and hence need not be involved in a fact. This feature of his view is independent of how one takes the sense in which logical objects are 'involved in' or 'connected with' propositions and facts. Aside from any distinction between a realist and a platonist, it is clear that insofar as Russell distinguishes xRS (or [ xR 1 y]) from xR 2y, and both of these from aR2 b and bR2 a, he recognizes logical objects. These are like universals in that they characterize facts and propositions into kinds. The recognition of logical objects is problematic. For, whether he takes logical objects to be further constituents of propositions or not, he seems to require that they be connected with the other constituents by a further tie or relation. Otherwise he has not accounted for the way in which such logical objects play a role in giving structure (form and order) to propositions (and facts). Moreover, directions (or ordered pairs) such as aR2 b and bR 2 a seem to be complex objects , since particulars are 'involved'. For, clearly , directions or ordered pairs involve particulars and an ordering or arrangement. One thus accounts for the order in a proposition (and a fact) by the introduction of an entity that is itself ordered. This feature also raises the problem of a Bradley-style regress in connection with the recognition of such objects, apart from any question of their relation to facts and propositions. Russell seeks to avoid one problem by giving the form the role of the exemplification tie. Thus, the form provides the connecting link in the proposition (fact) . But, then , it is difficult to see how he can avoid taking the fo rm as a constituent of the fact (proposition), unless ( 1) this merely means that a proposition ( or fact) is a complex th at contains constituent

208

entities as well as entities that are not constituents, or (2) he takes such a complex to be constituents in a connection and holds that facts (propositions) may not then be analyzed in terms of constituents. He seeks to avoid the problems about directions by holding that a is related to aR 1 b (or [ aR 1 b]) in one way and b is related to that fact (proposition) in another way. Hence he recognizes further facts expressed by a is in D I to [ a R 1 b], b is in D 2 to [aR 1 b], b is in D 1 to [bR 1 a] and

Thus , directions are construed as facts containing propositions (facts) related in a special way to objects. They are, then, clearly complex and j ust as clearly not constituents of their constituent propositions. This means that aR 1 b and bR 1 a differ in virtue of standing in different relations. Such a move is problematic and, for Russell, inconsistent. It clearly runs counter to his classic argument against individuation of particulars by relations in ' On the Relations of Particulars and Universals' of the same period, 1 9 1G-13 . That he consistently held to a line of thought that rej ected the appeal to relations to ground the difference of relata is clear from An Inquiry into Meaning and Truth of 1 940 and Human Knowledge: Its Scope and Limits of 1 948 . In both these later works he introduces absolute monadic spatial properties, as _ constituents of individuals, to ground the difference of such things without appeal to substrata or 'bare' particulars. The complexity of directions (or ordered pairs) also reveals a redundancy in a pattern that recognizes forms and directions. For, if we acknowledge x: R i5' and the directions aR 2 b and bR 2 a, we do not require xR.2y. One can take x:Ri5', as Russell later did, to 'involve' a relational form. To know what xR 1 y is (be acquainted with the relation) is to know that it is a two-term relation. That is why, in Fregean fashion, the relation is represented by the complex sign 'xR 1 y' and not by 'R 1 ' . Likewise, recognizing aR 2 b, or (a;b) , we also have the relational form brought in . Thus, a, b, x:Ri5' , and aR..2b appear to suffice to specify what condition (fact) would be the fact, if existent, that grounds the truth of 'aR 1 b'. In that sense we have a sufficient list of constituents. All we need add is that the fact is the exemplification of x: R i5' by a and b in the order aR 2 b. The Bradley-type 'problems' are obviously connected with the use of 209

something in a double role. Whether one takes the relational form to provide the predicative tie, as Russell does in 'Theory of Knowledge' , or whether one takes the relation (or the direction) to also provide the form for and the tie in the proposition or fact, as Russell did in 1 9 1 8, one avoids a question about the connection of the form to the constituents.By so doing, one also blocks a question about the connection of the relation and the direction to the other constituents. The use of entities in a twofold role to avoid a Bradley-type regress is characteristic of the purported nominalism of Wilfrid Sellars. Sellars holds that the truth condition for 'aR 1 b' is satisfactorily specified in terms of the two related objects, a and b. 1 2 His gambit derives from his understanding of Wittgenstein's remark in the Tractatus. 3 . 1432 Instead of, "The complex sign 'aRb' says that a stands to b in the relation R", we ought to put, "that 'a' stands to 'b' in a certain relation says that aRb". Sellars's position is purportedly nominalistic, by contrast with a realist who would hold that the fact is analyzable into two objects and a relation in an arrangement (exemplification). For Sellars, it suffices to note that aR 1 b is a matter of fact and, in recognizing a merely matter-of-fact relation, as opposed to exemplification, we do not reject nominalism.1 3 Yet, it is clear that the non-existent relation Sellars appeals to plays a threefold role.For, it (1) supplies the difference between aR 1 b and bR 1 a as truth conditions; (2) supplies the difference between aR 1 b and, say, aL 1 b, where 'xLS' , for an explicit realist, would stand for a different relation ; (3) supplies the basis for 'aR 1 b ' , rather than · --aR 1 b', being taken as true, or, to put it another way, it accounts for the fact being more than a class (the class { x:R.5' , (a; b)} , for example). The realist's arrangement performs only the latter function, if he recognizes directions, and only the first and third functions if he does not. In an obvious sense then , Sellars is no less of a realist about relations than one who explicitly acknowledges relations, and no less of a realist about directions than one who explicitly acknowledges such entities. For , the complexity of his appeal to related objects, as opposed to objects exemplifying a relation, allows him to avoid explicitly acknowledging relations and directions as entities. By contrast �ith Russell, Sellars runs together the relation x:R.5' and the direction aR 2 b by holding that ·aR 1 b' is more adequately stated by the use of a spatial relation between tokens of names. Thus, in place of 'aR 1 b' , Sellars employs, in his ideal language Jumbelese, something like

(a) a b. 210

That a token of 'a' occurs to the left of and at a certain distance from a token of "b'represents that x:R 1 y holds between a and b in that order. Sellars thus represents both the relation, x:R i9', and the direction , aR 2 b. Hence, in spite of his declared nominalism , he is clearly a realist about both relations and directions. In fact, he does not even avoid what he takes to be the nemesis of realism - exemplification. For, (a) not only shows what relation and what direction are involved but that the relation purportedly obtains or is exemplified in the direction. That his elaborate attempt to deny that he recognizes such entities fails I will not argue here , as I have argued the case elsewhere . 1 4 Here , it will suffice to note features of Sellars's schema and analysis that correlate with features of an avowed realist's schema. There is also a feature of Sellars's schema that may mislead one into thinking that he does not , on such a pattern, recognize relations. Although we can take (/3) X y in place of ·x:RS' to represent the relation involved, we do not have any sign to generalize from in (/3), as we do in 'x:R S', to yield a sign to represent the relational form that Russell recognized, x:R 2 y. This may lead one to believe that the category of relations need not be represented and, hence, is not recognized. But, recall that on a Russellian-type schema one need not take 'x:R 2 y', as representing anything, even though it is a sign pattern that may be arrived at by abstraction from 'aR 1 b'. Moreover, even though there is nothing corresponding to 'x:R 2y ' in Jumbelese, there is a pattern corresponding to 'x: R 1 y'. And, since Sellars has to distinguish between the situation represented by

ab and that represented by

b a, we can take the former to involve both the ordering represented by ' (a;b) ' and the relation represented by 'x y', whereas the latter encompasses the ordering represented by ' (b;a) ' as well as the relation x y. Thus, just as the explicit realist may think in terms of the use of the pattern aR 1 b as implicitly recognizing x:RS and aR 2 b (or (a;b)), so Sellars's pattern 211

involves corresponding commitments. Once we recognize that , we cannot avoid acknowledging another feature in

a b; that a and b in an order, (a;b ), stand in a relation, x y. Thus, we recognize obj ects, a relation , an ordering, and the exemplification, by the obj ects, of the relation in an order. In fact, Sellars' s n otation, while suggested by 3. 1432 in the Tractatus, can be viewed as carrying a Fregean theme to an extreme. Frege, to avoid a Bradley-type regress, did not recognize a predicative tie in propositions. Instead he took concepts to play a twofold role by supplying the content (and thus the difference between xRS and, say, xLS) and the predicative connection between that content and the individual senses in the proposition. This is one aspect of his holding that concepts are 'unsaturated' and his representation of concepts by expressions like 'xR1 y', which give the form of the sentences in which such signs become 'completed' . The awkwardness of such a representation is that 'xR1 y' is a complex sign in which 'R1 ' occurs. The use of the attached variables 'x' and 'y' in 'xR1 y' can then be understood to indicate the difference between monadic, two-term, t hree-term, etc. , predicates (and concepts) rather than the incompleteness of the predicate (concept). In short, predicates like 'Gx' and · xR1 y' represent concepts as well as the form of the propositions the concepts enter into. That is why, with 'xRi5' ' , 'xR 2y' is redundant on a list of the constituents of a fact (or proposition). On Sellars' s pattern, we cannot point to a distinctive part of the concept sign (such as 'R i ' in 'xRS' ) , as the concept sign has become a relation obtaining between names or, in the monadic case, a characteristic of a name. Thus, Sellars removes an awkwardness of a standard Fregean-style notation, and , hence, he dramatically represents the dependence of relations on obj ects. But this does not avoid relations, though it may well be taken to emphasize the moderate realism implicit in Sellars' s purported nominalism. In his most recent book Sellars presents an argument for nominalism that points to his hidden realism. He writes, Thus. one who is simply struck by the fact that we could use larger than b will be tempted to look for some aspect of (6)

�

which is doing the job done in (5 )

2 12

a larger than b

T to say that a is

by " larger than" , for example "the fact that •a' is above 'b'" or "'a"s being above 'b "' . It is absolutely crucial to appreciate that nothing in ( 6) , or about ( 6) , is doing the j ob done in ( 5 ) by ''larger than " . Many philosophers have stared this point in the face and missed it , thus failing to grasp its significance . Obviously the fact that •a' is above 'b' is essential to the semantical role (6) is playing. But that fact does not do the j ob done by "larger than " . Rather it does the j ob done in the case of (5 ) by the fact that ' a' and 'b' have a "larger than" between the m . Let me repeat : Nothing in (6) , or about (6) , is doing the job done by "larger than " . 15

It should be clear from the emphasized phrases that Sellars is taking the fact that a Jumbelese-style schema is constructible to be an argument for nominalism. 1 6 That aside, he is wrong in his claim. This is seen where we use 'xL 1 y' as the realist' s·· predicate''. Thus, ' x ' plays the role of 'xL 1 y' and , j ust as clearly, the space between 'a' and 'b'\akes the place of ' L 1 ' as the occurence of 'a' over 'b ' , rather than 'b' over 'a ' , replaces the linear order in 'aL 1 b' to represent the order in the fact. What Sellars is entitled to say is that no sign in ( 6) is doing the j ob done by ''larger than" . What he can add is that if you erase the tokens of 'a ' and 'b' in (6), the space between them disappears while, in (5), the token of ' larger than' would also have to be erased . This points to a difference between a token of a phrase and "a space'' , not to an argument for nominalism. Sellars is making use of two basic themes of moderate realism, as opposed to platonism. One only find s relations as obtaining between terms. Hence, without the terms we do not have the relation. And , since they are exemplified by terms, relations are not constituents of facts in the way the terms of the relations are. These are theses of realism that Sellars seeks to transform, by the formulas of Jumbelese, into nominalism. Sellars seeks to convince his readers that one does avoid recognizing properties by pointing out that ordinary statements, such as 'This is green', in ord inary contexts do not state that a relation obtains between an obj ect and a property. Since they do not do so, the realist construes a pattern of a purported perspicuous language, 'Ga' , in terms of a property being exemplified by an obj ect. Therefore, as Sellars sees it, the realist cannot claim that the pattern of the perspicuous language is a transcription of an ordinary statement. 1 7 In the case of the true sentence 'This is green' we'acknowledge a green obj ect and in the case of the truth ' This is to the right of that' we have spatially related obj ects. In neither case do we have additional objects (entities, terms): the color and the relation. B ut this is not an argument. It is a declaration that is equivalent to the familiar fiat that as long as one uses predicates predicatively no ontological commitments are made. The fact that one shows the dependence of attributes on obj ects by employing the spatial distances and j uxtapositions of Jumbelese, in place of normal predicates, 21 3

emphasizes, but does not support, the claim. One sees Sellar's gambit at work, simply and clearly, in an analogy he uses. It seems to me that the necessary equivalence but non-synonymy of a exemplifies triangularity with a is triangular is analogous to the necessary equivalence but non-synonymy of That a is triangular is true with a is triangular. That the analogy is more than a mere analogy is suggested by the fact that instead of saying that a exemplifies triangularity , we might with equal propriety say that triangularity is true of a, or holds of a. Now if a exemplifies triangularity triangularity is true of a triangularity holds of a are to be elucidated in terms of That a is triangular is true then exemplification is no more presen t in the world of fact in that narrow sense which tractarians like Professor Bergmann and myself find illuminating, than is meaning , or truth , and for the same reason . 1 8

The point is that the role of ' is true' is seen in the necessary eq uival ence of 'p is true' and 'p' . And this reveals that "is true' does not stand for a property of sentences or propositions. One must see the analogous connection of 'a exemplifies G(-ness) ' and •aisG' . Doing so, one does not take 'exemplifies' to stand for a tie or connection between a thing and a property. Hence, one need not recognize the property as a term of such a connection. N ot being a term, a property is not an entity. It is as if Sellars put s Bradl ey' s regress in the service of nominalism. The realist must recognize exemplification as a connection between obj ects and attribut es. But, we need not recognize exemplification as a connection, since "exempl ifies' functions like 'is true' . Hence, we may rej ect realism. Moreover, in place of "a exemplifies G(ness) ' we may use '" G' is true of a, " which is to be understood in terms of " ' Ga' is true" . Thus, the analogy is " more than a mere analogy" . By using "is true' we dispense with 'exemplifies' and, by the necessary eq uivalence of ' p is true' and ' p', we

2 14

purge 'is true' of ontological commitment . Thus, 'Ga' is true not in virtue of an obj ect exemplifying a property but in that a is a green obj ect ! In this way the nominalist dodges acknowledging the color of the obj ect as an entity. Such a line of reasoning avoids rather than resolves the problem of universals. For , the issue is about whethe r predicates , normal predicates or Jumbelese predicates , play a representative role. The realist makes a simple point. Since such predicative features are basic features of a perspicuous schema which purportedly allows us to represent the neutral facts we all recognize , Sellars must furnish an argument to show that they are exempt from being taken as representatives of something . To say that they represe nt objects, such as a, as qualified objects is to offer , as an analysis , what is to be analyzed . That Sellars elaborates his account by a sketch of the socio-psychological context in which we learn to attribute predicates to appropriate obj ects does nothing to resolve questions about the analysis of the truth conditions for se ntences like 'Ga' . The sentence is true , for him, simply because a is green . That , of course , is true enough . What is required , however , is that one explicitly state what a truth condition is (a fact , a complex obj ect, a coherent set of propositions , etc . ) . 1 9 Sellars does offer complex obj ects (objects as green , square, etc.) as such conditions. 20 But, he refuses to acknowledge that such qualified obj ects do not differ from the realist's facts . He also refuses to see that the re is a basic difference between 'is true', as a semantical predicate , and "exemplifies', as a term for an ontological tie . This difference is easily spelled out. The re are two problems connected with the notion of truth that have concerned philosophers. One is to give an account or theory or analysis of truth. This involves specifying what a truth condition is (a fact, an obj ect , a set of propositions, the acceptance of a belief, etc.) , characterizing how such a condition is construed (as a complex , a simple , a class) , and showing that such a condition so construed suffices as a ground of truth . A second problem concerns the introduction of a term into a perspicuous schema that will transcribe 'is true' and allow one to state that something is true without paradoxical consequences. Since both tasks are described as giving definitions of truth, they are often confused. Thus, if one introduces a predicate, along lines made familiar by Tarski, that satisfies the so-called convention T, he may thereby think he has proposed a theory of truth . Alternatively, if one offers a theory, such as a correspondence theory that recognizes both propositions and facts, one introduces a truth predicate for atomic propositions in terms of something like (I)

p is true = (3F) p ReF . F ,

2 15

where 'p' is a variable ranging over atomic propositions, 'F' is a variable ranging over atomic facts, and 'ReF. ' represents a relation between a proposition and a fact . Then, since (II)

p

= (3F) p ReF. F

must also be assumed, if we are to derive the required (III)

p is true

= p,

some conclude that such a theory of truth has no content . To say p corresponds to a fact is to say no more than that p is true, or, even, that p. But, (I) and (II) must be understood in the context of the relevant issues. (I) represents a version of a correspondence theory when we understand how 'p', 'F', and 'ReF. ' are taken. What is trivial is the introduction of a truth predicate, given that we require (III) to hold for atomic propositions, and not the correspondence theory that is reflected in, rather than stated by, (I). The statement of the relevant version of the correspondence theory requires the commentary about (I). That, in turn, requires the use of a background context involving references to propositions, facts, and connections between such entities. It is in that context that one also provides a discussion of the notion of a fact as a truth condition. In such a discussion, the question of an exemplification nexus arises. Thus, it is pointless to dwell on (IV)

a exemplifies G

as being necessarily equivalent to (V)

Ga

and

(VI)

· Ga' is true.

To do so is to make two mistakes. First, one mistakenly thinks that ' (3F) p ReF . F' is empty in view of (I), (II) , and (111 ) . Second, one fails to see that (IV) is involved in the explication of the notion of a fact. Just as some see (I) as trivial in that (II) is acknowledged, so Sellars tries to suggest that (IV) is trivial given its acknowledged equivalence (in some sense) to (V) and (VI). 2 1 There is an ironical note to this as well as an echo of the purported "paradox of analysis". For Sellars has criticized Davidson for making the very move Sellars himself makes, though he would not

216

acknowledge this since he sees another task to be achieved regarding 'is true' that is not achieved by merely satisfying the condition of convention T. One must elucidate the rule-governed activity of the use of tokens of "is true' in the social context.22 But this diverts us from the issue at hand. 23 It also reveals an implicit appeal to an "'ordinary language'' motif in Sellars 's argument. 'Exemplifies' has no ordinary language role, so there is nothing to elucidate. Thus , in view of (111)-(VI), we can confine ourselves to setting forth the rule-governed behavior relevant to the utterance of tokens of 'is true· in the '"causal order''. We noted earlier that Sellars explicitly appeals to ordinary usage when he argues that the realist fails to transcribe ordinary sentences , like 'a is green', by using ·exemplifies· in holding that a particular exemplifies a property. Here , too , we have an implicit appeal to the "paradox of analysis" . The paradox , recall, is that one cannot offer a statement as an analysis , since it will either trivially repeat the statement to be analyzed or state something different , and hence not be an analysis. In Sellars's hands, the paradox dispenses with the use of 'exemplifies' to render 'a is green'. This, in turn , dispenses with the recognition of a tie of exemplification. And, that , as we saw earlier, is taken by Sellars to rid us of the property green. Aside from any purported paradox of analysis , Sellars appears to argue that the realist's transcription of 'a is green' is not perspicuous since the realist reads ' Ga' in terms of the exemplification of G by a. But , what is it to furnish a perspicuous transcription? How can one defend nominalism by pointing to either the philosophical neutrality of ordinary usage or the grammatical fact that in 'This is green' neither 'green' nor 'is green' is used as a subject term? For , surely, the realist may acknowledge such claims. The issue is about whether one need appeal to universal properties and a tie in order to account for one predicate truly applying to two obj ects. Sellars allows an acceptable response to be the reiteration that the predicate truly applies because the objects are green. This is to claim that so long as predicates are used predicatively, no ontological commitments are made by such use . This Quine an theme explains Sellars's claim that 'exemplifies' is to be understood in terms of 'is true'. For , 'is true' is an utterly trivial addition to a perspicuous schema, since there is a clear sense in which the truth conditions for an atomic sentence s and a sentence asserting that s is true are the same . But , the use of ·exemplifies' , by a realist , signifies the latter's recognition of properties, facts , and obj ects. Thus, there is a critical difference between a use of the ordinary sentence 'This is green ' and a purported analysis mentioning a tie between an object and a property. The two are ''necessarily equivalent'' in that one is offered as part of a philosophical analysis in response to questions about the other. But this does not warrant Sellars's claim that 'exemplifies' is like 'is true', if we are thinking

2 17

of the sense in which 'is true' is a trivial addition to a schema. Note too, in the above cited passage, how Sellars slides from the claim of an analogy involving 'exemplifies' and 'is true' to the claim that 'exemplifies' is to be elucidated in terms of 'is true'. This, too, emphasizes his repetition of Quine's theme. 'a exemplifies green' is elucidated in terms of '"green' is true of a" ; the latter is elucidated in term s of "'a is green' is true" and that is ultimately elucidated in terms of (1) there being a green object and (2) the appropriateness of calling green objects green, in view of the roles 'green' and 'is true' have. In speaking of a green object and the predicate 'green' we do not speak of the color green as a term of a relation or tie. Hence, we escape from realism. This "defense" of nominalism thus forces Sellars to claim that every statement using a predicate as a subject term may be transcribed into a statement where no predicate is so used. 24 The success or failure of nominalism thus depends, for Sellars, on the success of such a program of transcription. Here, I do not wish to explore the pros and cons of such a program. I will merely point to a blatant inconsistency in Sellars's views. The realist is held not to transcribe 'a is green' as 'Ga' because he reads the latter in terms of the exemplification of a property by a particular . But, Sellars allows himself to transcribe 'Green is a color' in terms of " All green things are colored' or '"Anything of which a token of 'green' is true is a thing of which a token of 'is colored' is true" or some such statement. The implicit rule for the transcription of ordinary statements is clear. The nominalist , but not the realist , may depart from the literal transcription of ordinary statements. The realist may not Hparse" the pattern "Ga' in terms of 'a exemplifying green' since 'green' is then treated as a subject term , but it is not such a term in 'a is green'. The nominalist may replace " Green is a color' by a transcription not using a token of 'green' as a subject term, since that supports Sellars' s version of nominalism. In a recent work , Sel lars seeks to bolster his case against realism with a further argument. But , it might be argued , the platonist could capture the conceptual tie between exemplification and truth with the claim that truth is to be defined in terms of exemplification , thus That Fa is true

=

df a exemplifies F.

But this would be equally absurd , for it would be synonymous with That a exemplifies F is true

=

df a exemplifies F

which , as a definition ( though not as a necessary equivalence ) , is incoherent . 25

218

The argument is based on the assumption that 'Fa' and "a exemplifies F' ,, are " identical in sense and differ only in that 'Fa' does not use "an auxiliary expression'' . The assumption clearly confronts the realist with the paradox of analysis, since it involves the claim that the ordinary ,, statement to be analyzed is '' identical in sense with the statement reflecting a philosophical analysis. Thus, Sellars's argument fails on that ground alone. But, that aside, Sellars overlooks the simple point that what is "d efined" in an expression like p is true is the phrase ' is true' . Thus, whether we deal with a sentence with the auxiliary expression 'exemplifies· or not, there is nothing incoherent about holding that

'a exemplifies F' is true = df a exemplifies F. Sellars appears to be raising, knowingly or not, a version of Frege' s, and Bradley' s, argument against purported analyses of 'is true'. The realist holds that a monadic atomic statement 'Fa' is true on the condition that a exemplifies F. Sellars then asks if the realist's analysis applies to the sentence 'a exemplifies F'. And, it does, since the truth condition for that sentence is the same as the truth condition for 'Fa', that a exemplify F, and not, a la Bradley, that a and F exemplify exemplification. By (illicitly) identifying the sense of 'a exemplifies F' with the sense of 'Fa' , Sellars substitutes the former for the latter in ( S)

· Fa" is true = df a exemplifies F

and obtains (S 1 )

'a exemplifies F' is true = d fa exemplifies F.

Writing (S 1 ), as he does, as (SJ

that a exemplifies F is true = d f a exemplifies F,

it appears as if one uses 'exemplifies' in a circular way. 26 But that does not happen. As we see, by looking at (S1 ) , 'exemplifies' is only being used in the specification of the truth condition for 'a exemplifies F' . There is no more a problem of circularity with 'exemplifies' in (S 1 ) than there is with 'F' in (S) . But, if one permits (S1 ) , he may raise a problem, if he holds that a Bradley regress is viable in such a case. Then, if we admit 'exemplifies' 21 9

as a relational predicate, the condition for 'a exemplifies F' appears to be " (a ;F) exemplifies exemplification', and this purportedly initiates a Bradley-type regress. But it really does not . One need merely t ake the fact that is the truth condition for '(a ;F) exemplifies exemplification' to be that a exemplifies F. Or, what amounts to the same thing, one can take the appropriate sentence to be (S1 ) and not (S3 )

'a exemplifies F' is true =

df

(a ;F) exemplifies exemplification.

This would underline the insistence that exemplification is a tie and is not a relation among relations. 27 This point is made more emphatic by representing exemplification in a different way than one represents ordinary relations: by not allowing a term for such a tie or, at least, not allowing such a term to occupy the grammatical place of a relational predicate. Sellars, at best , thus merely reraises Bradley' s old objection and does not establish the incoherence of the realist's pattern. M oreover, not only does he mistakenly take the realist to be involved in a vicious circle, through not realizing that "is true' is being explicated , not 'exemplifies', and that 'exemplifies' is not used in the left sides of (S 1 ) and (S2), he also overlooks a further crucial point. One must recognize a different exemplification nexus or tie for monadic, dyadic, triadic, etc. , attributes. Thus, the pattern for " is true' is closer to "Ga' is true = dfa exemplifies 1 G • R1 (a, b)' is true = df ( a; b) exemplifies2 R 1 etc. Hence, if one at tempts to generate a regress, one should use •a exemplifies1 F is true = df (a �F) exem plifies2 exe mplification 1 , and all semblance of circularity disappears. (One should also keep in mind that since "exemplifie s 1 ' occurs inside the quotes, on the left-hand side of the biconditional, there is no circularit y on that ground alone. It appears that Sellars's rend ition, through the use of a "'that" clause , helps to confuse matters. But, even with such clauses, we still may distinguish use and mention.) The reali st, either by recognizing that the fact - a exem plifying F - is the truth cond ition for a whole series of sentences a exemplifies 1 F (a ;F) exemplifies2 exem plification 1 etc. 220

or by holding that such a series is illegitimate can take the purported B radley regress to be as harmless as the series Sellars acknowledges to be harmless a is F 'a is F' is true , '"'a is F' is true, is true etc. One need not , as Sellars does, seek to avoid a B radley-type regress by , taking the ""deep structure , of ·a exemplifies F-ness' to be ' that-Fa is true'. Gustav B ergmann once criticized the Fregean theme regarding the incompleteness of concepts by suggesting that signs for particulars may also be taken as incomplete. 28 Thus a name could be of the form · 0a' , rather than simply •a'. This would show that neither names nor predicates, by themselves, are used to assert anything and that they require being combined with signs of the other kind to form a sentence. It would also put names and predicates on a more equal footing, as befits a realistic position, since the Fregean-style notation supposedly emphasizes the dependence of concepts. As we have noted, Sellars's notation stresses that dependence even more emphatically. But B ergmann's claim overlooks a simple point. One would require a number of signs for each name to indicate the role the name may play in different sentence forms:

Alternatively, one could understand •0a' to cover all such cases. B ut, one need not do anything like that in the case of predicates used predicatively. What this shows is that the basic difference being depicted by the Fregean-style notation is that particulars ( or " senses" ) may enter into logically different kinds of facts ( or propositions) as subj ects, whereas properties and relations (concepts) may enter into only one kind as attributes (concepts). 29 This, of course, is a question about the form of facts (propositions) and not one concerning the dependence of attributes ( concepts) on particulars. What it represents is that particulars are not of logically different kinds, whereas attributes are monadic, dyadic, and so on. Since there are different kinds of attributes, in this sense, there are different kinds of facts (propositions). Hence, there is a point to taking the attribute ( concept) to "carry" the form of the fact (proposition) into which it enters as a constituent. In this sense, there is also a point to a Fregean-style notation, and even to Sellars' s extreme version of it.

22 1

If one holds that we need not appeal to either a fact or a relation to account for the truth of 'aR i b' , then questions about forms and directions would seem not to arise. In a recent book Michael Loux reiterates Quine's appeal to "is true of" and "satisfaction" to avoid attributes. 30 Thus, to account for the truth of 'aR 1 b' one need only note that a and b satisfy the predicate 'xR 1 y' (or that the predicate is true ofa and b ) . Oddly , if Sellars were to appeal to such a move, that would amount to holding that 'x y' is satisfied by a and b. And that would mean that one recognizes, explicitly, relational predicates in Jumbelese. 3 1 What is interesting about Sellars's schema is that using ' a b ' , as opposed to 'aR i b' , it appears as if we make use of a relation obtaining between the names to represent a relation obtaining between the objects. But, when we recognize the pattern with variables, as in 'x y' , we see that Sellars uses both the relation between the subj ect terms, as in 'a b', and the spatial distance between the subject signs. The predicate 'x y' is more like the predicate 'xR 1 y' than the sentence 'a b' is like the sentence 'aR 1 b' , and it is by concentrating on the latter pair that one may be tempted to overlook Sellars's use of a predicate term, albeit a strange one - the distance between subject terms. Insofar as one uses 'xR 1 y' as a predicate term, rather than " R I ' , one cannot have 'R i ' on a list of signs without being flanked by an 'x' and a 'y'. Likewise, one cannot have the space between the variables without letter tokens. But, then , 'x y' is as much a predicate sign as is 'xR 1 y'. To say this is not to claim that spatial juxtaposition represents a relation and hence that the spatial j uxtaposition of a and b functions as a predicate in 'a b' . This would let Sellars reply that the spatial juxtaposition of a and b is the sentence, not a constituent of the sentence. The point , to repeat, is that the space between 'x' and 'y' in 'x y' plays the role of ' R I ' in 'xR 1 y' . The only difference is that one can remove the letters •x' and "y' from 'xR i y' and be left with 'R 1 ' ; one cannot remove them from 'xy' and be left with the space between them. But, 'R 1 ' is not a term of a proper Fregean schema. This is why Sellars's schema is a Fregean schema that is both extreme and consistent� it is not a schema that perspicuously does without predicates. Whether we take 'x y' as a predicate sign or not, one clearly has to recognize an order among objects in the case of relational predicates. Thus, whether one employed 'x y' or "xR 1 y' , to employ the notion of satisfaction, in nominalistic fashion , would require one to recognize ordered pairs or sequences. Ordered pairs (or directions), and sequences generally, are not objects in the sense in which a and b are objects. Moreover , in the case of ordered pairs (as opposed to pairs) , the question of order or structure reappears. As we noted earlier , an ordered pair ( or a direction) seems to be a complex consisting of objects in an order or structure. Paradoxically, then , the nominalist's use of satisfaction appears to force the recognition of entities other than particulars -pairs -

222

and the order obtaining between particulars in an ordered pair. One might seek to analyze the ·order' in a standard way, along the lines of the so-called Wiener-Kuratowski procedure. Thus, an ordered pair is construed as a class of classes so that ( a ; b) is taken as the class { {a} , { a, b} } , or, since a and b are not themselves classes, but particulars , we could employ the •simpler' class {a, {b} } for ( a; b). 32 If one takes the pair as a class, one then takes the predicate, 'xR 1 y' or 'x y', to be satisfied by such an entity. Relational predication is then construed along the lines of monadic predication except that the kind of thing taken as the subj ect is different : a pair or class as opposed to an obj ect . Using classes , such as { a, {b} } or { {a}, { a,b} }, apparently allows one to dispense with an appeal to order, but, ironically, it introduces 'abstract' entities, i.e. , classes, and classes of classes at that, in the analysis of predicates applying to particulars. This is hardly in the nominalist's style. On the other hand , if ordered pairs are taken as basic, one introduces not only a new kind of entity but one that is complex in the sense in which the realist's facts are complex : the ordered pair (a;b) involves two objects in an order . Bergmann, who to Sellars is the arch realist on the contemporary scene , has recently made a decidedly nominalistic move with respect to the problem of relational order. 33 He seeks to construe the order of a relational fact along the lines of the set-theoretical construal of an ordered pair as a class. Bergmann introduces a new entity , a diad. Given any two entities, there is , 'eo ipso' , a third, the diad of which they are the sole constituents. Thus , given a and b, we have (a, b) , their diad. And, as in the case of sets , ' ( a, b) = ( b, a)' holds. One can distinguish the facts aR 1 b and bR 1 a in terms of the former involving a and (a, b) as constituents whereas the latter involves b and (a, b). Thus, one can analyze the fact aR 1 b in terms of a, (a, b) , x R S , and exemplification without appealing to an entity, such as (a;b) or aR 2 b, to account for the order. Bergmann thus, in nominalistic fashion, dispenses with an entity that is neither a particular nor reducible to particulars by appealing to a diad of particulars. Also, on his pattern, one need not appeal to the form xR 2y, since the diad (a, (a, b)) is taken as the subject of the relation x R S in the fact aR 1 b. That shows that xRS is a two-term relation on Bergmann's pattern. In the case of a monadic fact, a diad would not be the subject, and in the case of a three-term relation we would have a 'more complex' diad as the subject, i.e. , something like ( (a, (a, b)), c) or (c, (c, (a, (a, b)))). If Bergmann has succesfully analyzed order in relational facts, his gambit can be appropriated by the nominalist to avoid abstract entities, classes or pairs. Bergmann's diads, being complex particulars without an ordering or structure , should be more palatable to a nominalist. But even armed with Bergmann's diads , a nominalist must still propound the absurdity that a relation between a sign and a thing , a diad or an object, 223

constitutes the ground of truth of a sentence. And, he must face the obvious question about appealing to satisfaction as a relation. For, to stick with this pattern he must hold that (I) 'xR 1 y' is satisfied by (a;b) is to be construed as (II) '...is satisfied by ... ' is satisfied by ('xR 1 y' ; (a;b)). And this, of course, generates another version of a Bradley-type regress. For, on this nominalist's gambit, it is a relation between a sign and a pair that constitutes the condition of truth for'aR1 b' , (I), (II), and so on ad infinitum. Thus, (I) stands for a fact which we must analyze along realistic lines or which must be reconstrued along nominalistic lines.But to do the latter is to introduce (II). The nominalistic 'solution' thus invites the kind of question it purports to answer.Moreover, nominalists cannot invoke the kind of response that realists have employed in the case of a corresponding question about exemplification: that exemplification is a special kind of relation, a tie or nexus, that does not require a further connection to connect it to what it connects. To do so would be to recognize satisfaction as a special kind of entity. The version of nominalism we have considered thus falls victim to a vicious regress of the kind Bradley and Frege invoked for different purposes. My main concern in this paper, however, is not with the use of Bergmann 's analysis of order by a nominalist, but with the cogency of that analysis. Although suggestive and innovative, it is not viable . The basic problem with Bergmann's analysis of order has nothing essentially to do with his introduction of a new entity , the diad. Whatever problems and ad hoc features one may find in the introduction of such a complex but unstructured entity , there is a more basic problem with his analysis.It is a problem that would be present if he were to employ classes as constituents of facts and to directly make use of the Wiener-Kura­ towski procedure rather than parody it by the appeal to diads.Thus, I shall consider the issue in terms of classes. Since we deal with elements that are not classes, we may use the classes { a, { b} } and { b, { a} } to construe the ordered pairs (a;b) and (b;a), respectively . Let us then take R I to be a property of the class { a, { b} } , as Bergmann takes it to be a property of the diad (a, (a, b)). The question at issue is whether such enities, {a, {b} } or (a, (a, b)) , allow us to analyze order in facts and, hence, dispense with directions, forms, ordered pairs, and so on. We should first note that one chooses which class, { a, { b} } or { b, {a} } , is the constituent of aR 1 b. In fact, there are any number of other 224

choices one could make; { a, { a, b} } , { {a} , { a, b} } , and so on. This choice is really complex. Given that an ordered pair is taken as a constituent of a fact, one chooses which class is to be taken as such a pair. And, given such a choice, one makes use of an ordering imposed on the class. It is not the class alone that one employs in the analysi s, but the elements of the class i n an ordering. In a sense one implicitly treats the classes (or diads) chosen as ordered entities. It is true, of course, that the class { a, {b} } is not a structured or ordered entity since

{ a, { b } } = { { b } , a } is true, whereas, where a =I=- b,

(a;b)

=I=-

(b;a) .

But, in choosing to take (a;b) as {a, {b} } , in the analysis of the fact aR 1 b, one takes the element that is the unit class in { a, { b} } to be the second consti tuent of the class. That is, one takes the member of the class that is a unit class to be the correlate of the second term in the sentence ·aR 1 b'. The appeal to order is thus concealed in the interpretation rules, rather than inhering in the linear order of the expression. The Wiener­ Kuratowski procedure enables one to map sets onto sentences, { a, RP { b} } onto 'aR 1 b' rather than onto 'bR 1 a' ; but , in so using that procedure, there is an implicit appeal to order. Thus, the use of the procedure in a purported analysis of order in facts is question begging. This is not to say that the procedure is illici tly used in set-theoretical contexts. For the only requirement imposed for such use is that one construe (a;b) and (c;d) such that ( a ;b )

= (c;d) iff (( a = c) & (b = d))

holds. The procedure satisfies that requirement . Since we make use of an unordered entity, a class, it appears as if we have analyzed the order of a fact. But, all we have done is introduce a notation that does not appeal to an ordering of signs. To see this, let us replace 'aR 1 b' by (1 ) ' { a, { b} } E R 1 ' , taking R1 to be a class of an appropriate 'type' . We would also have (2) ' { b, {a} } E R1 ' for 'bR 1 a'. We may then note that (1 ) and (1 ' ) ' { { b }, a} E R1 ' state exactly the same condition, as d o (2) and (2' ) ' { {a} , b } E R/ . Thus, the order within the subj ect sign is irrelevant . We could also write our sentences as 'aR1 { b} ' and '{a} R 1 b' , under­ standing that the linear ordering of subject signs makes no difference because 'aR 1 { b} ' and '{ b} R 1 a' state the same thing, that-aR1 b. We t hen replace the appeal to the ordering of signs by the introduction of 225

class signs in our 'sentences' and the recognition of classes as constituents of facts. But, the ordering of particulars has not been eliminated, for we use the set-theoretical symbols to represent an ordering in a fact instead of using the ordering of the signs in a sentence to represent an ordering in a fact. Bergmann's gambit is thus startlingly like Sellars's attempt to dispense with properties and relations by means of a novel linguistic representation. Both believe that by representing something in a special way they do not represent it at all . Bergmann 's analysis also fails in that, first, he merely manages to represent order by a symbolism that does not involve ordered sign complexes, and, second, he must introduce an additional kind of entity, a diad, and declare that it is a complex but not structured. He seeks to make this latter claim palatable by holding that since any two things 'eo ipso' form a diad , the existence of a diad of two entities is a matter of logic and not of fact. Thus, given an object a and a property G, irrespective of whether a has G or even whether G is a property that may sensibly be attributed to a, there is a diad with a and G as constituents . However, it is obvious that Bergmann's diads are merely classes renamed. Given any two things, there is 'eo ipso' (or by a kind of comprehension rule) their pair-class. In effect, Bergmann thus takes cl asses to be constituents of atomic facts. To solve the problem of order at such a price is pointless. But , then, how is one to resolve the issue? If we introduce directions or ordered pairs, we recognize that they are complexes in a threefold way: there are constituents of such complexes, the ordered terms ; there is the ordering of the constituents ; there is the two-term form that is involved. We must then hold that questions comparable to those raised about the order of facts may not be raised about directions or ordered pairs and, hence, that Bradley-type regresses may not be generated. Alternatively, we may take a hint from Frege's decisive insight, which was explicitly adopted by Russell and Wittgenstein, and, ultimately, if not explicitly, by Sellars. Frege took it to be pointless to introduce a predicative connection as an additional constituent of a proposition. Thus, he let concepts play a twofold role. He was mistaken to think that no connection was then involved, since he let the concept provide the connection. (In a similar way Sellars was later to be mistaken in thinking that he avoided relations and exemplification by appealing to related objects . ) Nevertheless, there is the decisive insight that in dealing with the analysis of structured complexes something must play a complex role. In the case of monadic facts (or propositions) exemplification connects a term and an attribute, and as an asymmetrical connection, provides the form of the facts. In the case of relational facts, one can take the order (as a direction or ordered pair) as another constituent of such facts . One must then hold that exemplification connects the objects, the order, and the relation in the

226

fact. Exemplification is then taken to involve a further orde ring in that it connects the objects in the order into the relation. Thus, even with a direction as a constituent, exemplification takes on a dual role. However, one may hold that the objects do not exemplify the relation. Rather, it is the ordered pair that exemplifies the relation. On this view relation al exemplification is construed on the order of monadic exemplificat ion with ordered pairs as subject terms. B ut, ordered pairs are themselves complexes of the objects in an ordering. Hence, the objects as well as the ordering of the objects are, ultimately, constituents of such a fact. Thus, instead of taking exemplification to connect the objects, the relation , and the direction, one takes it to connect the relation to the pair, while taking the latte r to involve a connection between the objects. Alternatively, one can take the order among the terms as yet another aspect of the " predicative' tie, and n ot as a further constituent. This means that one may hold that facts (or propositions) are not analyzable in a sen se in which they would be analyzable if one acknowledged ordered pairs or directions. For, on such a view there is no constituent in virtue of which the fact that-aR1 b would differ from the fact that-bR 1 a. The facts are just different structural connections amon g the same constituents. Exemplification, on such a gambit , supplies the order , structure, and connection for constituents. A variant of the gambit would involve holding that the connection in aR 1 b would differ from that in bR 1 a. One then groun ds the difference in order in a different exemplification nexus. This would permit on e to hold that the facts differ in virtue of a different connection or nexus and not in virtue of a constituent direction. On either gambit on e recognizes different exemplification ties for different kinds of facts, in the sen se of monadic, dyadic, etc. , and , on the latter view, one would also acknowledge two dyadic ties, six triadic ties, an d so on. The recognition of different ties for different kinds of facts is not merely an ad hoc response to the questions regarding relational and monadic predication. We can see that when we note that B ergmann 's analysis of order must also incorporate several exemplification connections. He holds that there is only on e such connection, which holds between a monadic property or relation and something else. Thus, exemplification is, in a way, always a two-term conn ection for Bergmann. In the monadic case the connection would hold between a property and an object , in the case of a dyadic relation it would hold between a relation and a diad , in the case of a triadic relation it would hold between such a relation and a more complex diad, an d so on. I said 'in a way' above, because one can think of exemplification, for B ergmann, as always operating on one term, a diad . Such a diad would consist of a monadic property and an object , or of a two-term relation and a diad , such as (a, (a, b ) ) , and so on . But all this is misleading. For, both terms of such di ads are logically different in all such 227

cases. As one term we have either a monadic, a dyadic, etc. , attribute, whereas, as the other term, we have an object or an increasingly complex diad. These differences obviously point to another way of recognizing different exemplification connections. Insofar as one recognizes an exemplification connection in logically different kinds of facts, one recognizes a different kind of connection for each kind of fact. To hold that there is one connection that plays logically different roles or is 'multigrade' is to use different words, not to acknowledge fewer 'things' . There is yet another alternative. One can follow Frege and take the monadic and relational attributes to provide the connection, the form, and the order. This will do so long as one does not thereby think that he has not acknowledged that facts (propositions) i nvolve structure and order. All one does is let the attributes take on the logical roles of the predicative connection; one does not dispense with such roles. Thus, however we go about it, the resolution of the problem of relational predication requires that we recognize two things. First , complexes like facts (and propositions) are not analyzable since they are connections, i ncluding ordered connections, among constituents, though they are not connections in the sense of being connecting ties or relations. And , second, some constituent(s) will play more than one role in a fact or proposition. As to how one may choose between the various alternatives just outlined, I do not know. There are reasons for and against each. But none of these alternatives incorporate the delusion that order has been eliminated or the double delusion that order and relations have both been eliminated .

Notes 1

2

3

4

5

On this point see my 'Nominalism, General Terms, and Predication', in this volume, pp . 1 33 f. For arguments for the existence of properties and universals , so understood, see Russell's classic paper 'On the Relation of Universals and Particulars· , Proceedings of the Aristotelian Society ( 1 9 1 1- 1 2 ), pp. 1-24 , and my 'Russell's Proof of Realism Reproved', in this volume , pp. 196 f. For a discussion of Russell 's views on denoting and reference , see my Th ought, Fact, and Reference: the Origins and Ontology of Logical Atomism (Minneapolis: 1 978), chaps. 7 and 8. I wi ll use 'R2 ' as a variable ranging over two-term relations and the sign 'R 1 ' as a constant relational predicate. On Russell's view about senses ofrelations and logicalforms as subsistent universals, see his 'The Philosophical Importance of Mathematical Logic', The Monist 22, 4 ( 1 9 1 3) , pp . 48587, 492, and chap. 7 of his unpublished manuscript 'Theory of Knowl edge' . D.F . Pears discusses the relation of the latter work to Wittgenstein's Tractatus in 'The Relation between Wittgenstein's Picture Theory of Propositions and Russell's Theories of

228

6

7

8

9

10 11

Judgment' , Philosophical Review 86 , 2 (April 1977) , pp. 177-96. The lists make use of Russell's device of capping a variable to distinguish a constant sign :x:R 1 y' , standing for the relation (also indicated by 'R 1 ') or function, from the propositional function sign 'xR 1 y ' . (See ' Mathematical Logic as Based on the Theory of Types', in Logic and Kn owledge, ed. R. C. Marsh (New York : 1971) , pp . 59-102.) When referring to the relation in a discussion of Russell's views, I will use :x:R i5''. When I later discuss views of Frege and others, I will also speak of 'xR 1 y' as a 'predicate' expression . In 'Theory of Knowledge · Russell took the form to involve th e exemplification connection. Thus. he held that the proposition a is G had only two constituents. See 'Theory of Knowledge' . p. 183 . In keeping with this . Russell denied that forms were 'things' in the sense in which constituents . related in forms. were things. The basic modification will involve the recognition of facts as the referents of true propositions and the consequent distinction between relational concepts, as constituents of propositions , and relations, as constituents of facts. B. Russell. 'On the Nature of Truth' , Proceedings of the Aristotelian Society 7 ( 1906--7) , pp . 45--48. Theory of Knowledge' , p . 183. Theory of Knowledge· , p. 278 . Though Russell's view is actually more complex than my characterization of it , the difference is not important for the present issue . He holds that the complex [aR 1 b] should be represented by a definite description: ( Ly) (aD 1 y & bD 2y) . Thus , the relation R 1 is replaced by the relations D 1 and D 2 , which 'determine' R 1 , and he considers the atomic complex [aR 1 b] to be replaced by a molecular complex, since 'aD 1 y' and 'bD 2y' are terms of a conjunction. His move is designed to enable him to distinguish S judges that aR 1 b from S judges that bR 1 a

12

13 14 15 16

17 18

19

on the relational analysis of judgment he held at that time . (For a discussion of this analysis, see Thought, Fact, and Reference, pp. 309-32. ) His analysis will depend on the molecular expressions 'aD 1 y & bD 2y' and 'bD 1 y & aD 2y' having different atomic constituents whereas the complexes [ aR 1 b] and [ bR 1 a] have the same constituent entities. Sellars does not put it quite this way. But this is in keeping with what he says at a number of places, including 'Hochberg on Mapping, Meaning, and Metaphysics' , in Contemporary Perspectives in the Philosophy of Language, ed. P . French et al. (Minneapolis: 1978) , pp. 357-58. He acknowledges that a and b are related , but does not hold that he thereby recognizes relations as abstract obj ects . W . Sellars , Science and Metaphysics ( New York, 1960), p. 137. See my 'Sellars and Goodman on Predicates, Properties, and Truth ' , in this volume , pp . 185 f. W. Sellars , Naturalism and Ontology (Reseda : 1979) , p. 58. Sellars has denied that he does this , and I had earlier agreed with him (see his 'Hochberg on Mapping, Meaning , and Metaphysics' , p. 353, and p. 165 of my ' Mapping , Meaning , and Metaphysics' in this volume , pp . 157 f. but his new book makes it clear that he does so argue. W. Sellars , Science, Perception, and Reality (New York , 1963) , pp . 243--44. Ibid. , pp . 244-45 In Substance and A ttribute (Boston : 1978) , p . 367 , M . Loux also overlooks this simple fact in his disarming restatement of the nominalistic line of argument (though he argues for realism on other grounds) :

229

We ask why it is true, for example, that Socrates is wise ; and the answer we are given is 'Because he is wise' . What the realist contends is that this is not to provide a genuine explanation ; it is merely to restate what has to be explained .. . but I think that the extreme nominalist would be right to ask what other sentence the realist would have him use here . Is he supposed to say that 'Socrates is wise' is true because Secretariat is a horse or because Jimmy Carter is president? Pretty clearly not; if 'Socrates is wise' is true , it had better be because Socrates is wise ; and we can be sure that any theory of predication that suggests anything to the contrary is false. 20 21

22 23 24

25 26

27

28

29 30

JI 32 33

See 'Hochberg on Mapping , Meaning, and Metaphysics', p. 358. For Sellars, this merely means that we acknowledge green obj ects in the ordinary context. I speak of 'equivalence in some sense' , rather than follow Sellars's use of 'necessary equivalence' , since there is an important issue regarding the sense in which statements embodying philosophical analyses are equivalent to the statements of ordinary usage which they purportedly analyze. Not taking them to be necessary, as I do not, would provide yet another stumbling block to Sellars's gambit. But to go into such an issue would involve too long a digression. See 'Hochberg on Mapping, Meaning , and Metaphysics' , pp. 357 , 359 . This is so since , ultimately, the 'rules' direct us to ascribe predicates to appropriately characterized obj ects. See the correspondence between Sellars and Loux printed in Naturalism and Ontology. Sellars, Naturalism and Ontology, p. 1 08. Keeping the 'paradox' of analysis in mind, one can turn any 'definition' into an 'incoherent' pattern along Sellars's lines. Since , by stipulation, the definiens and the definiendum have 'the same sense' we may replace any defined pattern by its defining statement in ' ... = df ' . . • and obtain an 'incoherent' result, since a statement is then its own definition. Clearly patterns like ' . . . = df ' . . ' are not to be so used. Here Sellars may obj ect that I now do exactly what he does when he uses ·a is green· to express the truth condition for 'that-a is green is true'. The difference , of course, is that I have acknowledged a tie , exemplification . that plays a unique role, whereas Sellars seeks to have truth conditions at no ontological expense. This does point to the traditional problem of realism associated with Bradley's name and to how Sellars uses the Bradley regress to support nominalism. G. Bergmann , 'Propositional Functions', Analysis 17 ( 1956), pp. 43--48. Russell took this difference to distinguish particulars from universals. Loux, Substance and Attribute, p. 36. J. Wilkin has argued that Sellars's Jumbelese contains predicate terms in 'Sellars on Bradley's Paradox', Philosophical Studies 36 ( 1979), pp. 5 1 -59 . Since we are concerned only with ordered pairs of individuals (and not pairs of sets) . we avoid the standard counterexamples to construing (a;b) as { a, { b} } . G. Bergmann , 'Notes on Ontology' , in No!LS ( l 98 1 ).

Elementarism , Independence , and Ontology

Elementarism is the thesis that all undefined predicates of an improved (ideal) language are of the first type. 1 Some philosophers adhere to an explication of ontology , ( 0 1 ) , whereby the answer to the question 'What exists'? is provided by the referents of the undefined descriptive (non­ logical ) terms of one ·s ( implicit or explicit) ideal language. Proponents of (0 1 ) would then take the thesis of elementarism to assert the nonexistence of properties of properties. Recently Professor Bergmann has claimed that ·the only philosophical reasons which one could with some plausibility adduce in favor of elementarism, or, rather, the only such reasons I can think of, are specious'. 2 To produce philosophical reasons for a claim is, in Bergmann's terms, to show that it follows from one or several philosophical principles by themselves or in conjunction with commonsensical truths. To accept a proposition as a principle is to refuse to defend it directly and to argue instead, first, that the things it mentions are all commonsensical, and, second, that without accepting it one cannot solve all the philosophical problems'. The philosophical principle that Bergmann holds to be speciously involved in proofs of elementarism is the pri nciple of acquaintance - the contention that all undefined descriptive constants of an improved language must refer to things with which we are directly acquainted. I shall not, in this paper, be concerned with the ·specious' use of this principle that Bergmann argues against. What I shall try to show is that (1) a closely related principle supplies what may be considered a reason for accepting the elementaristic thesis ; (2) these considerations, in turn , may lead one to hold that 0 1 by itself, does not provide an adequate basis for the ' explication of ontology ; (3) these same considerations connect the elementaristic thesis with ( a) the view that relations do not exist, (b) fact ontologies, and ( c) nominalism. I shall use the phrase ·principle of exemplification', hereafter PE, for the two-part requirement ( 1 ) that all undefi ned descriptive predicates of an improved language , L, refer to characters that have been exemplified at least once and (2) that every particular named by a proper name of L has at least one primitive non-relational descriptive property . PE is thus a rule guiding the interpretation of L. A sentence will be called 'syntactically atomic' if it contains no defined signs and no logical signs (connectives, operators, variables). Thus a syntactically atomic

23 1

statement is of the form ·---( ...) ' where '---' stands for an undefined predicate term and '. .. ' stands for the subject term (s) (undefined predicate (s) or proper name (s)) of an appropriate type. A syntactically atomic statement S will be called 'independent' if given that S is meaningful and true it does not follow from the rules for interpreting L that any other statement which is not a logical consequence of S is meaningful and true. A state of affairs is atomic if it is referred to by a sentence that is independent. Let L be a phenomenalistic language. This means that the undefined signs of L name phenomenal entities and properties of such. L then contains independent sentences. For example, let ' 1 f t ' be an undefined first-level (nonrelational) predicate and 'a' be a proper name of L . The sentence (a) ' 1 f 1 (a)' is an independent sentence. Hence, the state of affairs it refers to is atomic. Where '2 0 1 ' is an undefined second-level (non-relational) predicate, the sentence (/3) '2 0 i ( 1 f 1 ) ' is syntactically atomic.But it is not independent and hence it does not refer to an atomic state of affairs.We can see why it is not independent i n two ways - the second may be considered a specific case of the first. (I) For (/3) to be meaningful and true PE requires that a sentence like (a) be true.For, in order for (/3) to be meaningful ' 1 f 1 ' must be, and for this to be the case " 1 f 1 ' must refer to a character that has been exemplified at least once.Hence there must be (have been) some particular such that a sentence like ( a) is true. To put it another way, if one who coordinates statements of L to somebody's contents now coordinates '2 0 1 (1f 1 ) ' , then he either has, or could have, coordinated ' 1 f 1 (a)' to an earlier content of that subject.To put it still differently, if what (/3) asserts is the case then so is (was) what a sentence like ( a) asserts. The acceptance of PE as a principle guiding the interpretation of an ideal language thus implies that (/3) is not independent and hence the state of affairs to which it refers is not atomic. This, in turn , points up that a sentence"s being syntactically atomic does not ensure that it is independent. (II) Essentially the same point is involved in a more specific form of the argument.Let 20 1 and 1 f 1 both first be presented to a subject at the same time and such that 1 f 1 exemplifies 2 0 1 • Then there must also be, by PE, an individual that is simultaneously presented and which exemplifies 1 f 1 • Hence, we know that if the state of affairs represented by (/3) occurs, then, at the same time, so does one represented by, say, (a). Hence , (/3) is not independent. Its not being independent provides, I suggest, one of the sources of the rejection of undefined second-level descriptive predicates by those who accept PE , 0 1 , and logical atomism.Logical atomism, whatever else it i mplies, has, I believe, implicitly involved the view, (A), that syntactically atomic sentences refer to ·atomic states of affairs' , or, in other words, that all syntactically atomic sentences are "independent' . Yet, atomic states of

232

affairs, as characterized above, are referred to by some but not all syntactically atomic sentences. Thus, as we saw, the concepts of ' independent sentence· and ·syntactically atomic sentence' do not jibe, and, consequently, (A) is rejected. But rejecting (A) destroys a symmetry between an improved language and 'the world' that has always appealed, if only implicitly, to logical atomists. This may be thought to weaken the atomists' case.However, there is an alternative to rejecting (A). The alternative is to reject (/3) , and all sentences like it, as syntactically atomic.Elementarism provides a ground for such rejection, since , upon the elementaristic thesis, all second-level descriptive predicates would be defined.This suggests that logical atomism provides a 'reason' for elementarism. It makes explicit what may have implicitly led some logical atomists to propose elementarism. But this is not the whole story.Before proceeding three points should be noted about the argument that (/3) is not independent. First, the connection between (/3) and a sentence like ( a) is not the result of an empirical law about phenomena and their properties. It follows, so to speak , from a ·principle' of one's philosophy, PE - what, in a more traditional manner, might be called a metaphysical principle. Second, the connection between (/3) and (a) is not that of logical implication. Neither (/3) nor (a) logically implies the other, since neither ' 1 f1 (a) :::J 2 0 i ( 1 f1 r nor '2 0 i ( 1 f 1 ) :::J 1 f l (a)' is analytic.Rather, to repeat, from its being the case that (/3) refers to a presented fact and from PE's being a principle or rule for interpreting L, it follows that what a sentence like (a) refers to is also the case.Note that in speaking of ( a) and (/3) I always referred to (/3) "determining' a sentence like (a) to be true and not (a) as such.I did this because, strictly speaking, what follows from (f3)'s being true and PE's being a principle or rule for interpreting L is that (y) '(3x)1 fi (x)' is true. For, we only know that some particular exemplifies 1 f 1 , not that a does.I used (a), and the consequent awkward phrasing, to emphasize the point that if we know that the fact to which the syntactically atomic sentence (/3) refers was presented then we know, by PE, that another fact, which could be referred to by a syntactically atomic sentence, was also presented. Thus, in a sense, PE may be said to 'connect' facts referred to by syntactically atomic sentences. All this does not bear on (f3)'s not being independent, since, like (a), (y) neither logically implies nor is implied by (/3) . Third, what was just said may be taken as an illustration of the point that 'to follow logically' an d 'to follow from the rules for a language' are two notions an d not one.For, from PE it follows that if (/3) is true so is (y).Yet, (y) is not a logical consequence of (/3).This shows the ambiguity of the phrase ' language rule', since we may distinguish formation rules, transformation rules, and rules,like PE, which guide the interpretation of a language. 233

We saw that an undefined second-level predicate cannot refer to a constituent of an atomic state of affairs. This fact may lead some to hold that the existence of properties of properties is suspect. Elementarism expresses this contention. We may then consider an elementarist to implicitly assert the thesis that 'to be is to be the component of an atomic state of affairs'. The elementarist thus proposes an alternative explication of ontology, which we may put 'linguistically' as follows: (Oz) , to exist is to be the referent of an undefined descriptive sign that can occur in an independent sentence. Yet, there is a certain awkwardness in distinguishing among the undefined descriptive signs with respect to ontological significance- in asserting (Oz) at the expense of (0 1 ) . But, in an elementaristic scheme (0 1 ) and (Oz) might amount to the same thing. For, in such a scheme only first-level predicates and proper names would be undefined descriptive signs. This, to repeat, may be thought to provide a motive and an argument for elementarism. Only in an elementaristic language could syntactically atomic sentences correspond to atomic facts. Yet, even in an elementaristic language, (0 1 ) and (Oz) need not amount to the same thing. If one holds to the second part of PE, they will not. To see this consider the sentence (8) ' 1 Rz(a, b )', where ' 1 Rz' is a two­ term relation sign of the first type and 'a' and 'b' are proper names. By the second part of PE, it follows that if (8 ) is true then there must be two states of affairs referred to by sentences ascribing n on-relational primitive properties to a and to b. Hence, by an argument similar to the one about (/3) , (8) is not an independent sentence and, consequently, does n ot refer to an atomic state of affairs. Thus, on (Oz), " 1 R2 ' would not designate an existent, while on (0 1 ) it would ; (0 2 ) may thus reveal a motive behind the rejection, by some, of relational properties as existents. It also establishes a connection between such rejection of relations and the similar rejection of properties of the higher types. This ' ontological connection' has an interesting corollary. The Wiener-Kuratowski procedure for defining first-level relations in terms of classes involves, we recall, classes of higher types. Similarly, any attempts to define all higher type descriptive predicates in terms of first-level ones would certainly involve relational first-level predicates. This means that we seem to be forced to have either primitive relational predicates or primitive predicates of the higher types. Consequently, for one who holds to both parts of PE, some syntactically atomic sentences are not independent. Hence, (0 1 ) and (Oz) cannot jibe! All this seems significant in that while (0 1 ) reflects one historical motif in classical ontology, simplicity, (0 2 ) reflects another , independence. Some philosophers have thought of existents, in some ultimate sense, as in dependents. Those philosophers who propose to explicate ontology in

234

terms of (0 1 ) have , I believe, ignored this other equally important theme. One strand of this theme may be thought to be reconstructed in the following: to be an independent entity is to be the referent of a sign that can occur in an independent sentence. Thus only individuals and first­ level non-relational properties would be independent entities. Another , strand of the explication of "independence is reflected in the second part of PE. One might suggest that this principle expresses a basic ingredient in the concept of 'individuality'. For, by it, to be an individual is to be, in a sense, independent of the relations among individuals. The second part of PE makes that sense explicit. One concerned with 'independence' might point out that in yet another sense no component of an atomic fact is "independent'. For, by PE, particulars require properties and vice versa. No particular is presented "bare' and no quality is presented unexemplified. This could lead one to hold that the only candidates for truly "independent' existents would be the referents of independent sentences - atomic facts. Such a one might then propose a further explication of ontology, (0 3 ) , 'to be is to be the referent of an independent sentence' . This brings us to the connection with a fact ontology that I mentioned above. A proponent of (0 3 ) may then contend that atomic facts are not only independents but simples, since independent sentences, being syntactically atomic, do not contain other sentences as parts. This point, of course, makes use of a different sense of "simple' from that used when we say, for example, that particulars are simples. Be that as it may, for such a fact ontologist two alternatives present themselves. He may hold that all independent sentences refer. True ones refer to facts and false ones to, perhaps, 'false facts'. But if he holds that the referent is the meaning of an independent sentence then he is forced to hold that a sentence means different things depending on whether it is true or false. This is unpalatable. Hence, he may separate the meaning from the referent of an independent sentence. He then no longer needs false facts, and may hold that only true independent sentences refer. Thus he is forced to the second alternative. Upon it a true independent sentence would have a meaning and a referent, a false one only a meaning.The meaning of a sentence would be the same whether a sentence was true or false. Thus a fact ontology would be 'complex ' in that each true independent sentence would, a la Frege, involve two entities - a sense and a referent - or, perhaps, a complex entity. 3 In any case making independent sentences the key to ontological commitment abandons an ontology of 'simples' in the sense in which particulars are simples. For, one may hold that proper names mean what they refer to, whereas, as we have seen, a fact ontologist is forced to abandon the identification of meaning and referent. To avoid associating a meaning and a referent with a true independent sentence a fact 235

ontologist might suggest that true independent sentences are not the only signs or sign combinations that refer . He may hold that while there are facts there are also things. The meaning of an independent sentence is then determined by the signs that make up the sentence - by the things that make up the fact. This would acknowledge that facts are not simples in that they are composed of individuals and properties. But they would not be complex in the sense of being associated with two things, or one complex thing - a meaning and a referent. Such a fact ontologist could also remind us that individuals and properties, while simples, are not independents in one of the senses we have considered . Hence, to do justice to the various motifs of simplicity and independence we have considered, he might suggest that an adequate explication of ontology should be threefold, embracing (0 1 ) , (Oz), and (0 3 ) . Or, to put it metaphorically, one might say that there are different l evels of existence which are revealed respectively by these three explications. To insist on the use of any one to the exclusion of the others would miss an important theme in ontology. On the other hand, we note that since the classes of entities selected, respectively, by (0 1 ) , (Oz) and (0) are different, there remains an irreconcilable conflict. To put it cryptically, in all the senses of those slippery terms 'simple' and 'independent', all simples are not independent and all independents are not simple. Two further points remain to be made. First, I spoke above of a fact ontology possibly involving a complex entity - a combination of meaning and referent. I spoke this way in order to point out a connection between fact ontologies and a familiar medieval pattern . In this pattern one distinguishes the essence and existence of an object. Existents are then objects to whose essence existence has been added. The essence provides the ground for the conceptualization of the object . Similarly , we have the meaning of an independent sentence, whether it be true or false . But if it is true , there is a further "component' , the referent. In a fact ontology this corresponds to the existence of the essence. We may also note the similarity of fact ontologies to the Aristotelian pattern , where material substances are composed of form and matter . Since forms of material objects and prime matter are not independent existents, the "simplest' 'independent' material existents are composites of form and matter. Aristotle's material substances are thus "facts' compressed into " things' . Second , I mentioned a connection of this discussion with nominalism. To see this consider a hierarchy of undefined predicates of types / th rough · a sentence (� "'"' ) 'nf 1 ( n- 1 f 1 ) ' , b y th e argument n, ' 1 f 1 ' , 'Zf 1 ' • • . 'nf 1 ' • G 1ven showing (/3) not to be an independent sentence, (l) 's being true and meaningful 'implies' that a sentence of the form •n-lf 1 (n-Zf 1 ) ' be true . Similarly this latter sentence's truth requires a sentence involving a predicate of the next lower type to be true , and so on. The regress 236

continues until we arrive at a sentence with a proper name. Proper names n ame individuals . Hence individuals may seem to anchor the chain of syntactically atomic sentences. This may lead one to stress the ontological importance of individuals . They may even seem to be the basis for their being independent sentences . Thus the notion of an independent sentence may reveal a thread, albeit a minor one, in the complex of motives that could lead a philosopher to nominalism.

Notes For discussions of elementarism see J . Weinberg, 'Concerning Undefined Descriptive Predicates of Higher Level', Mind, 53 . pp . 338-44 ( 1954) ; G. Bergmann , ' Elementarism' , Philosophy and Phenomenological Research, 18, pp. 107-14 ( 1957) ; L . Palmieri , 'Higher Level Descriptive Predicates· , Mind, 54 , pp . 544-47 ( 1955) , and 'Second Level Descriptive Predicates·, Philosophy and Phenomenological Research, 16, pp . 505-1 1 ( 1956) ; H . Hochberg. 'Professor Storer on Empiricism' , Philosophical Studies, 5 , pp . 293 1 ( 1954) , and ' "Possible " and Logical Absolutism' Philosophical Studies, 6, pp . 74-77 ( 1955 ) . � Bergmann, p . 108 . 3 This discussion does not presuppose that all descriptive signs , or sign combinations , of an ideal language require an entity that is their meaning - only those signs ( sign combinations) whose meaning is not somehow specified in terms of others . 1

Ontology and Ac quaintance

Is a principle of acquaintance a guide in ontology? The question is raised since some claim to be acquainted with such things as substrata (bare particulars), universals, and the nexus or structural tie of exemplification. 1 A principle of acquaintance reflects the idea that to use simple signs (undefined or primitive predicates and proper names) as referring signs we must be acq uainted with the referents of such signs. Since it is also held that such simple signs refer, in a properly clarified language, to 'ontological simples', the elements of an ontology, the principle also involves holding that one is acquainted with such entities. Complex signs, defined in terms of other signs, may indicate things 'indirectly' by means of the signs that define them. A definite description is an example of this latter kind of sign. Seeing two white patches I am, in a perfectly unproblematic sense, acquainted with two obj ects and can simply (or directly) label (or name) them. Likewise, distinguishing the qualities white and square from each other and from the patches, I can , simply or directly refer to them by the predicates ·white and ·square' . To use a metaphor, one can 'point with the mind's eye'. Avoiding the metaphor, we note that all that is involved is being able to distinguish, in experience, two obj ects and two qualities. A definite description must distinguish an object by its constituent predicates in order to be used to indicate it. This is one point involved in a principle of acquaintance. Another concerns the use of signs fo r qualities not distinguishable in experience. I do not know what the predicate " Blipy' means since I have not experienced a quality of things which it refers to nor am I familiar with any definition of that sign. In short the principle of acquaintance, while introduced into philosophical discussions, involves our talk about ordinary things and qualities, not our discussion of the ontological makeup of things. Consider three alternative ontologies: (a) considers a white square patch to be a substratum and a group of universals connected by the nexus of exemplification; (b) holds that it is a set of quality instances (this whiten ess, this sq uareness, etc. ), perfect particulars as some call them, connected by some structural tie; ( c) holds it to be a set of universal qualities combined by a nexus. N o one who adheres to (b) or (c) would hold that when he sees the patch he is acquainted with it as a set of quality in stances or as a set of universal qualities. A reasonable proponent of (c) 238

would acknowledge that since he is acquainted with the patch and it is, on his analysis , a set of universals connected by a structural tie, he is, in a sense , acquainted with a composite of universals. But this is not to say that he is acquainted with the obj ect as a composite of universals. It is to say no more than that an object of acquaintance , the patch, is considered to be composed in a certain way. To be acquainted with an object is not to be acquainted with it as a particular ontological analysis holds or reveals it to be. This is to confuse philosophical analysis, which is dialectical in character , with a ludicrous sort of phenomenology. Further, to be acquainted with a quality is not to be acquainted with that quality as an ontological position holds it to be - a universal or a perfect particular, for example. Recognizing this, one recognizes that the claim of some to be acquainted with substrata and the tie of exemplification is hopelessly, if simply, confused. Part of the confusion about acquainance stems from a metaphysical use Russell put it to . It led him and others, following him, to hold that they were acquainted only with phenomena , not physical objects. Since phenomena were the objects of acquaintance, physical obj ects were held to be either constructions from such phenomena or unknown things related to them in some way. One who took the first alternative could then come to fuse 'being an object of acquaintance' with 'being what was really there'. Phenomena became the elements into which physical objects were analyzed. Standing in this tradition, one who proceeded to analyze phenomena into particulars (substrata), universals , and the tie of exemplification took it as a natural move to hold that such things were obj ects of acquaintance. Note though that physical obj ects were analyzed into phenomenal ones since only the latter were held to be objects of acquaintance. But substrata and universals are held to be obj ects of acquaintance since phenomena are analyzed into them ! The reversal is instructive . It reveals that one holds that he is acquainted with entities his metaphysical arguments lead him to, since they are constituents of what he is acquainted with in a quite ordinary sense. That is why we cannot argue about seeing a white patch (though we may argue about its analysis) ; but seeing a substratum is another matter. Of course, if one insists he is acquainted with the substratum since he is acquainted with the patch then it is clear he assumes that what is a constituent of an obj ect of acquaintance is itself an object of acquaintance. (' Constituent' here refers to the relation in an ontological analysis ofthe kind involved in (a), (b) , and (c), not to that between a chair and one of its legs.) With such an assumption the principle of acquaintance becomes vacuous as a guide in ontological investigations. For, under such an assumption, one may be acquainted with anything from monads to Absolutes : with anything introduced by metaphysical arguments to 'account for' the ordinary 239

things and facts we start from. But without such an assumption there is no ground for holding that one is acquainted with substrata, if such there be. If one who adheres to (a) still insists that he is acquainted with a substratum since he is acquainted with the patch, then either he merely restates the assumption or says nothing. The talk of reduction of physical objects to phenomena may be thought to pose a problem in speaking about acquaintance with ordinary objects. For, surely, physical objects are quite ordinary, yet some philosophers have held we are 'really' acquainted with phenomena on those occasions when we speak of experiencing physical things. Similarly , one might suggest we are really acquainted with what phenomena and physical objects are ultimately composed of, substrata, universals, and the nexus that connects them. One difference between the two cases we just noted. Second, without going into the complex of issues that surrounds the phenomenalism-realism dichotomy, we can note that the relation between phenomena and physical objects poses a different sort of question from that posed by the relation of substrata to things that supposedly contain them. In one perfectly clear sense both phenomena and physical objects are ordinary things we start from. But no one could claim this about substrata, universals, and exemplification . The attempt to reduce physical objects to phenomena, or vice versa for that matter, is an attempt to hold that one sort of ordinary thing is ·reducible' to another sort. In short 'constituent' is used quite differently when one holds a physical object has phenomena as constituents and when one holds a substratum is a constituent of a thing. Part of the problem goes back to the historic confusion , from Berkeley to Moore, between a particular phenomenal thing and phenomenal qualities - between a white patch and the quality white. Hence the reduction of physical objects to phenomena was fused with the analysis of things into qualities. This reinforced, in turn, the identification of objects of acquaintance with elements of one's ontology. The reintroduction of substrata by some in the empiricist tradition then understandably led to confused talk of acquaintance with such things. 2 A third point to note is that to get at the issues we are considering we can confine the discussion to phenomenal entities. The white patch we started out with may then be thought of as a phenomenal, rather than a physical thing. We noted that just as one does not experience such a patch as either a substratum exemplifying qualities or as a combination of qualities, one does not experience the qualities as either universals or instances of such. Qualities, like things, are what we start from, in experience. The principle of acquaintance functions to differentiate a quality like white from one like blipy ; it does not assure us, in so doing , that we are acquainted with universals. The identification of a quality as a universal is

240

a matter of philosophical argument or dialectics . Hence , while the assertion that one is acquainted with a substratum involves the gratuitous assumption that a constituent , on an ontological analysis, of an object of acquaintance is itself such an object, there is no such assumption in speaking of qualities and universals. What would correspond to such a claim in the case of qualities would be to assume that we are acquainted with qualities as universals. This would be as confused as the corresponding claim of acquaintance with substrata. Yet being presented with a quality is sufficient for the use of a primitive predicate sign, just as being presented with a white patch is sufficient for its being named. But being presented with a patch is not sufficient for a constituent of it, a substratum, being named - if we adhere to a principle of acquaintance. This fundamental difference reflects the difference between a patch and a substratum, on the one hand , and a quality as a universal , on the other. The substratum is part of the composite patch : a universal is not part of a simple quality ; it is what we come to argue that the quality we find in experience is . The name of a substratum does not function as the name of an object of acquaintance. Hence for Russell such a sign would really be an implicit description. Moreover the description would be something like "the substratum in Peter' , where "Peter' named the white patch. But then what about the property of being a substratum? Clearly one cannot define it in terms of ordinary qualities one is presented with. This leaves two alternatives . One must either claim acquaintance with it or acknowledge that such a notion (whether one calls it "individuality' or ·particularity' makes no difference) is meaningful only in a contextual sense - in the context of a metaphysical position and its contrast with alternative views. The case of universals is different. Indeed 'universality' like 'particularity' is given meaning by a contextual setting of metaphysical questions and alternative answers. But, since qualities are presented , such a notion is, as it were, grounded in experience. For , upon the claim that qualities are universals, the notion of universality refers to (or characterizes or corresponds to) distinguishable qualities that are presented. That is, somewhat like the empirical interpretation of an abstract axiomatic system, the presented qualities ground or anchor the metaphysical concept of universal. Nothing corresponds to this in the case of substrata. 3 The point can be driven home when we recognize, if we are honest, that we must refer to the substratum in Peter by such a phrase and not by a simple name. However, we need not (and cannot) refer to the quality white by a corresponding phrase involving the notion of 'universality'. (I say cannot because doing so would involve us in further confusion by replacing neutral ordinary terms with a particular metaphysical apparatus. But this point is not crucial here.) Distinguishing white as a quality from other qualities, we can simply label 24 1

it. Ordinarily speaking we even distinguish the quality from the (ordinary) thing that has it. This is obviously implicit in noting that a thing has several qualities. Again nothing corresponds to this in the case of substrata. But this is simply to note a fundamental difference between universals and substrata as ontological elements. To adhere to a principle of acquaintance then forces one to abandon substrata since, on it, one must 'know' by acquaintance the simple elements he speaks about- be acquainted with the referents of the simple signs that designate substrata. Alternatively, one may, to retain substrata, abandon the principle of acquaintance. But, in view of the distinctions we have noted, one need not abandon such a principle in order to have primitive predicates refer to universals. It is sufficient to be acquainted with the qualities. One need not be presented with the fact that qualities are universals. (Recall that a universal is not an 'unknown' part of a presented quality. ) To abandon the principle of acquaintance in order to retain substrata is thus to include, in a perfectly clear sense, an 'unknowable' element in one's ontology. This may bring one to reconsider the need for substrata in view of alternatives (b) and (c), 4 for these latter require no such 'unknowable' . This, of course, reflects a basic theme of the idealistic gambits of both Berkeley and Bradley. But ontologies like (b) and (c) are not intrinsically idealistic. That, however, is another story. One last consideration. Claiming substrata to be presented, some have also claimed exemplification to be presented. Recognizing that the claim about substrata is confused, we also see the claim about the nexus to be so . But, even if we consider things with which we are acquainted to be composites of universal qualities, with which we are also (as explained) acquainted, one could not claim acquaintance with the nexus that combines qualities into things . In spite of this, there is no problem corresponding to the one about substrata . For such a nexus is not an entity or element and no sign for exemplification need be tied to anything. This point is reinforced by recalling that to consider exemplification as a relation among relations leads to Bradley's regress. What is involved is that a composite of qualities is more than a class of such, or , to put it another way, a fact is more than the constituent entities that make it up. Yet a sentence like " Peter is white' indicates a fact, or is about a thing, in virtue of the signs in the sentence . A sentence, like a defined sign, refers, if it does, indirectly - in terms of other signs. 5 There is no need to be acquainted with a nexus for it to do so. In short the linguistic device of predication, while taken to reflect some ontological feature, need not be tied up with an element of experience for us to know what a sentence says. However, since we are acquainted with things and qualities but not with a nexus, while holding that there must be one, we may take this to

242

reveal what force there is in Bradley's claim that a nexus is incomprehensible. In this sense we do not , ultimately , know what a sentence says since we do not know (are not acquainted with) what connects the referents of the constituent terms of the sentence. 6 One might suggest that since we are acquainted with things , which are held to be composites of qualities, there is a sense in which one is acquainted with the nexus. Not being an ingredient of the thing , as a substratum is held to be, the nexus is experienced since one experiences the thing. But whatever differences there are between a substratum and a nexus, both differ from qualities in that we can distinguish , in experience, qualities from each other and from things ; but we certainly do not so distinguish either a nexus or a substratum. Insofar as this is so, both Bradley and Berkeley have made their respective points. Notes 1

For example see G. Bergmann , Logic and Reality ( Madison : University of Wisconsin Press , 1964 . 1965 ) , p. 47 ; also E. Allaire , Philosophical Studies. 16, pp. 1 6--2 1 (January-February 1965 ) . I t i s interesting to note how the pressure to make 'unintelligible' substrata 'knowable' or 'intelligible· has led from the prime matter of Aristotle to the extended substrata informed by 'corporeal light' in Grosseteste and the materia signata of Aquinas , to extension as the essence of corporeal substance in Descartes , to Moore's material objects as 'parts of space' (Some Main Problems of Philosophy, Chapter 1 0) , to Bergmann's recent identification of the substratum with the 'area' of a thing like a white patch. ' We have seen why it will not do to say that the presented difference ot the two things does this , in view of the required assumption that what is part of a thing , on an ontological analysis, is presented if the analyzed object is presented. The point is that a quality , which is itself presented , is identified dialectically as a universal. A thing which is presented is claimed dialectically to contain a substratum as a constituent. The thing is not identified, in any sense, with the substratum. (This may be overlooked since on the substratum view the propername of the obj ect is taken to 'really' name the substratum.) One might claim that since things are presented one may consider a notion of 'individual' or ·particular' to be derived from experience. But, even so, this would not be 'individual ' in the sense of 'substratum' or bare particular. Nor would those two notions even coincide or apply to the same entities as would , say, ·quality' and 'universal' . It is no accident that Bergmann not only claims acquaintance with substrata but with 'individuality' , which such things exemplify (Logic and Reality, pp. 47-48) . 4 For such a reconsideration see my Th ings and Qualities' in the proceedings o f the annual Oberlin philosophy meetings, 1964 . 5 This is not to confuse a sentence with a defined sign. 6 On alternative ( c) , when Peter is considered as a composite of qualities, the nexus connects, not the referents of the terms 'Peter' and 'white ' in the sentence 'Peter is white' , but the constituent qualities of Peter , of which white is one. (For details of this see reference cited in note 4. ) This complication has no bearing on the point. All that is involved is that , on such a view, predication, as a linguistic device, does not correspond to a relation between two independent entities, as it would on the alternative adhering to substrata , universals , and exemplification.

243

Things and Descriptions

In A n Inquiry into Meaning and Truth, Russell proposed to consider a phenomenal thing like a square white patch as a collection of qualities. He did so to avoid an alternative view which would hold that the patch consisted of a bare particular (substratum) related to qualities (universals) by the special relation (ontological tie) of exemplification. To Russell his alternative eliminated an 'unknowable' that had bothered philosophers since, if not before, Aristotle's introduction of prime matter. 1 Russell recognized that a problem arises in trying to distinguish adequately one collection of qualities from another - two white squares, for example. In short , he faces the problem of individuation. This we shall take up later, for I hope to show that certain arguments purporting to establish that substrata must be recognized in order to deal adequately with the problem of individuation are unsound. 2 Before doing so we shall discuss some related questions that arise in the analysis of the idea of one thing. Consider a white square that has only the additional property designated by 'P 1 ' , with 'W1 ' and '5 1 ' standing for "white' and ·square'. If 'Socrates' were the name of the white square, then , on Russell's view , that term may be considered as an abbreviation for either a definite description of a second order class or property specifying that only P 1 , W1 , and 5 1 are members of it or for the set sign ' { W 1 , 5 1 , P 1 } ' . [Let (a) stand for the description and (�) for the set sign. ] Thus Socrates, like all individuals , becomes a class or property of properties. 3 Such a proposal involves several difficulties. To say that Socrates is white would be to assert either 'W 1 E( a)' or ' W1 E(/3)'. 4 But both statements are analytic truths. To put it loosely, classes being what they are , the assertion that there is one and only one class having as its only members W1 , 5 1 , and P 1 and that W1 is a member of that class is analytic, as is the assertion that white is a member of a class defined by enumeration to include white. The same would hold for assertions ascribing 5 1 and P 1 to Socrates. Since the sentences asserting the existence of such a class, using either ( a) or (/3) , are also analytic truths, that there is such a thing as Socrates also becomes an analytic truth. There are even stranger consequences. Assume that there were no white circles and no black squares, but that there was a black circle indicated either by a description analogous to (a) or by the set sign ' { B 1 , C 1 , P 1 } ' where ' B 1 ' and 'C 1 ' stand for "black' and 'circle'. Having

244

such terms and properties one could then construct (descriptions or) class signs like · { B 1 , S1 , P 1 } ' and · { W1 , C1 , R 1 } ' . Again, classes being what they are, statements ascribing existence to such classes would also be analytically true. To put it paradoxically, on the view that reduces things to classes of qualities, nonexistent things become necessary existents . The problem lies in the incompatibility of the logical properties of classes with the analyzing of things into classes of qualities. To avoid all this one would have to have some way of distinguishing between classes that were things and classes that simply were classes. A special property, say existence, at the level of properties like W1 would not do. Instead one might suggest the introduction of a higher level property that existent classes would exemplify. That some such thing must be done points up the peculi arities of the position. This point also serves to contrast sharply Russell's view, which turns the sentence 'Socrates is white' into a statement of class membership, with the view that predication in language reflects an ontological tie between elements of a fact or of a complex thing. Class membership as a linguistic device reflects no such tie, nor is there any on Russell's view. 5 This is precisely what leads to the problems we just considered. To avoid such problems some device must be introduced which connects the members of some classes of qualities into individual entities. But then this connection will furnish the ontological tie, not class membership. 6 One who holds that exemplification is a relation between a bare particular and a universal property might also claim that so viewed exemplification is something he comprehends in that he is acquainted with such an ontological connection. Class membership, however, is a logical relation he understands only in terms of predication and not in terms of direct acquaintance. Such a claim introduces a principle of acquaintance as a crucial theme in one's metaphysics. The issues surrounding such a principle will not concern us here. 7 A proponent of exemplification and bare substrata might also contend that the ontological tie must connect or relate entities that are independent of the relation. This is not so for class membership, where a class is specified in terms of its members. In part this reflects the concern over the analyticity of sentences like 'Socrates is white' on Russell's view; but it also reflects other concerns that we shall take up later. A consequence of statements like 'Socrates is white' and 'Socrates is square' being analytic is that all statements truly ascribing properties to Socrates are logically equivalent and, in that sense, say the same thing. Furthermore, if, contrary to the simplified case we are considering, Socrates was discovered to have additional properties, then we could not, on Russell's view, truly predicate them of him unless we altered our analysis of Socrates. One can only truly ascribe properties to Socrates that are included in the

245

specification of the class which Socrates is. One might then conclude that to say anything about Socrates is not only redundant, but involves knowing everything about the patch. If one adds relational properties to the class then one is on the road to internal relations and Bradley's absolute. This holistic theme will also occupy us later in another context. Quine has proposed the replacement of proper names by definite descriptions. He has some explicit motives for doing this. One that is not so explicit may be the same as Russell's . If we use a 'regular' description instead of (a) to define 'Socrates' we would have (y), '(1x) [ W1 x & S 1 x & P 1 x] '. If one uses (y ) instead of the proper name 'Socrates' one might feel that he is not required to recognize an entity that the proper name designates which is distinct from, independent of, and in addition to the properties specified in the description . For, following Russell, the meaning of a description is specified by the predicates in the description; a description is not a 'denoting' sign. The description can refer to something indirectly without, like a name, being connected to that thing.8 A name to be used independently of any description of what it names must be connected to something directly and not by means of other terms. The thing named and the connection of the name to it provide the ground for a sentence in which the name occurs, being about the thing named or referring to some fact. A description may be said to be about something in a different way. It is connected indirectly through the specified properties. This may lead one to hold that the use of a description, when there is a thing fulfilling it, reflects a consideration or analysis of that thing in terms of its properties. A name, not making use of any properties of a thing, lends itself to the idea that it refers to something about the thing other than the properties of it - the substratum or bare particular. The use of descriptions, as opposed to names, would go along with a view that considered an individual to be composed solely of universals or properties in combination. The description ( y ) does not reflect the turning of Socrates into a class of properties, but it may be thought to reflect his analysis into a composite of qualities. On the bare particular analysis, Socrates would be a bare particular tied by exemplification to universals: white, square , and P 1 • On the class analysis the white patch is simply a class of qualities. A third analysis is to consider the white patch as a composite of qualities in a special structural connection or tie that would correspond to exemplification on the bare particular analysis. On this alternative the ontological tie would hold only between universals and not between universals and a further kind of thing, a substratum. The analysis of the term 'Socrates' by ( y ) may be taken to reflect this view. Such a view recognizes particulars, but not as either simples or substrata. Since this alternative recognizes an ontological tie that connects qualities into things, one may hold that it does not get rid of particulars in the way 246

that Russell did. On Russell's view the lack of an ontological tie may be considered to reduce individuals to their constituent universals in a way that the present alternative does not do. [It is perhaps relevant to recall that Russell considered classes to be ·logical fictions' . ] However, the present alternative does get rid of bare individuals or substrata . If defensible, it thus reaches Russell's goal . Before pursuing the above point let us compare the use of (y) with the use of a proper name like "Socrates'. Consider three sentences, using (y), to assert · Socrates is white', ·Socrates is square', and 'Socrates is P 1 ' . None are analytic truths. Nor would a sentence asserting that the description (y) is fulfilled be an analytic truth. This points to a radical difference between (y) and (a). Furthermore, the problem about the nonexistent black square does not arise in the case of (y) . Yet the three statements asserting that Socrates is white, square and P 1 are logically equivalent to each other and to the assertion that the description (y) is fulfilled. In this sense all these statements, using (y), say the same thing. 9 But this is neither surprising nor detrimental. To name Socrates is only to indicate him. Where the term 'Socrates' is a name, the sentences 'Socrates is white' and 'Socrates is square' each state that what is indicated by that name has a certain, and different, property . A description does not merely indicate. It indicates by means of properties purporting uniquely to determine an object. Hence, to ascribe such properties to the object, indicated by means of them, is to do something different from indicating by means of a name. Moreover, there is an analogous , though not explicit, feature in the use of names. When seeing a white square and naming it or referring to it by a purely indicating sign like "this' or a proper name, we do so in virtue of something we notice about it. That is, one does not come across a substratum or particular apart from properties. One confronts it, if at all, exemplifying properties. In applying a name to a bare particular or substratum proponents of such things think of distinguishing them from the properties they exemplify. This may be looked upon as j ust another way of saying that the sign used as a proper name has only an indicating function. Hence, a bare particular becomes a hypostatization of this function of a sign. The use of descriptions makes explicit the fact that where something is indicated properties of that thing play a role. Proposing the use of definite descriptions in place of proper names may thus reflect the rej ection of bare particulars as elements of one's ontology. This may be gotten at, alternatively, by holding that proper names name complexes of qualities, yet are simple, primitive signs. One who argues in this way rejects the notion that language must picture objects and hence that simple signs cannot designate complex objects. The proponent of descriptions, as opposed to names, as indicators of comp ! exes of qualities might then be

247

thought to accept, to a degree, such a 'picture principle' of language while rejecting bare particulars as entities. Since his individual objects are composites of universals, the signs corresponding to such individuals must be composites of signs which refer to the universals involved. The composite sign is linked to the complex object it corresponds to since the signs it is composed of refer to the entities the object is composed of. Furthermore, while descriptions, as complex signs, would correspond to complexes of qualities as complex entities, they woul d do so only to a certain degree. That is, a description need not contain predicates indicating all the qualities of the indicated object. An object described by (y) on the present alternative could have further properties. Thus another problem that arose on Russell's view does not arise. Suppose , a bit more realistically than our simplifying assumption that Socrates has properties in addition to W1 , 5 1 , and P 1 , but that predicates referring to these additional properties do not enter into his description in (y). To ascribe any such additional properties to him will not in the least be to say 'the same thing' as ascribing properties included in the description. Nor will any two such ascriptions say the same thing as each other. Thus a further difference to Russell's view is involved. Moreover, one must not be misled by the following argument. Suppose W1 , 5 1 , and P 1 suffice to individuate Socrates, then if in addition he has R 1 and Q 1 , the set of properties W1 , 5 1 ,P 1 , R 1 , and Q 1 will also individuate him. Let (y 1 ) stand for a description constructed from this latter set of predicates. Then the identity ' (y) = (y 1 ) ' will hold. We could then replace the definition of the term ' Socrates' by the more extensive descriptive phrase . One could do this for all the descriptive properties of Socrates and hence turn any statement ascribing one of these to him into a statement logically equivalent to any other statement ascribing a different property to him . The point to be made in reply is that the above identity statement is synthetic. Those who advocate a bare particular analysis and the naming of such things still , generally, hold that one indicates past objects or things one is not now acquainted with, and hence not capable of being simply and directly indicated , by definite descriptions. Hence , to say about such things that they have or had certain qualities involves the same seeming redundancies. On a bare particular analysis one avoids these seeming redundancies at the price of introducing two kinds of entities. First , there is the mysterious bare particular which is named. Second , this entity combines, by predication , with a universal to form a fact which determines the truth of a sentence asserting that the thing has the quality . Such a true sentence may even be thought to refer to this further thing , the fact. On the alternative analysis , since the white patch is considered a combination of qualities no further entity is involved. A true sentence

248

ascribing a quality specifies a constituent of the complex indicated by the descriptive phrase. It does not assert that something is related to what is indicated and hence refer to a third, complex , entity comprising the two relata in a relation of predication. Holding that Socrates is a composite of qualities one might say that it is the complex thing that makes the sentence true . All this brings us to the question of individuation , for it is on this issue that the bare particular analyst bases his case. Supposedly, alternative views cannot account for individuation and difference. Quine sought to solve the problem of individuation by introducing peculiar individuating properties and basing descriptions on predicates referring to such properties - "Pegasizing'. He made one mistake in holding that such primitive properties could be introduced for unfulfilled descriptions. Thus he had signs whose only function was an indicating one with nothing being indicated. But we are not here concerned with questions about references to non-existent 'things' . Talking about existent things having a unique property, which serves solely to individuate them, simply puts proper names (and bare particulars) at the predicate level . Such properties of individuation are certainly as puzzling and mysterious as bare substrata, and, like these latter, are hypostatizations, in a more devious way , of the simple indicating function that some terms may have. Russell tried another approach. Consider two white patches , Socrates and Plato. Assume that they are alike in all non-relational descriptive properties and that Socrates is to the left of Plato . To distinguish them qualitatively Russell assigned to each the property of being at a place in the visual field . Some would object to this on the grounds (a) that they are not acquainted with such properties of things, and (b) that space is relational, while on Russell's proposal it is not 10 . But is Russell's suggestion so outlandish? Given a succession of phenomena cannot one recognize that a patch is in the same part of the visual field as a previous one , just as he recognizes that it is the same color as the previous one? However, given a succession of exactly similar visual fields one might also have to have recourse to similar properties of a temporal kind. Perhaps such properties cannot stand up to philosophical probing. But this question can be waived here, for one has had recourse to such properties due to the acceptance of certain arguments that forbid the use of relational properties for purposes of individuation . These arguments, I wish to show , are not cogent and, consequently, one may employ relational predicates in descriptive phrases to indicate Socrates and Plato . Suppose that in order to distinguish the descriptions of Socrates and Plato, where Socrates and Plato are two exactly similar patches with 249

Socrates being to the left of Plato, one proposes to include in the description of Socrates the predicate 'being to the left of Plato'. This would immediately be seen to be inadequate since the definition of such a predicate would have to include the descriptive phrase indicating Plato. This in turn would have to include the predicate 'being to the right of Socrates' or it would not have been uniquely specified as distinct from Socrates, for, recall, all its non-relational properties are shared with Socrates. But this problematic situation is easily changed. Suppose one introduces the property L, with 'R' standing for 'right of, by

Lx = df (3y) (Ryx & Wy & Sy) and suppose further that no white patch is to the right of Plato. For a thing to have L is to have a white square to its right. To the use of such a property the proponent of bare particulars retorts that to use a relation we must already have terms standing in that relation , and hence to use a relational property as a constituent property in forming a definite description is illegitimate. This argument is confused on two counts. First, it confuses 'being about' a white patch in the sense of saying that something is white with formulating a descriptive phrase to indicate or be about the thing . Thus in formulating the descriptive phrase to indicate Socrates one refers to Plato only in that one talks about something being to the right of Socrates. It is not a question of using the descriptive phrase for Plato in the construction of a relational property to individuate Socrates and of using the descriptive phrase for Socrates in the construction of such a property for Plato. Consequently no circularity is involved . Second, one might, to use a cryptic and loose phrase , say that logical and temporal priority are confused. Perhaps the point can be clarified in the following way. One must indeed notice Socrates as distinguished from Plato. One can also refer to these different things by different names. This does not mean, having done this, that one cannot consider an analysis of the things in terms of properties and of the names in terms of descriptions composed of predicates referring to those properties. Two patches are noticed to be different and in a spatial relation. To notice them in a spatial relation does involve that instance of the relation depending on there being things related. But this does not mean that the things related do not also depend on that instance of the relation. Without there being some spatial relation between them 'they' would be one and not two. Nor does it mean that such mutual dependence prohibits the use of the relation in an analysis of the things . One might think otherwise ifhe confuses a relation with an instance of it and thinks that ' L' must be defined in terms of "being to the right of Plato' instead of in terms of 'R'. 250

Also, if one holds that exemplification is the on tological tie or relation and that such a tie must hold between independen t relata, he might hold that all relations require ontologically independent relata. It is understandable why one would hold that exemplification req uires independent and simple relata. Recall the sentence "Socrates is white' . If the subj ect term is held to refer to a composite entity that contains whiteness, and thus depends on the universal, one might feel that predicating "white' of Socrates is redundant, or empty or analytic. This we discussed earlier. Furthermore, one who adheres to a picture principle of language would naturally believe that predication in language ought to reflect the ontological tie between the simple elements of an ontology. On the view adhering to substrata , universals, and exemplification it seems to. On the alternative view it apparently does not since the description (y) refers to a complex in which both the tie and the universal are con stituents. The ontological tie on this view combines simple universals into things� it does not combine a substratum and a universal in to a fact. This difference is behind the fear that predication without substrata is empty. B e that as it may, that relations are not presented without relata does not mean that one cannot analyze particulars in terms of q ualities and relation s. Ontology is not phenomenology. Thus even if one may n otice something without being aware of what relations it stands in or properties it has that fact has no bearing on the issue. To the bare particular an alyst it might. Thus he may argue for numerical difference as distinct from conceptual difference by holding that he can apprehend our two patches as different with out noticing how they differ. One might then think that if he n otices simply that two things differ, they must differ in simples. Since they do not differ, in our example, in simple nonrelational properties, an d, since relational properties 'involve' terms (and hence are thought incapable of grounding simple difference) , they must differ in simple bare particulars. From what one simply notices one is thus led to a kin d of ontological simple. Since, via a principle of acq uaintance, one convinces oneself that a bare particular is an obj ect of acq uaintance, the circle is closed and the knot is tied. The bare particular is the ontological representative of apprehended differen ce and the ground of numerical difference. We may protest that all this is far too simple. For some might feel that they j ust don't know what it is to apprehend two particular things as different without apprehending them to be different in some way. But that aside, the above argument confuses what I apprehend with the analysis of what is apprehended. To put the matter slightly differently we may say that without there being a difference in property (including relations) between Socrates and Plato there would be no apprehension of simple difference. That this is so points up that what "makes' them different is one question. Whether they may be apprehended as simply

25 1

different is another. It is relevant to notice that for the bare particular analyst there is the possibility of having Socrates and Plato be different while not differing in any properties or relations , for bare particulars may be simply different, and, I take it, could be apprehended as such. That this does not happen is, from the perspective of the bare particular analyst, a fact about our world. Looked at from another perspective it reveals an inherent absurdity in the bare particular analysis. In any case , as I am said to notice something being simply different from something else , it is then thought that this difference must not , ultimately, be grounded or analyzed in terms of properties or relations. In addition to the problematic nature of this assumption, there is the sheer question­ begging element involved in the notion of 'something'. If 'something' is taken as a bare particular then , of course, from the very role such a thing plays it is simply different from another such thing. But if the something is taken as the patch from which we start our ontological analysis, then the matter is not closed. We may still consider it to be a composite of qualities. If at this point one argues that it cannot be analyzed as a composite of qualities, because it is just seen as a distinct thing, not as a composite , I hardly know what to say, except to repeat that ontology is not phenomenology. Of course Plato and Socrate5, are apprehended as distinct things. This is where analysis begins. But this does not mean that the bare particular analysis is the correct one. Some may think that it does - that to say that bare particulars account for the difference of Plato and Socrates is simply to say that the latter are different. This explains why one may also convince himself that he is acquainted with bare particulars . It also points up that in attempting to defend bare particulars as entities and as objects of acquaintance ontology may be reduced to triviality. We may then reject the arguments we have considered in favor of bare particulars and allow a description which employs a property like L to indicate Socrates. For objects, one need not recognize numerical difference as distinct from conceptual difference . For what makes Socrates differ from Plato may be held to be a quality occurring in his description. In this context, we might note something about numerical difference. Suppose in order to reflect that relation we introduced into a language schema a predicate ' D ', as a primitive predicate distinct from '=F ' , the latter being defined in the Russell-Leibniz fashion as · ---- (F) (Fx Fy)'. Whatever else the property D would involve it would be such that (8) '(x) (y) [x =F y :J Dxy]' would be true. Otherwise one would have conceptually different things being numerically the same. But it is difficult to conside r what that would mean . I cannot. This is part of the peculiarity about 'D'. It is, supposedly, a primitive predicate, yet (8) is hardly an empirical generalization. Hence (8) becomes either a synthetic a priori truth, a partial or implicit definition, an addition to the analytic

=

252

statements of the language, or what have you. In short it becomes some form of necessary truth. Consider two things that are numerically different , a 1 and a 2 • a 1 will not be numerical ly different from itself, hence there will be a context involving the property D that disting'uishes a 1 from ai - That something does not differ numerically from itself will , I take it, be a •necessary' truth like (8). Hence, it will "follow' from two things being numerically different that they are conceptually different. For , to put it paradoxically, D is a concept among concepts. Therefore, (E) '(x) (y) [Dxy � x y]' will also be a kind of necessary truth. [Just as "follow ' , above, reflects a kind of "inference·. ] Thus · D' and · * ' are , in some sense, "logically' equivalent notions. All this points up both the redundancy and peculiarity of · D' , as a primitive predicate , and why numerical difference is represented in the language schema , not by · D' , but by the occurrence of different signs for different things. Numerical difference, like conceptual difference, is said to be " logical' , but in a different sense. The relation expressed by · * ' is logical in that it is defined in terms of logical signs-connectives, variables , and quantifiers. Numerical difference is logical by analogy to predication. The latter is reflected in a language schema by a syntactical device , say j uxtaposition and the type distinction , to show that what is reflected is a structural or ontological tie rather than an ordinary relation and to acknowledge and avoid the puzzles associated with Bradley. But the problems about · D' are not on a par with Bradley's puzzles about predication. Any ontology must acknowledge some connection or tie , corresponding to predication , and distinguish it in kind,to avoid Bradley's problems , from what it connects or ties. Numerical difference is another matter. To put it cryptically one doesn't need either it or bare particulars in the way in which one needs a tie like predication. Hence the peculiarities about 'D' reflect , not a general or logical feature of ontology , but puzzles arising from a particular solution to the problems of ontology . This may be ignored by assimilating numerical difference to predication as a basic logical or structural or categorial feature of reality. To the view that Socrates and Plato are composites of qualities , including relations, it may be obj ected that this forces one to acknowledge that a thing changes when what it is related to changes. In short one is involved with internal relations. Thus if L is included in the set of properties that constitute Socrates, the replacement of Plato by a red square would mean that Socrates was no longer what he was. Yet, if one is talking about phenomenal things he may well , for a variety of reasons , hold that such things neither change nor persist through change. That Socrates does not persist through a change - the disappearance of Plato ­ would be perfectly in keeping with this contention . Even if one does not restrict himself to phenomenal entities, but rej ects continuants, the point

*

253

would be the same. Actually, the bare particular analyst faces the same problem in his own way. Given that Socrates and Plato are bare particulars and that the latter disappears then to hold that Socrates persists through time forces him (1) to acknowledge that qualities, relational or otherwise, are attributed at a time rather than simply predicated, (2) to hold that the persistence of Socrates is not to be taken literally since another particular exactly like Socrates is what is to the left of the red square after the change, or (3) to divide, but not reduce, Socrates into parts coincident with the period before and after the change ; this would involve there being three bare particulars - Socrates and his temporal parts - related in certain ways. (1) and (3) introduce continuants. (2), in rejecting continuants through relational change , reflects the point about internal relations. Here we are not concerned to explore the issues surrounding (1), (2), and (3) on a bare particular analysis. The point simply is that the same sorts of problems face such an analyst. Further, certainly a variant of (1) is possible on the analysis of Socrates into a composite of qualities if one makes use of predication at a time. Unlike (3), however, a composite, as opposed to a bare particular , is not literally the same composite through a change of a constituent.But the peculiarity of (3) reveals the problem of acknowledging continuants on a bare particular analysis. The issue is then whether there are continuants through relational change.If one is convinced that there are no continuants at all, then the analysis of Socrates into a composite of qualities including relations would cause no anxiety.If one is convinced that there are continuants, then perhaps he will argue for bare particulars on such a ground. But this is to base his contention on arguments differing from those we are rejecting in this paper. A proponent of bare particulars may raise two further objections to the view we are considering.First, he might hold that the use of descriptions begs the question since (1) zero level variables are employed, and (2) the connection of predication is used. But , while zero level variables are used, composites of qualities not bare particulars are, as some say, the 'values of the variables' . And, the linguistic relation of predication need not reflect the ontological connection between a bare particular and a quality.What it reflects is the ontological connection between qualities in a composite. Further, it is used to specify a member of that composite in a true sentence about it.Second, he might point out that to hold that things are composites of qualities is to turn predication between zero and first level signs into something different from predication between first and second level predicates. To say that green is a color is not to say that green, a simple, is a composite containing color as a constituent.Yet one might wonder if exemplification between a bare particular and a universal is the same thing as exemplification between two universals.It 254

is the same in the sense that it holds between simples in both cases and that in both cases the simples are thought to combine into facts. It is different in that the simples are of logically different types. Moreover, only the substrata and first level universals are held to be constituents of the things we started out to analyze. This crucial difference indicates that exemplification , on the substrata analysis, does not function uniformly. The question invites detailed analysis. The point here is that it is not obvious that the bare particular analysis requires fewer ontological ties. If that analysis does require fewer such ties then this would be a point in favor of it. But this is a different argument for bare particulars than those we are rejecting. A definite description need not specify all the properties of what is described, only those sufficing to individuate. One might feel that it should mention all , eventually, since then the descriptive term will reflect, in language, what sort of entity is being described- its structure so to speak. For, it will show that it is a composite and what it is composed of, j ust as, on a bare particular analysis , a proper name referring to a bare particular reflects, in language, that it is and what it is - a simple particular . Thinking of the predicates in a description as furnishing a meaning for the descriptive phrase, one then thinks the addition of predicates to a description reflects the growth of meaning. Including relations of all kinds can lead one to hold that to know what a thing is , i. e. , to know what the term indicating it means, is to know everything. Hence including relational properties in the descriptions of things leads one to Bradley's holistic Absolute. Also , one is led to the complete redundancy of all statements about something when using such a comprehensive description. Alternatively taking a bare particular as the meaning-referent of a proper name avoids such redundancy. But one must not be misled by the way this matter is put. A fulfilled description indicates or is connected with a thing by means of the predicates in the description. We can then distinguish two aspects of a description: first, its indicating role and , second, its meaning in the sense in which the latter is specified by the predicates it contains. In view of the second one may be led to hold that of two descriptions of the same thing the one that specifies more properties provides more meaning. But all this could mean is that since one answers the question 'What is this?' by listing properties a more complete description provides a ful ler answer to that question. If we keep separate the questions (a) 'What is it?' in the sense of what are its non­ relational properties ; ( b) 'What is it?' in the sense of all the things one can say about it ; (c) 'What is it?' in the sense of what properties need be specified in order to individuate it ; (d) 'What is it?' in the sense in which the answer is simply 'a composite of qualities' ; (e) �what is it?' in the sense in which one asks 'What is being indicated?' and in which the answer

255

might be simply 'this' ; then no harm is done. Mixing different senses of this'question' can lead one, on the one hand, to invent bare particulars, and, on the other, to Bradley's absolute. That one could indicate a complex of qualities by the term 'this' or a proper name does not mean that there need be an additional simple entity which is what is really named. To think so is to accept uncritically the principle that a simple term must indicate a simple thing. One who feels forced to indicate a composite of qualities by a complex sign such as a definite description also, in a way, accepts this principle. But such a principle is not bound to the view that things are complexes of qualities. One may on rejecting this principle indicate such complexes of qualities by proper names - signs with a purely indicating function. This is another matter, though we may note here that it is an alternative that could avoid the consideration of relational properties in the complex that constitutes the thing. 1 1 (Two complexes of non-relational qualities would just be different as bare particulars are just different. ) One might suggest that both the bare particular analyst and the holist accept the principle that language must picture reality in an extreme form. Just as one invents a simple thing to correspond to a simple term, the other holds that the corresponding term must reflect all the complexity of the complex thing. Hence, in order for a description to indicate a composite of qualities it must indicate all of them, including relations. As our knowledge of a thing grows what the thing is is revealed by the term that indicates it - by its meaning or definition, which is specified in terms of its constituent predicates. We then arrive at absolute idealism and the notion that all statements about a thing are redundant or analytic. All that is then left to be said about " what it is' is that it is what it is -'Reality is reality' . This illusion and that of bare particulars are opposite extremes of the same kind of mistake. Kenneth Barber has recently argued against themes in this paper and in my article 'Moore and Russell on Particulars, Relations , and Identity'. Barber argues, rightly I believe , that the appeal to bare particulars (substrata) to resolve the purported problem of individuation is specious. 1 2 He also argues, wrongly as I see it, that two alternative resolutions I had considered, ( 1) the appeal to relational properties and (2) the use of relations under a ·reconstruction' of the problem, do not succeed. Since his paper is admirably clear, it is not difficult to restate his two arguments. Consider a domain limited to two white squares that we presume to have no further non-relational properties. Let the domain be construed in terms of a visual field so that we may also take one object to be to the left of the other without any question arising about a reference point. We then have Wa, Sa, Wb, Sb, aLb 256

characterizing the relevant facts . Let us assume that, to resolve the problem of individuation, objects like a and b must be analyzed in such a way that some constituent of each grounds their numerical difference. For this paper I may ignore questions about the meaning or cogency of the italicized terms and phrase. Barber's dispute with me does not depend on such matters. If one assumes the above condition - call it (1) - and rejects bare particulars or pure individuators, he will ( 1) naturally look for some property to be a ground of difference and (2) take a and b to be complexes of constitue nt properties. In our limited domain W, S, and L do not enable one to resolve the problem as W and S are common to a and b and L is not, as matters stand, a viable candidate for a constituent property. I had suggested taking " (3y) yLx' and · (3y) xLy' as monadic predicates standing for relational properties, as distinct from relations. To be clearer, let us use an abstraction device ' " ', borrowed from Russell, so that · (3y) yLi , ' (3y) iLy' , and ·xLy' stand, respectively , for the two monadic properties. right-of-something and left-of-something and the relation left-of. We assume, then, that there are relational properties, like (3y) yLi, in addition to the relation iL_y. Given such an assumption, we can take a and b to be, respectively, complexes (bundles) of properties, where the corresponding sets of constituents will differ in that one contains (3y) yLi and the other contains (3y) iLy. Neither will contain .i:Ly. Barber raises two objections , though he does not clearly separate them, to this gambit. First, he dismisses the ·creation' of properties by the formation of predicates like '(3y) iLy'. Second, he holds that there are no such properties since the fact that a has (or exemplifies) (3y) iLy is the fact that a stands in iLy to b. Hence the constituents of the fact are a, b, and iLy, and not a and (3y) iLy. Suppose, in the place of the above gambit, one rejects (I). He may then hold that to account for the difference between a and b it suffices to point to a difference that holds between them, such as that - aLb but neither that - bLa nor that - bLb. This assumes that our concept of an object is such that a and b are acknowledged as different if it is acknowledged that some condition obtains for a but not for b . Needless to say, one does not prove that a differs from b without assuming, or starting from, for example, 'aLb' and ' -- bLa' , '--aLa' , etc. as true. Barber objects to this second proposal by claiming that I presuppose thatiLy is irreflexive. He holds that iLy being irreflexive is 'synthetic' and that "aLb' and 'a=/:- b' provide a confirming instance of such a claim expressed as '(x) (y) (xLy � x=/:-y)'. Therefore, a Lb does not ground the truth of 'a=/:- b' and hence does not account for the difference between a and b . This second line of criticism rests on an obvious misunderstanding. My claim was that, given that we use the term 'object' in conformity with the principle of the identity of indiscernibles, where the latter is understood

257

in terms of '(x) (y) [ (3f) (Jx & -fy) :=) xi: y] ', then, trivially, we can take 'aLb' and '--aLa' as true only if we admit that 'a i= b' is true. Barber makes two mistakes. First, he mistakenly insists that from 'aLb' it must follow that 'a -:f=. b' , if 'aLb' is to ground the difference of a from b . But , if it so follows, then '(x) (y) (xLy :=) x-:f=.y)' is taken to be 'analytic' and not 'synthetic' . He holds the latter is synthetic, not analytic. Hence no relevant implication holds. Therefore, no ground or account is given. There are obvious problems, in connection with ' (x) (y) (xLy :=) i=y) ' , with his use of 'synthetic' and the implicit claim t hat all synthetic generalities have confirming instances. But we need not probe into such matters, for the crucial claim is not that xLy is irreflexive but that '(x ) (y) [ (3f) (Jx & --fy) :=) xi=y] ' holds and that 'aLb' and '--aLa' are true. Not realizing this is his second mistake. Of course, from t he last t hree mentioned sentences 'a -:f=. b' does follow. But ' (x) (y) [ (3f) (Jx & --fy) :=) x-:f=. y ] ' is not taken as 'synthetic' in Barber' s sense. One does not confirm it . One takes it to be involved in the notion of (analysis of) an obj ect . All Barber does is repeat what I had admitted , that one cannot "get ' 'a i= b ' from 'aLb' . Barber's first criticism is more interesting, and to respond to it requires some preparation. Russell once dismissed the so-called Wiener-Kura­ towski procedure for eliminating relations (as sets of pairs) as a trick without philosophical importance. In this he was wrong. Assume that we take properties and two-term relations to be classes: properties as classes of things (that have the property) ; relations as classes of pairs. In addition to things and classes we have recognized a further entity, a pair. Hence, one must recognize a new kind of entity. Such a "thing', relative to the obj ects standing in the relation, is complex.Alternatively, one can avoid the appeal to such things by construing the relation as a higher-type class. Thus, instead of appealing to the pair (a, b) , we might consider the class whose members are a and b . A relation would t hen be a class of such classes. Corresponding to the monadic predicate " G x' we would t hen have a class of obj ects, including, say, a and b . Corresponding to the dyadic predicate 'xLy' we would have a class of classes. What the Wiener­ Kuratowski procedure reflects is the difference between taking a property 'in extension' as a class and taking a relation in extension. To put it another way, relations cannot sensibly be taken in extension if the relevant class is to contain the obj ects that stand in the relation. One must introduce a new entity, a pair, or 'ascend the types'. In either case t he members of the class are not the related t hings. 1 3 This in turn reflects an old philosophical claim. Relations are not exemplified in the same sense that properties are. The analysis of the concept of an obj ect is terms of the notion of a property complex, or bundle, represents such a point if, on the analysis, one also holds t hat 258

relations cannot be constituents of such complexes. Such a one will naturally be led to hold that relations, not being constituents of the complexes, are exemplified by the latter. Thus sentences like 'Ga' and � aLb' , if perspicuously presented, will reflect this difference in the kind of predication and not merely reflect the number of terms involved. , Properties are not simply " one-term relations. If we represent a property bundle (obj ect) by ( 1 ) C/G,H, . . . ) where " C 1 , represents a combining function (nexus, tie, compresence relation), we distinguish C1 from the connection between a property and the complex obj ect. Thus, given an obj ect, as in ( 1 ), we may say that such an obj ect contains G. But contains is not Cr Moreover, neither C 1 nor contains is the predicative connection expressed in 'aLb'. The Wiener­ Kuratowski procedure is a formal reflection of the notion that monadic and relational predications are fundamentally different . It also reflects the idea that relational predication presupposes the exemplification of nonrelational properties. It is clear, on the bundle view, in what sense relational predication presupposes nonrelational predication, while the reverse is not the case. The need to appeal to a further entity, a pair, or to ascend the types weighs against the taking of properties and relations in extension, j ust as the points involved are reflected in , and hence support, the bundle analysis of an obj ect. This latter analysis is in direct conflict with the taking of properties in extension. For obviously one cannot construe a property as a class of particulars, while taking a particular to be a bundle of properties, even though bundles, are not classes. While obj ects, as bundles, are construed not as classes but as complexes of properties, such complexes have been assumed to share a basic characteristic of classes. N o two can have all constituents in common, as no two classes can have all members in common. Barber challenges this assumption, as I had challenged it. But for purposes of developing a point, I had assumed in my earlier paper that it was a condition to be met. Hence in the case of our simple domain one must 'find' a further constituent property belonging to one obj ect but not the other. I assumed , further, that predicates like '(3x) xLy' are to be taken as standing for properties. Thus, as one might put it , complex properties are introduced. I spoke, in my paper, of defined predicates standing for properties, since I introduced the shorthand expression 'R 1 x' to abbreviate ' (3x)xLy'. But the issue is not about the creation of properties by defining signs. It is about whether the expression ' (3x) xLy', which is not an eliminable expression, stands for a property. Assuming that it does, we can take a and b to have different

259

constituents. Barber's first objection is thus not an argument, since I did not suggest that we'create' properties by definition but suggested that we recognize that certain expressions carry'ontological commitments'.His second argument is an argument. In it he claims that to predicate '(3x)xLy' of b is to assert ' (3x)xLb' and, more emphatically, if we take 'aLi' as a predicate, to predicate it of b is to assert 'aLb' . The latter sentence is true, given the fact that -a Lb . That fact also grounds the truth of the ascription of 'aLi' to 'b' . Obviously, since in both cases 'aLb' expresses the claim that is made. In the case of'(3x)xLb' it is also the fact expressed by 'aLb' that grounds the truth of the existential claim. Hence there is, if I may so put it, no monadic fact. But if there are properties corresponding to the predicates '(3x)xLy' and 'aLy' , there should be such facts. Hence there is a dilemma. In the case of 'aLy' being ascribed to b , either that is not equivalent to claiming that 'aLb' holds or there must be a further fact.In neither case can there be a further fact, since in the one case, to ascribe 'aLy' to b is to assert 'aLb' while, in the other case, the fact that aLb suffices to ground the truth of'(3x)xLb'. So runs Barber's argument. I had explicitly assumed that we take predicates like'(3x)xLy' to stand for properties. Such properties, on a bundle view, are taken to be constituents of objects. Hence, on such a view, such properties are not exemplified by objects, as relations like 'iLy' are exemplified by'pairs' of objects. In short, on a bundle view, there are no monadic facts on the order of facts like that -aLb . An object, represented by a sign complex such as (J ), is construed as a complex that is categorially different from a complex such as a fact.We can consider each constituent property to be contained in, or be a constituent of, the object and take such claims to express further facts. Then, given that (2) Ci W,S, (3x)yLx) depicts the object a , in a sense it follows that ' (3x)aLx' is true. Likewise, given that 'aLb' and '(3x)aLx' are true, it follows that a relational property, (3x}yLx, belongs to a - on the kind of analysis in question.By assuming that the only fact involved is that expressed by 'aLb' , Barber offers no argument against the bundle view as construed above.For, he merely rules out the claim that (3) (4) (5) (6)

260

b contains (3x)xLy R1 b (3x)xLb aLb

are used to make different (though logically connected) assertions in the context of a bundle view. Of course one can claim that it is a defect of a bundle view that it recognizes both a relational fact ' that -aLb ' and a complex object containing a (monadic) relational property. For there is a redundancy of truth conditions, or, to put it another way , there are necessary connections between such conditions. But this is not surprising, given the recognition of relational properties in addition to relations. In this connection we should also note that when a predicate like ' (3x)yLx' occurs in subject place it cannot be replaced by occurrences of the relation predicate 'xLy' (or 'xLy'). For example consider the sentences (3f) (f= (3x)xLy) and

One may say that such properties enter into some facts as subjects and not as exemplified by objects. On a bundle view relational properties like (3x)xLy and (3x)yLx are taken to be subjects of a combinatorial nexus that serves to combine properties into particulars. Given the analysis of a bundle view and the recognition of relational facts like that -aLb, such facts are not logically independent of an object, a, containing a monadic relational property. This is a price the bundle theorist must pay for the recognition of relational properties. It is not a surprising price. Barber's argument reduces to the claim that the price is not acceptable. Notes 1 2

3

4

Bertrand Russell An In quiry into Meaning and Truth (London : Allen and Unwin, 1 956) , pp. 97-99. Arguments for particul ars occur in G. E . Moore , ' Identity, ' Proceedings ofthe Aristotelian Society, vol . 1 ( 1 90 1 ) and Bertrand Russell , 'On the Relations of Universals and Particulars , ' reprinted in Logic and Kno wledge, ed. R . C. Marsh (Allen and Unwi n , 1 956) . Moores's parti culars are quality instances and Russell's may also be , but the arguments are essentially the same as those for bare particulars that later are suggested throughout G. Bergmann , Logic and Reality ( Madison : University of Wisconsin Press , 1964), and in essays by E. All aire and R. Grossm ann in Essays in Ontology (The Hague : M. Nijhoff, 1 963 ) . The description i s ( ,F 2 )((W 1 EF & 5 1 EF& P 1 EF2] & ((f1 ) [(f EF2) = ((f = W1 )v(f1 = S 1 )v(f = P 1 ) ) ] ) . For purposes of th is paper 'P-'can be read as either a second l evel class or property sign and ' E ' as either the class membership sign or the predicative 'is' . Sometimes '(a) ' and '(�) ' will be used to refer and sometimes to abbreviate the description and set sign .

261

5

Class membership is not significant for ontological questions. Part of what is involved we have just seen. That Russell's view can be stated in terms of properties rathe r than classes does not affect this. With 'Socrates' as a second order property defined as in n. 3 , predication would not reflect any ontological tie, since all ascriptions o f properties to Socrates would become disjunctive identity statements that are either analytic truths or falsehoods. Where a predication is so transformed one may say no ontological connection is involved as no connection among entities is required for the statement to be true. 6 One might suggest a second order property of 'co-exemplification' that the first order properties exemplify when there is an individual thing. This, however, would mean that Socrates was no longer a class of qualities , but qualities in a certain relation. Such a relation would then furnish the ontological tie and hence turn the view into one we consider later . This is disguised since such a view would also require an exemplification tie so that the qualities could exemplify the relation of co-exemplification. Moreover, predication would no longer be trivial as in n. 5 unless, of course, one defined 'co-exemplification' by extension. Russell problematically spoke of 'compresence' in this connection see note 1 1 . 7 For a claim of acquaintance with substrata and the ontological tie see G . Bergmann. op. cit. , pp. 47-48. 8 Russell , in keeping with his formula that a description is an incomplete symbol, would probably not accept this formulation. 9 The sense in which these formulations say the same thing is like the sense in which '2 + 2 = 4' and '7 + 3 = 10' say the same thing. Also with respect to intentional contexts none of these statements say the same thing as any other one. 1 0 G. Bergmann, ' Russell on Particulars' , reprinted in The Metaphysics of Logical Positivism (New York : Longmans , Green & Co . , 1 954) . 1 1 What I have called 'Russell's view' is not, literally, what he proposed, since he treats the sign 'Socrates' as an undefined sign ( a proper name) to avoid turning sentences like 'Socrates is white' into analytic truths. But as he analyzed obj ects into classes of coexistent qualities and did not recognize the need for ( or role of) an ontological tie, his use of names for classes is specious. To argue this point is beyond the scope of this paper, though part of what is involved is touched on in notes 5 and 6. The use of proper names for objects considered as complexes, but not classes, of qualities is discussed in 'Things and Qualities' in Metaphysics and Explanation , ed . Merrill and Capitan (University of Pittsburgh Press , 1966) . 12 Kenneth Barber , 'On Representing Numerical Difference' , Philosophical Topics X, 2, pp. 93-1 03. 1 3 I ignore the question of order and hence ordered pairs. On this matter see ' Logical Form, Existence , and Relational Predication· in this volume , pp. 204 f.

Universals , Particulars , and Predication

One sees two white square patches. Assume for simplicity, first, that we are talking about phenomenal. rather than physical, things, since the phenomenalism-realism question will not concern us, and, second , that the patches have no other non-relational properties. Call them Plato and Socrates. A metaphysician may then ask ( 1 ) what are they composed of­ wh at is their ontological analysis? (2) what are the qualities that we .attribute to both of them? (3 ) h ow are the qualities connected or related to the things? Consider three alternative solutions. One, call it (a) , holds that the patches are composed of particulars or substrata and universals (squareness and whiteness) combined in the structural tie of exemplification. I n so answering (1 ) , (a) also answers (2) and (3) by asserting that properties are universals and that they are connected to the substrata by the tie of exemplification. A second solution, call it (/3) holds that the patches are composites of universal qualities. In so doing it answers (2) and (3) by holding, with (a), that qualities are universals but, unlike (a) . that the universals are related to the things in that they are parts of them combined together by a structural tie which we shall call combination. Combination, as opposed to exemplification, combines qualities into things: exemplification combines substrata and qualities into both facts and things. Hence (a) and (/3) disagree about the answers to (1 ) and (3). These disagreements constitute the subj ect of this paper. What we shall see is, first, that there are radical differences between the arguments that lead to the acceptance of particulars and those that lead to the acceptance of universals. These differences will reveal the former to be specious. Second, we shall consider differences between the ontol­ ogical ties on (a) and (/3) . Finally we will discuss relational proper­ ties in connection with (a) , (/3) and a third view , (y), which considers things to be combinations, not of universals, but of instances of qualities. B oth (a) and (/3) agree that there are universals - that qual i t ic'.', a re universals. To say that the quality white is a universal is to say, in part, that one and the same thing is connected in some way to both Plato and Socrates and accounts for the truth of the sentences 'Plato is white' and 'Socrates is white'. To put it another way, the term 'white' in both sentences refers to the same entity. What arguments are there for such a view? Russell elegantly put forth the classic argument in ' On the Relations of Universals to Particulars' . 1 To deny universals is to assert 263

that the quality attributed to Socrates is not one and the same with the quality attributed to Plato.The quality 'in' each is numerically distinct and, furthermore, no one thing accounts for these distinct qualities being of the same kind.[I mention this latter point since one might, on a version of (y), hold that there are particular qualities as well as universal qualities . The former account for the whiteness of particular patches; the latter for such particular whitenesses being just that. ] One must then hold that such particular qualities are related in some way, since they are the entities in virtue of which we truly assert that both things are white.One must then specify such a relation. The obvious point is that such a relation will either be taken as a universal or a particular instance. If a particular then the original problem recurs when we introduce a third white patch or a pair of black patches. If admitted as a universal then the view finally accepts universals, albeit relational ones. No other alternative can answer the original question - to account for Socrates and Plato having the same color. That is, no alternative acknowledging only individuals that denies the two patches are connected, in some way, to one and the same entity - can prevent the recurrence of exactly the same kind of question we started out with.Let us consider the case of particulars. One may argue that just as universals account for the sameness of quality, something must account for the difference of two patches which, conceivably, have all their nonrelational qualities in common. 2 This something, the ground of numerical difference, is considered to be a substratum which stands in a unique relation or connection with universals to form or constitute facts and the things we started out with, Plato and Socrates. These ordinary things are thought of as composed of a substratum and universals connected together. Facts about such things may be composed of the substratum and one universal . The facts are about the things since the same substratum is a constituent of both sorts of entities. Such substrata account for the difference of the two patches; for there being two and not one thing. These substrata, in turn, are held to be simply different. At this point one may balk. If substrata are held to be simply different , why bother with them at all? Why not hold that Socrates and Plato are composites, not classes, of universals, and that they are different composites of the same universals? They just simply differ. If substrata can simply differ, why may not composites of universals simply differ? The proponent of substrata must retort that since Plato and Socrates are composite entities they cannot simply differ but must be held to differ in a constituent. Only simple entities can simply differ.Let me call this assertion the axiom of difference . The first point to note is that it is a necessary assumption in the argument to establish the need for substrata in an adequate ontology. We must then inquire why some accept, implicitly or explicitly, such an axiom . 264

One reason for adhering to the axiom of difference might be the belief that in order to 'account' in one·s ontology for a certain feature of experience one must introduce an entity of some kind. Thus as universals are said to account for the experience of qualitative identity, substrata account for the experience of numerically different things. Further, an ontological tie like exemplification accounts for the existence of facts and for the experience of one and the same thing having several qualities. Not to so locate or account for features of experience is, on this view, not to engage in ontological analysis. Of course one cannot argue with stipulations. If it is stipulated that the difference of Plato and Socrates must be accounted for by an entity in each that is simply different from any other entity of the same kind, and whose difference in turn need not be accounted for, then nothing more can be said. Yet one might point out that considering things like Plato and Socrates to be complexes of universals also locates, in one's ontology, an existent that may be taken to account for the difference of things that we notice. The proponent of substrata might reply that this provides no such account since the thing then accounts for itself . That is, one does not account for the difference of two things by holding that they are simply different combinations of the same universals. This is just to assert that they are different and not ·account for' the difference, since one does not hold them to be different in virtue of different constituent entities. Adhering to such a conception of ontological analysis one is naturally led to assert the axiom of difference . Yet why does one hold to such a notion of ontological analysis? One might argue that to hold that a complex of universals accounts for or grounds difference is to abandon the idea that the existents of one's ontology must be simples, as opposed to composites. Hence, to locate an existent to account for difference must be to locate a simple entity or element. But this is not an argument since it merely restates that a simple existent must account for difference, and we are concerned with why this axiom is adhered to. There is, I believe, a fundamental reason that leads the proponent of substrata to hold that the axiom of difference is necessary for ontological analysis to be significant and meaningful. Providing an ontological analysis, a metaphysician tells us what the 'parts' or 'constituents' of things are. In so doing he accounts for the 'facts' of ordinary experience. Yet there is a problem about his use of terms like 'part', 'constituent', and 'account'. Substrata are not 'parts' of things either in the sense in which a leg is part of a chair, an atom is part of the leg, a soldier is part of an army, or , for that matter, as a sense datum is held by some to be part of a physical object. How does the metaphysician use the term 'part' when he proposes an ontology like (a), (/3), or (y) ? Suppose he feels obligated to explain his use by means of ordinary terms and uses. He

265

might point out that we use 'part' in various senses , but they all h ave one thing in common : no two 'wholes' can be composed of exactly the same 'parts' . Consider the two ordered pairs [ 1 ;3] and [3 ; 1 ] . The pairs are not composed of the sam_e parts , for, in addition to the numbers , there is the ordering relation which is different . On the view (/3) two things are held to be different and yet composed of exactly the same qualities in exactly the same ontological tie or nexus . This points up what is involved in holding that the ordered pairs have different parts. Part of a thing, on an ontological analysis, is the nexus or tie that relates the other constituents . Keeping the feature of the part-whole dichotomy that requires two wholes to differ in a part lets the metaphysician use his special sense of 'part' in a way fundamentally like other, ordinary uses. Not keeping this feature might lead him to feel that his use of the dichotomy is meaningless or empty. If so, his whole enterprise is nonsensical. Thus to speak meaningfully of 'whole' and 'part' , two wholes , to be two , must differ in a part. We thus restate the axiom of difference . Using 'part' and 'whole' in a way fundamentally like all other uses, the metaphysician may then feel he need only further explain how he uses the terms referring to the entities he takes as parts of ordinary things - 'substrata' , 'universal' , 'nexus' , 'exemplification' , 'instance' , etc . One may hold that this must be done contextually , within the context of a metaphysical position and by contrast with alternatives. He may further point out that this is legitimate since such terms do not have ordinary uses . Hence no ..alternative is open . But 'part' and 'whole' , being ordinary concepts , must retain , when used by the metaphysician , the fundamental features that characterize their use . This I take to be the crucial argument for the axiom of difference . But it neither does what it purports to , nor need what it purports to do be done . It does not do what it purports to in view of the very special sort of thing the nexus of exemplification is which combines "parts' into "wholes' . Such a nexus enables a universal to combine with two distinct substrata into two distinct facts and to be a common part of two spatially distinct things . No ordinary thing is a part of two other spatially separate ordinary things . On (/3) one complex of universals is enabled to be different from another complex of the same universals since universals and the nexus of combination are the special "things' they are . Exemplification , on (a) , permits two substrata to combine with one and the same universal into two things and one substratum to combine with two universals into one thing . The point is that a nexus and ontological elements play quite special roles in each case . It is misleading to speak as if one metaphysician's use of "part' and 'whole' is more ordinary , since he does not acknowledge that two complexes may share all constituents. Ordinary analogies do not help to explicate the sense in which a substratum , a universal , and a nexus are "parts' of a white patch .

266

Moreover, one may point out that the metaphysician's use, without the axiom of difference, still corresponds in certain ways to ordinary uses of 'part' and "whole' . The relation of whole to part is (a) one-many; (b) so used with terms like "simple' and ·complex' that the whole, relative to its parts, is complex and, they, relative to it , are simple; (c) such that the parts are combined by a relation ( the nexus) to form the whole; ( d) the simples of the system can only be parts and not wholes. In short, ·constituent', "element' , ·complex' , "simple', and ' nexus' are such special notions in metaphysical analysis and so intimately connected to the part­ whole dichotomy that the latter can hardly masquerade as an ordinary distinction. An ad ditional correspondence with ordinary uses that is provided by the axiom of difference does not really add anything, and, in view of (a) , (b), (c) and (d), is not needed . Adherence to the axiom of difference might indicate that the 'class­ member' distinction dominates one's conception of part-whole. On the one hand, two classes to be two must differ in a constituent or member, and , on the other hand, several classes can have a common member as several things can exemplify one universal. Perhaps one who thinks in terms of the axiom of difference implicitly thinks of things as collections of substrata, universals and a nexus. This is suggestive if we consider that many philosophers nowadays get at ontology through language, improved or otherwise. Hence in considering complex entities and their parts one might insist that simple signs in language must correspond to simple things in ontology and complex things must be indicated by complex signs. Moreover, the simple signs in a complex sign should indicate the simple things in the complex entity. If one holds that language should 'picture' entities in this way, then one will be able to consider the constituents of two complex entities by noting the constituent signs in the complex signs that indicate them. A complex sign would provide a list of its referent' s parts. But if two complex entities had all parts in common, we would have one list and not two. Hence the complex sign could not be coherently used to indicate two things. Language would not picture reality. Insisting that it must can lead one to the axiom of difference. In any case, we have seen that the argument holding substrata to be necessary for individuation requires the axiom of difference. The argument for universals needs no corresponding assumption. One does not assume that something common must be referred to by the term 'white' when one says'Plato is white' and ' Socrates is w hite': one argues that this must be so for the two sentences to ascribe the same quality to both subj ects. Of course this means that one must refute all nominalistic gambits. But this is not done by assuming a correlate to the axiom of difference. If the term 'white' in each sentence does not refer to the same thing, then one does not say the same thing of

267

each particular, unless some further connection and entity is introduced. This is the basic theme in Russell's argument for universals. In view of this an objector might insist that ·there is a correlate to the axiom of difference, for one is assuming that to be used to say the same thing in two sentences a predicate must refer to one thing. But this, if it be an assumption, is not at all like the axiom of difference. It simply states, first, that terms must be connected or refer to entities in order for sentences to be about or true of things, 3 and second, that reference must be consistent if what is said, by two uses of a term, is to be the same. The terms 'Plato' and ' Socrates' can refer to different things and, consequently, the sentences 'Plato is white' and 'Socrates is white' can be about such different things if and only if there are two things. What, if anything, makes them different is a further question. The axiom of difference is relevant to this further issue. The term 'white' can be used to say one and the same thing, in two sentences about two things, if and only if there is a common thing that it refers to (or is connected to in some way) in both occurrences. This corresponds not to the axiom of difference but to the trivial point about two things being required for the two sentences to be about two things. What would correspond to the axiom of difference would be an assumption about the need for some additional thing to account for the quality's being the same in both cases. But no additional thing accounts for that ; to say that it is the same in both cases is to say that it is a universal. Perhaps the point we are concerned with can be made in another way. Since there are two distinguishable white patches, we have no problem connecting the names 'Plato' and ' Socrates' to their referents. The term 'white' is connected to a quality of each. We may then ask if the quality attributed to Plato is the same as that attributed to Socrates. If we say 'no' we are faced with the problem of using the same term (to say the same thing) to refer to (or be connected with) different entities. As Plato and Socrates are distinguishable, there is, to repeat, no corresponding problem. One might believe there is if he thinks that we are led to acknowledge substrata by asking 'what makes things different' just as we are led to universals by asking 'what makes things the same in a respect'. This way of putting the question leads us to assume that something must make them different. However , if we ask why we can truly predicate the term 'white' of two distinct things, we are led to universals. But, if we ask why two names can name distinct things we are led to assert, trivially, that there are two distinct things -not two distinct substrata. In effect the same point comes up in another way when the proponent of the axiom of difference insists that the difference of substrata is not to be accounted for since such entities are simples. Such simples account for the difference of complex entities, the patches, but their difference need not be accounted for in turn. The case of universals is quite different. These entities 268

account for the sameness of respect of the two patches. But to ask about a universal, what accounts for its being one, would be either to ask, nonsensically, why it is the same as itself or to inquire, legitimately, about the connection between universals and particular things. 4 While this raises a question about the nature of the nexus, 5 one need not introduce a stipulation, like the axiom of difference, to avoid a question. For while the question substrata were introduced to answer may be re-raised about them, this is not so for universals. To avoid this one can only stipulate that the question is inadmissible. This suggests not only that there is a significant difference between arguments for universals and those for substrata , but also that the problem of individuation is specious. The second issue I wish to discuss concerns ontological ties like exemplification and combination. On view (/3) we may consider a white square patch to be pictured linguistically by C 1 (W , S)

(1)

where ·c 1 ' stands for the nexus 'combination'. The linguistic picture is misleading since it appears to turn the nexus into something like an ordinary relation, but this can be ignored for the sake of simplifying the presentation. On ( a) the picture would be

(2) with "E 1 ' for the nexus uniting the substratum s to its universal qualities, W and S. For the proponent of ( a) there are facts as well as things. The relevant ones here would be those referred to by the sentences 'Ws' and "Ss'. One might then think of an alternative 'picture' of a white patch like Socrates as

Ws & Ss.

(2 ')

This shows that E 1 combines the two-fold function of exemplifying, as a relation between a substratum and a universal , and combining, as a relation that connects several facts into one thing. 6 Of course the combining role is shared by the substratums, which is the sort of thing that can exemplify several qualities. Thus we might think of two relations or ontological ties being involved. We would have E 1 which connects elements into a thing and, say, E 2 which connects a substratum and a quality into a fact. Thus we would have, in addition to (2) E 2 (s, W)

(3)

and E 2 (s, S)

(4) 269

However, one might point out that given (2), E 1 , and the categorial difference between s, on the one hand , and W and S , on the other, we can generate both (3) and (4) and, hence, E 2 . E 2 would hold between any two elements combined by E 1 , where the elements are of different ontological kinds - particulars and universals. Given this 'construction' of E2 we can derive (3) and (4) from (2). Thus one might insist that only one nexus is involved . Alternatively, one may feel that to claim that only one nexus is required due to such logical connections between the two is to think of ontological ties as being too much like ordinary relations or mathematical functions. For, even granting that to hold that only E 1 is required is to think of things l ike Plato and Socrates as complex facts, there are still facts, l ike Ws, and complex facts, l ike Socrates , which are things. Having two sorts of complex entities, even though they both are complexes constructed from universals and particulars, may well be considered to require two ontological ties. Be that as it may, ( a) at least requires both an additional entity, s, and a nexus that has a twofold function. The need for a nexus with such a complex function simply reflects that (a) acknowledges facts as entities as wel l as things. On (/3) there are no facts. When one says (5) 'Socrates is white', where 'Socrates' refers to the complex 'pictured' by ( 1 ), one claims that the universal white is part of the complex. The complex thing is what makes the sentence true , not a fact composed of two simple entities in a structural relation. Facts are no longer needed. But in the sentence, predication no longer neatly reflects a nexus as it appears to do on (a) . On this latter view we have . in (3 ) for example, the sign 's' corresponding to the substratum, the sign "W' to the universal and 'E 2 ' ( or juxtaposition of 'W' and 's') reflecting the nexus. Three linguistic aspects correspond to three ontological entities. In virtue of this correspondence the sentence , as a whole, corresponds to the fact made up of the three ontological entities. By contrast, if we transcribed (5) on view (/3) , with 'C2 ' for 'contains', as C2 (Socrates, W)

(5 ' )

the sign 'C2 ' would not correspond to any ontological entity or feature. Predication in language truly holds between the sign 'Socrates' and the sign 'W' , since ·Socrates' refers to a composite of qualities that are combined by C 1 and that composite includes W. The sign 'C2 ' would be no more than a linguistic shadow of the nexus C 1 • C 1 only has the function of combining elements into things, not elements into facts. Hence its sole function corresponds only to one of the functions exempl ification performs in (a) . But it is in vi rtue of this single function that some sentences are true and others false. On (/3), however, language no longer neatly pictures what that view holds to be " reality' , but then the ontol ogy

270

has not been constructed on the linguistic model of subject , predicate, and predication. Furthermore, the ontology of (/3) is not so complex in that the ontological tie only has a single function. 7 We may complicate the issue by acknowledging universals of the higher types, color for example. With ·col2 ' for 'color' , consider the true sentence Col2 (W) .

(6)

Clearly, since W is held to be a simple entity we do not then mean that color is a constituent of W as W is a constituent of Socrates. Hence must not (/3) introduce a further ontological nexus for such a case, corresponding to exemplification in (a), and thus also introduce facts? Before discussing this let us see how (6) would be analyzed on (a) . Recall that to avoid having both E1 and E2 on (a), we considered the possibility of constructing the latter from the former, of deriving (3) and (4) from (2) . But clearly no such derivation is possible from something like (2) to something like (6' ) Hence, one must acknowledge E1 , which ties first level qualities and substrata into things like Socrates, and E2 , which ties things and qualities of different levels into facts. Further, only some of the latter facts may then be constructed from E1 . One could introduce a further connection, E3 , to unite universals of different types into facts. Then we could hold that all E2 connections were constructable from E 1 , but not E3 connections. I n either case two basic connections would be involved, since not all cases of the one would be derivable from the other. This reflects a point many philosophers have expressed, though in quite different ways: predication in 'F(x)' is different from that in 'F2 (f)'. The issue can perhaps be clarified in the following way. Suppose, on (a) , one decided to include secon d level universals, say color and shape ('Sh2 ' ) in the composition of Socrates. Then, instead of (2) we would have E 1 (s, W, S, Col2 , Sh2 )

(2' )

picturing Socrates. One could not then specify, simply in terms of kinds of elements, which entities went to make up facts, for this would not serve to differentiate between 'Col2 (W) ' and 'Col2 (S) ' . One could of course different iate between kinds (not types) of universals an d reflect this difference syntactically, j ust as the differences among 's ' , 'W', and 'Col2 ' syntactically express the differences among particulars an d universals of

27 1

different types . But this would simply provide a linguistic smokescreen to cover up the fact that something other than E 1 is needed to generate 'E 2 (Col2, W) ' from (2' ) . Besides, considering Col2 as a constituent of Socrates is already a bit strange , since color is not a property of Socrates (or of s) . If there are universals of the higher types a proponent of (/3) must also acknowledge a nexus in addition to C 1 • This second nexus will connect universals of different types into facts . But this does not mean that ( a) is a simpler ontology than (/3) . For not only does (a) have a nexus with a two­ fold function , E 1 , and bare substrata , but ( a) must, to deal with universals of the higher types, introduce a further nexus and , hence , recognize two kinds of facts . If, in attempting to avoid this, the proponent of ( a) considers properties like color and shape as constituents of Socrates i n order to generate E2 (Col 2 , W) and E 2 (Sh 2 , S) from (2' ) then clearly the proponent of (/3) can do something similar by considering something like 'C 1 (W , S, Col 2 , Sh 2 ) ' to replace ( 1 ) . We must not be misled by an apparent difference between (a) and (/3) . The adherent of (/3) acknowledges that the sentence 'Socrates is white' is true since one is a constituent of the other , but that 'White is a color' is true , not since one is a constituent of the other, but since one exemplifies the other. The proponent of ( a) holds that both are true in view of an exemplification connection that gives rise to facts. But speaking of exemplification in both cases covers up the fact that two ontological connections are needed. Nor will it do to hold that at least on ( a) they are of the same 'sort' , exemplification , while on (/3) one is a part-whole relation and the other exemplification . This , again . is to be misled by linguistic and not ontological differences between the two views . On (/3) a nexus holds between the constituent qualities of things , it does not connect a part with a whole , but several parts into a whole . It holds only between universals as ( 1 ) shows . The second nexus also holds among universals , but between universals of different types such as Col 2 and W . Hence one nexus combines universals of the same type into individual things , the other combines universals of different types into , say , facts . Speaking of part and whole enters only when we consider the linguistic relation of predication between the subject and predicate signs of the sentence 'Socrates is white' . The predicate refers to a part of the complex entity named by the subj ect term . On (a) both sentences , 'W(s)' and 'Col 2 (W) ' , are true because of connections between the simple referents of the subj ect and predicate signs . But the connections are different , as we have seen . Hence (a) has two kinds of facts , one kind corresponding to 272

each kind of sentence. We may conclude that the existence of higher level qualities does not provide an argument for (a) on the ground that fewer ontological ties are needed. The proponent of (y) is forced, even without higher level qualities, to recognize an additional nexus. On the basis of the argument for universals ( y) must acknowledge relational universals to account for instances of whiteness all being whitenesses. Not being related to one universal entity, they simply relate to themselves by a relation of similarity, or some such thing. Socrates is then a construction from an instance of whiteness combined with an instance of squareness. Call these entities ·w 1 , and ' s1 ' , respectively, and the corresponding , but distinct, instances in Plato "w2 ' and " s2 ' . Let " C* ' stand for the nexus that combines such entities into the complex things - the patches. We may then picture Socrates and Plato respectively by

and

Since w1 is related to w2 in some way, but not to either s1 or s2 , and as s 1 has that relation to s.:, , we must acknowledge not only the relation but the exemplification of it by the pairs of instances. Thus in addition to the instances and the relational universal, R, we must introduce a further nexus, E* , and facts pictured by

and

Just as C* may be said to be a ·combining' relation, E* may be spoken of as an ' exemplifying' one. It would m ake no difference if one sought to identify R, the similarity relation, with the nexus E *. The point is, an ontological tie is needed in addition to C* . This complicates (y) beyond that of (/3) . With respect to (a) , however, the proponent of (y) might point out that E1 is an ontological nexus with a two-fold function, exemplifying and combining, so to speak. For on the one hand it permits a substratum to connect with one universal to form a fact, and together with the substratum it enables several facts to be about the same thing . Alternatively, a proponent of (a) may insist that on (/3) the nexus C1 not 273

only combines but serves to individuate and, hence, has a two-fold function as well. This returns us to the argument about the axiom of difference. But even if C 1 be a nexus with a two-fold function, E 1 is also. Moreover, ( a) involves a further ontological kind, substrata, and the two-fold function of E 1 gives rise to two types of entities, simple facts and complex facts (things). C 1 , by contrast, only combines qualities into complex things. Thus far we have confined the contrast of (a) and (/3) to nonrelational qualities. Suppose the sentence ' Socrates is to the left of Plato' is true and that proponents of both (a) and /3) hold that 'left of' is a simple relational universal. A proponent of (a) might then argue that on (/3) one must either introduce a further nexus to account for the truth of relational sentences or include relational qualities as constituents of things like Plato and Socrates. Assume the second alternative will not do. 8 On (/3) one must then hold that a relation like 'left of' (hereafter L) is exemplified by two complexes of non-relational universals, Socrates and Plato. There mustthen be, in addition to C 1 , a tie, say Ex, that connects L, Socrates and Plato into a fact (or whole other than a 'thing'). This fact may be pictured by

Ex (L, Plato, Socrates) . On (/3) we then have two ties and relational facts as well as complex things. But the proponent of ( a) must do something similar. For if he omits L from the analysis of Socrates and Plato, from (2) , he must introduce a tie in addition to E 1 to combine substrata and relational universals into relational facts. Again, ( a) has no advantage over (/3) , and relational universals , no more than universals of the higher types, need lead us to bare substrata. The proponent of (a) gains nothing by holding that the nexus which gives rise to relational facts may be identified with the one which connects universals of different types. For the adherent of (/3) may make a corresponding claim. The handling of relations on (/3) suggests some points in connection with classical arguments about the 'reality' of relations and the distinction between essential and accidental properties. In that relations are not included as constituents of things one might take this to be a significant difference between relational and non-relational qualities. This may be taken to explicate why some hold that relational qualities are not existents. It is interesting to note that Moore held that properties of complexes of qualities were non-existent (non-natural) 'adjectives' . 9 But as some relational qualities are held to be simple elements in one's ontology they are, in that basic sense, existents. Not including relations as constituents of things one holds them to be 'external' in one clear sense of

274

that term. Taking them to be external in that sense, we note a further difference between relational and non-relational qualities. It follows from a things·s being a composite of qualities that it cannot change in one of its constituent qualities and be one and the same thing. This is involved in the notion of "thing· on (/3). But it does not follow that a thing cannot change relations and still be one and the same. All relations are in this sense ·accidental' as opposed to 'essential '. This connects 'accidental' with ·external' and 'essential' with 'internal'. Extending this distinction one might hold that not all non-relational qualities of a thing need be constituents of it. Hence some might be held to be exemplified by one thing as the relation L is exemplified by two things. Such properties would , l ike all relations , be accidental as opposed to essential. Again, in a clear sense , a quality that was a constituent of a thing might be held to be substantial as opposed to adjectival, though not substantial in the sense in which a complex of qualities - a particular thing - would be said to be substantial. (/3 ) thus offers the possibility of incorporating and explicating several classical themes in the history of metaphysics. In his most recent works 'Notes on Ontology (Including a reply to Professor Hochberg) ' 1 0 and 'Notes on the Ontology of Minds' , 1 1 Bergmann runs into a problem with the axiom of difference , what he calls 'the fundamental principle' of ontology. In a highly innovative and suggestive move he recognizes an entity , called a 'diad' , to resolve problems about ( 1 ) order in facts, (2) the grounding of truths of i dentity and difference, (3) the connection between a thought and its intention, (4) the traditional problems associated with the axiom of infinity in a logistic analysis of elementary arithmetic , and (5) the nature of classes. Consider two objects, a and c. For Bergmann, the two objects now do not exemplify a relation of diversity. This would involve recognizing a fact that would ground the truth of "a =F c'. But such a fact is problematic on two counts. First , it would seem that the existence of the fact would presuppose the existence of diverse objects, a and c, and hence not ground the truth of 'a =F c'. Second , the attempt to ground the truth of 'a =F c' in the existence of a fact would not fit with the apparent necessity of the truth of 'a =F c'. For Bergmann , the truth must be a matter of logic and not of fact. Hence , the two objects are held to form a 'diad'. Such a diad is a thing, though it is complex . Any two things 'eo ipso' give rise to a third , their diad. The existence of such a diad grounds the truth of the relevant statement of difference. And, since 'eo ipso' any two things give rise to such a diad , the existence of such a diad is a matter of logic and not of fact. Bergmann suggests that a thing can also form a diad with itself that is then potential rather than actual. The potentiality of the diad of a thing with itself is what grounds the truth of identity statements l ike 'a = a' and is the basis for their necessity. Thus, neither 'a =F c' nor 'a = a' states a fact , and

275

not stating facts, though being true, they are logical truths. The truths being grounded, respectively, by the actuality of one diad and the potentiality of the other. Since any two objects or entities 'eo ipso' form a diad, there is a simple derived benefit of Bergmann 's pattern. He has always been concerned to justify an axiom of infinity. Now, given two objects, a and c, we have their diad, represented by ' (a, c)'. By the claim that for any two entities we have a third, we then have a further diad, the diad of which a and (a, c) are constituents, the entity (a, (a, c)), and so on. Thus, two objects generate an infinite number of objects. This , of course, is reminiscent of the use of the unit set operation and a familiar objection to Fregean senses. Moreover, if Bergmann takes seriously his recognition of the potential diad (a, a), and he must to ground the truth of "a = a' on his pattern, he can also generate an infinite series from one entity, so long as he holds that a -=I= (a, a). This raises an obvious problem . Given that (a, a) is potential or possible, is (a, (a, a)) actual? Note also that Bergmann's diads, like his facts, are both actual and potential or possible. This obviously casts doubts on his introduction of diads and his claim that expressions like '(a, c)' are not disguised sentences ( merely ways of writing sentences of the form 'a -=I= c'). Aside from such issues, Bergmann's use of an iteration device to justify an axiom of infinity is essentially the same as the pattern for the iteration of properties to justify such an axiom in 'Properties, Abstracts, and the Axiom of Infinity' in this volume, and both patterns, of course, derive from the unit set operation employed for such purposes by von Neumann. Bergmann also attempts to use diads along the lines Wiener and Kuratowski employed to analyze ordered pairs as sets to dispense with the need for an ordering entity in the analysis of facts. This analysis of his is considered in detail in 'Logical Form, Existence, and Relational Predication' in this volume. His attempt to use the idea that there are complex entities which do not contain a connecting relation is also employed in his analysis of the nature of classes and of the intentional connection between a thought and the state of affairs that is its truth condition. The former issue is not relevant to the concerns of the present essay . The latter issue is relevant. Bergmann considers all particular thoughts that a is F to share a common property. Let us indicate this property by 0. (For the details of Bergmann ·s view and a criticism of it see 'Belief and Intention' in this volume . ) Yet, he holds that the property, which is fundamentally like a Fregean proposition or thought (except for Bergmann an individual mental state of a person would exemplify the property, while for Frege a person would stand in a relation to the proposition), picks out its corresponding state of affairs as a matter of logic and not of fact. Hence that 0 picks out the fact - Fa is not, itself, a 276

fact or state of affairs. Thus, Bergmann holds that the two entities, the property 0 and the fact - Fa (whether actual or possible) form a com plex entity which he calls a meaning circumstance. D iads are also circumstances; circumstances of diversity. Circumstances are the complexes which are not constituted by connections holding among their other constituents. They are simply complexes of constituents without connecting ties or relations. Thus, his fundamental intentional relation M is held to be of the same status as =I= . Neither relation enters into a complex diad or meaning circumstance, and, hence, both are declared by Bergmann to be 'literally nothing'. · =I= ' thus stands literally for nothing. . . . in this world "M' stands for literally nothing ... On his latest view , the truths that "0 means that- Fa' ('0 M Fa') and that , " 0 =I= the fact Fa would be grounded respectively by the actuality of the meaning circumstance composed of 0 and the fact (potential or actual) Fa and the actual diad composed of the same two entities. But, since M and =I= are literally nothing and do not enter into the complex entities ( the meaning circumstance and the diad), what distinguishes these complexes? Bergmann has no answer or must, in ad hoc fashion, hold that the fundamental principle only holds for certain kinds of complexes, facts or property bundles, and not for his circumstances and diads. The problem becomes even more acute when we consider 0 and Fe . The fact Fe and the property 0, being different entities, will form an actual diad . Yet, since 0 does not mean ( or intend or pick out) the fact Fe, the m eaning circumstance composed of 0 and the fact Fe will be potential. But, what distinguishes the actual diad from the potential meaning circumstance? Either Bergmann must abandon ' the fundamental principle' of ontology or much of his latest analysis. Notes 1

2

3

4

Reprinted in B. Russell , Logic and Knowledge, ed . by R. Marsh ( London : 1 956) . This view is held by G . B ergmann in several essays in Mea,p,ing and Existence ( Madison , 1 959) and Logic and Reality (Madison : 1 964). However, as the elaboration of this view , and the arguments for it which follow , are what the present author takes to be involved in it , I do not wish to attribute them to anyone . This does not preclude , by stipulation , attem pts to link sentences as wholes to entities. One would have to argue that such a gambit must also ulti mately recognize universals , since propositions or facts will have to be complexes containing other entities. But this involves arguments removed from the issues of this paper. Part of what i s involved is that while substrata are held to be 'parts' of 'complex' ordinary things , universals are not 'parts' of qualities.

277

5

6

7

8 9

10 11

One may, as B radley did , feel that a nexus is incomprehensibl e . The point is that if there are problems about a nexus they weigh equally for (a) and ((3) , and such problems do not serve to put universals and substrata on a par with respect to a stipulation like the axiom of difference . This may lead one to speak of a thing like Socrates as a conjunction of facts or as a complex fact, as opposed to constituent simple or atomic facts. Questions may be raised about the naming of complex entities on ((3) and as to whether sentences like 'Socrates is white ' become analytic on that view. I have attempted to deal with such issues in 'Things and Qualities . ' For a discussion of some issues surrounding this assumption and of the use of relational qualities to differentiate obj ects see 'Things and Descriptions' in this volume. See my ' Moore's Ontology and Non-natural Properties' , Review of Metaphysics, XV, 3 ( March 1962) , pp . 365-95 . For a consideration and defense of Moore's and Russell's early arguments for universals see A . Donagan , 'Universals and Metaphysical Realism ' , The Monist, 47 , 2 (Winter 1 963) , pp . 21 1 -246 . In Nous, May , 1 98 1 . I n Foundations ofAnalytic Philosophy, ed . b y P . French e t al . , (Minneapolis: 1 98 1 ) .

Facts and Truth

Recently, the correspondence theory of truth and the appeal to facts that i t i nvolves have been attacked as involving a logical paradox. The criticism, in the spirit and style of Frege' s attack on the correspon dence theory, and reminiscent of Bradley's holism, has been advanced by Donald Davidson. Employing a pattern of argument that Quine has used for other purposes, Davidson claims that if there are facts, there is exactly one fact: the Great Fact . The argument purporting to establish this conclusion goes as follows: Indeed, e mploying principles i mplicit in our examples , it is easy to confirm the suspicion . The principles are that if a statement corresponds to the fact described by an expression of the form ·the fact that p' , then it corresponds to the fact described by 'the fact that q, provided 'p' and 'q' are logically equivalent sentences, or one differs from the other in that a singular term has been replaced by a coextensive singular term . The confirming argument is this. Let 's' abbreviate some true sentence . Then surely the statement that s corresponds to the fact that s. But we may substitute for the second 's' the logically equivalent '(the x such that x is identical with Diogenes and s) is identical with ( the x such that x is identical with Diogenes)' . Applying the principle that we may substitute coextensive singular terms , we can substitute 't' for 's' in the last quoted sentence , provided 't' is true . Finally , reversing the first step we conclude that the statement that s corresponds to the fact that t, where 's' and 't' are any true sentences . 1

To see why the argument is fallacious, let us consider it in a slightly more explicit form. Let (P 1 ) and (PJ be the principles Davidson employs: the first dealing with the replacement of logically equivalent sentences, the second concerning the replacement Qf a sentence by another that differs only i n containing a different, but coextensi ve, singular term. Let us use the expression " ' s' Den. s" to state that the sentence ' s' corresponds to or stands for the fact that s. The argument may then be put as follows: 1. s

2. t

3. 4. 5. 6.

's' Den. s (t x ) (x = Diogenes & s) = ( tx)(x = Di ogenes) (t x)(x = Diogenes & t) = (tx) (x = Diogenes) (t x) (x = Diogenes & s) = (t x) (x = Diogenes & t)

279

7 . 's' is logically equivalent to ' (tx) (x = Diogenes & s) = (tx) (x = Diogenes) ' 8 . 's' Den . (tx)(x = Diogenes & s) = ( t x) (x = Diogenes) 9. 's' Den . (tx)(x = D iogenes & t) = ( t x)(x = Diogenes) 10. 't' is logically equivalent to ' (tx)(x = Diogenes & t) = (tx)(x = Diogenes) ' 1 1 . 's' Den . t . Hence , where 's' and 't' are any true sentences, either one stands for the fact we might think to be expressed by the other. Consequently, there is no point distinguishing, or trying to distinguish , among so-called facts. Steps ( 1 ) , (2) , and (3) may be taken as premises of the argument . Steps ( 4) and (5) follow from the premises, by standard logical procedures , with two implicit assumptions . We assume , first , that we are employing Russell's theory of descriptions , with respect to the understood expansions of the definite descriptions , in context, and , second , that the sentence 'E! ( tx)(x = Diogenes) ' is true . Step (6) is a consequence of steps ( 4) and ( 5) by the standard rules for identity . Step (7) also involves standard logical procedures together with two implicit assumptions . Agai n , we assume we are employing Russell's theory of descriptions and , moreover , we assume that the sentence 'E ! (tx) (x = D iogenes)' i s not only true but logically true . Step (8) is j ustified by (P i ) and steps (3) and (7) . Step (9) follows from steps ( 6) and (8) by use of (P 2 ) . Step ( 10) follows for the same reasons as step (7) . Step ( 1 1 ) follows from steps (9) and ( 10) by the application of (P 1 ) . Since the argument makes implicit use of Russell's theory of descriptions, it is necessary to examine it in a form where such contextually defined signs are replaced by their definitions . But , if one expands the descriptions in ( 4)-(8) , it immediately becomes obvious that the argument is invalid . There is no way to get from the sentences , so expanded , to (9) . In fact , (P 2 ) becomes totally i rrelevant . It is easy to see why the problem arises here . The sentence ( 4) is merely a rewriting of the conj unction 's & (tx)(x = Diogenes) = (tx)(x = Diogenes)" . The same holds for the corresponding sentences i nvolving 't' and both •s' and 't' . If one , then , considers an argument expressed in terms of such conj unctions , it becomes clear that one could get to the purported conclusion if one employed a totally specious principle . That principle , call it (P 3 ) , would be somethi ng on the order of: any true sentence may replace any other true sentence to the right of 'Den . ' in contexts of the form '" . . . ' Den . . . . " . We may thus go directly to ( 1 1 ) from ( 1 ) , (2) , and (3) . But , then , the 'problem' , along with Davidson's · argument' , has disappeared . The apparent plausibility of the argument , as originally formulated , depends on the use of a descriptive phrase , containing a 280

conjunction sign, as an apparent subject term. Definite descriptions allow for a typographically condensed way of expressing a conjunction in an apparent subject-predicate form. Thus, one can apparently apply the principle (P 2) . B ut , that prir\ciple may be used only if it reflects a valid argument in unabbreviated form. In this case it does not , and the form of notation disguises the fallacy, since it appears to permit the use of (P J. Since such an application is blocked, in the expanded form of expression, we need not feel arbitrary in rejecting apparent descriptions like ' ( t x)(x = Diogenes & s )' for replacements in identity contexts by using (P ,,) . The validity of an argument cannot be made to depend on its occur""'ring in typographically abbreviated form. Moreover, we can reject Davidson's argument without having to claim that ( P 1 ) or (P 2 ) must be abandoned since the argument involves nonextensional contexts, i. e. , the statements containing 'Den. '. One may suggest taking the descriptions to have primary scope in ( 8) and ( 9) . It would be illicit to do this in step ( 8) , since we are using ( 4)-(7) in the argument. Expanding the descriptions in ( 4)-(7) prohibits the use of primary scope in ( 8). This would involve an additional assumption that one may do so in such contexts, whatever that would mean. This suffices to reject Davidson's argument irrespective of questions of scope. But , to see another point, let us assume such manipulations of scope are permitted. If we adopt primary scope in ( 8) , we can get (9). This is easy to see. We have s, t and ( 3 x) (y) [ ( (y = Diogenes & s)

= x = y) & 's' Den. x = x]

as an expansion of ( 8) , with primary scope. All we need do is extract 's' from ( 8) , so expanded , and replace it by 't'. This can be done by moves of standard logic. But, we still have no way to get to ( 11) . Hence, the argument still fails, even with the assumption about scope. What we can do, however, is get a 'peculiar' result from ( 8) , expanded as above, or (9) , expanded as ( 3 x) (y) [ ( ( y = Diogenes & t)

= x = y) & ' s' Den. x = x] .

This, given the standard uses of '=', permits us t o get t o '" s' Den. Diogenes = Diogenes" . We, thus, have 's' standing for the fact that Diogenes is Diogenes. Hence, the illicit move to primary scope does 'cover' something. We can see what by considering the following pattern: 1. s 2. 's' Den. s 3. s :J [ (pv ---- p)

= s] 28 1

=

4. (p v -- p) s 5. ' s' Den. p v -- p. With ( 1) and (2) as premises and (3) and ( 4) obtained by propositional logic, we get (5) by employing a replacement based on material equivalence. Doing so, we end with ' s' standing in Den. to a logical truth, as above . This is what the shift to primary scope amounts to in the above pattern. Once again, we have an illicit use of material equivalence as the basis of a replacement. That is what the use of descriptions covers up in the argument. Hence , we see what is peculiar about the use of true sentences, 's' and 't', in such arguments. Given the truth of a sentence , it is materially equivalent to a logical truth, which then replaces it. We might j ust as well adopt (P3 ) or a general rule for replacing true sentences with true sentences. The use of primary scope brings all this out when we recall that on Russell's theory there i s supposedly no difference between the use of primary and secondary scope when the descriptions are fulfilled, except for nonextensional contexts. To take the contexts involving ' Den. ' to be extensional in that there i s no difference regarding scope i s to explicitly acknowledge the use of (P3 ) , since i t i s covertly employed where we expand the descriptions with secondary scope. To seek to establish Davidson's conclusion by the use of an explici t primary expansion of the descri ption is to assume, by such an expansi on, that we may replace ' s' by a tautology in contexts involving 'Den . ', an d hence make use of (P 3 ) in the assumption of primary scope. What we learn from Davidson ' s argument is that some singular terms, in the form of definite descri ptions, can only be unproblematically 'replace d' in some argument patterns i f we accept (P 3 ). All thi s is not surprising when we recall other uses of Davidson's borrowed argument. As Quine put it: " Anything x, even an intension , i s specifiable in contingently coincident ways i f specifiable at all. For suppose x is determined uni quely by the condition ·cp x' . Then it is also dete rmi ned uniquely by the conj unctive condition "p · cp x' where •p' is any truth, however irrelevant. " 2 The use of · specifiable' covers a gap which some seek to turn into a mine of philosophical problems. As many put it, the contexts ' cp x & p' and 'cp x' , given that ·p' is a true sentence, have the same extension or are sati sfied by the same " obj ects'. Suppose that such contexts uniquely determine an obj ect, a. The interesting que stions are: (A) Doe s the description LX)(

Gx] ' . All such uses may be regarded as alternative formulations for statements where negations occur only in front of atomi c forms. Hence, no problem arises for the ontological accounts of negation, either in terms of a negative nexus or in terms of negative properties. Yet, one may raise a question about the ontological status of the logical equivalences used i n such transformations. This, in effect, re-raises the question about the ontological status of logical truths. We may also note that for the issues in this paper it is not necessary to distinguish between the particular white patch and another sense of 'particular' whereby the patch contains a particular substratum or ground of individuation. The latter would be what stands i n the exempli fication nexus to the properties and thereby constitutes the ordinary particular. Such a substratum would then also be what stands in the relation of negative exemplification to a property. If one holds that an ordinary particular, like the patch, is to be analyzed only in terms of a set of qualities in a relation, then the tie of negative

30 1

exemplification would be taken to hold between the complex particular (the patch) and a simple property it does not have, i. e. , which is not a constituent of it. There would then be a fundamental asymmetry, on such a view, between the analysis of 'a is white' and 'a is not black'. In the former case, the predicate would be taken to indicate a property joined with other properties by a tie or nexus to constitute the object: in the latter case, the nexus of negative exemplification would connect the property to the complex particular composed of properties. The two sentences 'Wb' and '--- Wb' then reflect the two possibilities that b is connected to the property W either by exemplification or negative exemplification. That the sentences are contradictories reflects the impossibility of a particular standing in both ties to one and the same property. This way of putting things points to a further difference between the case of 'Wb' and '---Wb', on the one hand, and that of 'Bb' and 'Wb', on the other. For, it would seem that if one held to a synthetic a priori necessity or incompatibility between B and W this would be reflected in the claim that a particular cannot simultaneously stand in the exemplification nexus to two properties that are incompatible, rather than not being able to stand i n two relations to one property. But this difference disappears if we accept a third alternative account of negative facts. One may hold that there is neither a negative element nor a negative nexus and seek to account for negative facts by means of negative properties. Thus a negative fact would be analyzed in terms of a particular standing in the exemplification tie to a negative property. On this view both synthetic a priori incompatibilities and logical incompatibilities would involve a particular not (possibly) standing in the exemplification tie to two ( or more) properties . Yet, one could hold that the sense of 'possibly ' would still be different i n the cases of synthetic a priori and logical impossibility, but this difference would be reflected by the distinction between properties and negative properties , rather than by that between exemplification and negative exemplification. For, incompatible properties , like B and W, are not related as are a property and its negative correlate , say B and ---B. This assumes '---B' is not taken as a 'simple' sign referring to a 'simple' property , but that the former is defined in terms of ',....., ' and 'B' and the latter is taken as a complex property having B as a constituent , in some sense. If the negative properties are taken as simple one abandons the expFcation of logical necessity propounded by Wittgenstein in the Tractatus. For , in effect , one will base logical truths on necessary relations between simple properties , and, consequently , consider logical truths analogously to the way some philosophers have spoken of synthetic a priori truths. It is not the purpose of this paper to explore alternative accounts of negation and negative facts, but analyzing negative facts either in terms of negative 302

properties or negative exemplification , preserves a distinction between logical incompatibility and synthetic a priori incompatibility . This difference may go along with another. One who adheres to synthetic a priori necessities recognizes an additional kind of fact . Thus , the incompatibility of B and W would be reflected by the addition of something like · I (W , B )' to the list (a) , where 'I' stands for a primitive notion of incompatibility . Here we are not concerned with the question of negative facts nor with whether or not 'I' is a problematic notion . We merely note that holding to synthetic necessities requires a new category of facts . By contrast , one may attempt to speak of analyticity or logical necessity in terms which do not require such an addition . This would involve holding that logical truths like ·--(p&--p )' do not require an ontological ground , either in the form of logically necessary 'facts' or the fact that the world has a certain "logical form' . If successful such an attempt would reveal a further difference between synthetic and analytic necessities . In this discussion I have merely mentioned but not considered alternatives to the introduction of negative facts , such as possible facts . Hence , questions about the structure of the latter or whether such alternatives differ only verbally are not taken up . The reason for this is simply that the point being argued is that some type of entity , in addition to atomic facts , is needed to ground true negative sentences, and each alternative recognizes such a ground. Moreover , I suspect that , as long as one confines oneself to the issue at hand , no real difference will emerge between the alternatives . As one argues against the need for conj unctive facts one can seemingly argue against existential facts . What makes a sentence like ' ( 3x) Wx' true is the fact expressed by 'Wa' . Again , this is reflected in the valid pattern for existential generalization . With universally general facts there seems to be a differe nce . There is the point of Russell's that Sa & Sb (x) S (x)

won't do as a valid argument form . 2 Hence , we must add that a and b are all the obj ects. This would give us , in effect , (y)

Sa & Sb & ( x ) (x = a v x = b) (x) Sx .

But then we have a generality as part of the basis of inference . Yet '(x) (x = a v x = b ) ' is a special type of generality . For any finite list like (�) all one needs is a generality stating that the individuals listed constitute the 303

totality of individuals. Hence, if we take the valid inference pattern of existential generalization to be the key to the avoidance of existential facts, we can maintain, in view of (y), that there is only one general fact for a finite universe: that pertaining to how many things (at most) there are. For an infinite domain this pattern will no longer do. In that case we may hold that the very notion of a list is specious. We will consider such a domain shortly. More pertinent for the moment is another type of case. Consider two black crosses against a small white background enclosed by a circular boundary. One sees that there are two black crosses. One also sees that every cross in the circle is black or that no cross in the circle is red. D o these 'facts' indicate that there are existential and general facts corresponding to the relevant true existential and universal sentences? One might be tempted to say so if he thought that the experiences in each case were both different from and irreducible to others that could be described by atomic sentences. Since the experiences are not so reducible, what is meant or intended by ( 1) 'Every cross in the circle is black' is not the same as (2) 'a is a black cross in the circle and b is a black cross in the circle'. Nor, is it the same as (3) 'a is a black cross in the circle c and b is a black cross in the circle c and only a and b are in c'. Here the term 'meant' can be taken two ways. First, one can point out that the universal quantifier is not reducible in meaning to a conjunction. Second, one can contend that a thought or intention expressed by ( 1) is different in kind, from one expressed by either (2) or (3). This latter contention is independent of the issue of the definition or reducibility of the quantifiers. For, even though one defines 'p � q" as ' ----- p v q', one may hold that a thought whose content is expressed by one is different in kind from a thought whose content is expressed by the other. Hence, led by the definitional irreducibility of the universal quantifier to a conjunction, one may reject (2) as a transcription of ( 1); led by the irreducibility of intentional contexts, one may also reject (3) as a suitable transcription. Thus, one could come to hold that (1) indicated a general fact. But this would be a mistake. For the issue, recall, centers around what facts are necessary to ground the truth of ( 1). The issue is neither whether the meaning of the universal quantifier can be captured by definition nor whether thoughts expressed by a general sentence are different from and irreducible to other thoughts. To point out that ·w a or Ba' does not mean 'Wa' and that thinking that Wa or Ba is not the same, in intention, as thinking that Wa, is irrelevant to whether or not there are disjunctive facts. Likewise, holding that ( 1) means neither (2) nor (3) is irrelevant for the question of general facts. Assume there are four objects , a, b, c and d, and properties referred to by 'C 1 ' for "cross' and 'C2 ' for 'circle' , in addition to W, B, and S and a relational predicate 'I' for 'in' . The issue is whether or not a suitable 304

exhaustive list, (8) , of atomic and negative sentences containing "C 1 a', 'C 1 b' , 'C2c' , 'lac', etc. , together with a general sentence stating that there are only four objects or, alternatively, the assumption that the list is complete so that the fact that there are only four objects may be said to show itself, suffices to ground the truth of the generality that all the crosses in the circle are black. Since such a list and such a general sentence do suffice, no general fact is needed to correspond to and ground (1) . Are there then general facts? One who is impressed by the point that generality is required only in the form of a general statement about the number of things there are and by the contrast of this case to that of negation might say · no· . He might even feel that this requirement is already taken care of by the specification of the list of atomic and negative sentences. That is, even the general sentence as to the number of things there are is not one that belongs to the list and is, in fact , not even necessary, since this is implied or shown by the list and the stipulation that the list is complete. But this just seems to be a less explicit way of making a claim about the number of things that there are. Hence, one who is impressed by the need to stipulate something in the form of a general sentence, whether added to the list or made about it , might feel forced to acknowledge general facts; i. e. , one general fact about individuals for any finite universe. For we cannot explain such generalities away in the manner we employed for conjunctive and disjunctive facts. An attempt might be made to strengthen the case against general facts by insisting, in Wittgensteinian fashion, that the number of individuals there are is something that cannot be stated but must show itself. Hence, it would be inappropriate to even consider whether there is a general fact corresponding to the true sentence 'There are only four individuals in the universe which is depicted by (8) ' . Being inadmissible, such a sentence could not indicate a general fact. This would put the matter far more strongly than merely holding that such a general sentence is not required . Following such a line, one might insist that if we hold that a list like (8) depicts a model universe, and if we further employ an interpretation rule or, i f you will, picturing principle, according to which one sign (name or predicate) represents one thing (individual or universal) ; it follows from (8) 's being a picture or representation that there are only four individuals in the universe represented or pictured . One is thus led into holding that there is a general fact only by not realizing what it is to be a picture or representation. But thi s does not alter the point that no set of atom ic sentences will suffice to ground a universal generalization. We may then wonder if it is really crucial whether one adds an explicit premise in the form of a sentence as to there being at most four individuals or holds that thi s shows itself. Clearly generality has not been eliminated . To meet this objection, one might suggest , following standard logic texts, that in a

305

finite universe a universal quantification can be taken to express a conj unction. Hence , in a universe of two things , ' (x)Sx' would follow from 'Sa & Sb ' , since the former is merely an abbreviation of the l atter . Here one would have to acknowledge that the very meaning of the universal quantifier changes depending on the number of things there are , i. e. , its sense is relative to the domain we are using it to speak about. But even if there are no objections to this , we could still not express that there would be at most so many things . Hence, one would stil l have to claim that that 'fact' showed itself. Consequently , one might suggest that the peculiarity of a sentence like ' (x) (x = a v x = b ) ' , even if admitted in order to derive ' (x)Sx' from 'Sa & Sb & (x) (x = a v x = b )', need not commit us to the recognition of a general fact . B ut this seems more in the nature of announcing a decision than providing an argument. What is peculiar would be the nature of the general fact . For , it might not seem amenable to analysis into components as did the atomic and negative facts . Or, to reflect one of Wittgenstein's worries , it is difficult to see how such a fact could be depicted. However, we m ight consider generality to reduce to a relational property holding among the individuals a and b. That a and b are all the things there are is a relation they exemplify . Hence , the need for only one universally general sentence like ' (x) ( x = a v x = b ) ' permits us to consider generality as a property of things . (It is even tempting to speak of a 'logical' property) . This would have the advantage of enabling us to consider the structure of a general fact analogously to the structure of atomic and negative facts. One may even speak of the general fact as a kind of atomic fact since it is analyzed in terms of a primitive relation exemplified by particulars . Alternatively , since such a relational property is unlike any ordinary relation , there is a sense in which such a fact is different , in kind , from any atomic fact . Actually , there would be an infinite number of such relations . Each would be an n-place rel ation , where n would vary according to the size of the domain whose members exemplified the relation . Thus, while each would play the same role in a different possible domain , there would be two aspects involved. To say of a universe of two things that they are all that there are is to say something similar and something different from saying of a domain of three things that the three are all that there are . In a way this simply reflects the varying and the constant sense of the universal quantifier when it is taken as equivalent to a two term conj unction in a domain of two and a three term conj unction in a domain of three item s . B u t there is a crucial difference . For, the relations are not taken as 'the meaning' of the universal quantifier, since the similar roles played by the various n-term relations is relevant . Thus , the 'sense' of the quantifier is not given by a specific conj unction or n-term rel ation but, in part , by the fact that each possible domain involves an n-term relation . 306

Just as one may feel forced to recognize a universally general fact corresponding to there being only a certain number of individuals , one seems obliged to acknowledge an existential fact corresponding to there being at least a certain number of objects . In short , to state that there are exactly two obj ects , for example , requires both an existential and a universal sentence . Moreover, such an existential sentence will not follow from any set of atomic sentences and negations of such , as " (3x)Wx' fol lows from ·wa· . Here , one may rehearse the same arguments about certain things showing themselves or being implicit in the notion of a list or representation of facts as were raised about the universally generalized sentence . In fact , one may reinforce the point in the case of the existential sentence by pointing out that accepting "(3x)Wx' as a consequence of ·wa· requires holding that signs like 'a' must designate . Yet , we would hardly give ontological status to the designation relation . To connect a name to an obj ect requires us to recognize the obj ect , but not the connection , as an entity . Thus , it may be held that if on the list of atomic and negative sentences characterizing the model universe the different names · a' and • b' occur, and no other names occur , this suffices to ·show' that there are two obj ects . In effect, it is argued that j ust as one requires that names designate one also requires that diffe rent names designate different obj ects . The second rule should have no more ontological significance than the first . But , while it is true that we may consider both rules to be on a par in that we require a symbolism to conform to the m , there is a crucial difference . To require names to name does not state a fact about the world but simply lays down a rule for a language to be used about it . To require that different names name different things can be taken that way also . But that there are two obj ects rather than three is a fact about a world . Hence , we must not let the rule about different names for different things implicitly assert this fact while pretending that it is simply a rule. In short, there is an existential fact about at least how many things there are j ust as there is a universal one about at most how many things there are . Granting such an existential fact we face the question of its structure . Here a symmetry with the purported analysis of the universal fact suggests itself. Difference or diversity has been considered a relation holding between any two different things . But, in a way , to say something is different from something else is to say there are at least two things . That is , there is at least a logical equivalence between the two sentences in that if there are two things , then something will be different from something else , and if the latter is the case it follows that the former is. The same can be said in the case of three or more obj ects . Just as one speaks of two obj ects being different one may speak of a three term , four term , etc . , relation holding among the diverse things that there are . And we may 307

correspondingly hold that the existential fact as to the least number of things that there are is to be analyzed in terms of an n-term relation of diversity among the objects of the model universe. The existential fact thus corresponds, in structure, to the universal one, and the relation involved in the universal fact may be thought of as the dual of the notion of diversity or difference. Or, perhaps better, we might think in terms of two relations, one involved in the existential fact, the other in the universal, and together they reflect the idea of diversity in that both are involved in sentences like 'There are five different things' and 'Consider a universe of five objects'. Like the relation involved in the universal fact, the one in the existential fact is peculiar in that it is not an n-term relation where n can be specified. Speaking of relations corresponding to the general and the existential fact may seem excessively ad hoc. In part, I share the feeling but do not see that it is any more ad hoc than speaking of exemplification as a special kind of relation. Perhaps there is a difference in that one who says exemplification is a special kind of relation does so under the pressure of an argument like Bradley's regress. By contrast, I am suggesting that we consider quantification in terms of certain special relations, not under the pressure of an argument, but to enable us to consider the general and existential facts as basically similar to other facts. But whether such a difference reveals that one 'solution' is ad hoc and the other not or that different issues are resolved in different ways is an open question. A relational property of diversity being involved in an existential fact may seem problematic if we consider a domain of only one individual. For here there would seem to be an existential fact but clearly no relation of diversity to hold among different objects. Thus it appears that one has to treat such a domain as a limiting case. This leads to a few further points. We have been treating the assertion that a is diverse from b as involving the claim that there are two things. Thus the relational property of diversity is, in effect, a property of existence. This may be thought to be brought out in the limiting case of the domain of one where there appears to be a one term property rather than a relation. Since such a property is taken to be involved in the existential fact, and this fact is expressed by the use of the existential operator , one might say that the existential operator reflects a property of existence even though it is not a predicate. (Just as one can hold that a relation of diversity may be indicated by the use of different signs for different things rather than by a predicate). This, in turn, would merely reflect the point that such a property, like the relation of exemplification or that of identity, is not an ordinary property or relation. In the limiting case of a domain of one object, the universal quantifier also reflects a one term property. But this property would still differ from 308

that reflected by the existential quantifier, for there is a clear difference , between the notions of · at least one' and ·at most one . That there are two properties rather than one, even in the limiting case, is thus seemingly indicated by the need for both a unive rsal and existential statement. Thus the equivalence of the universal and existential quantifier in a domain of one does not seem to point to the collapsing of the two properties into one in such a domain. However , there is a further problem. To say that there are at least two objects is to employ the existential sentence '(3 x) (3y) (x =I= y)' or, perhaps " (3 x) (3y) (x = a & y = b & a =I= b )' ; but to state that there is at least one would involve something like ' (3x) (x = x) ' or ' (3x) (x = a) ' . Yet. ' (3 x) (x = x)' . unlike " (3x) (3y) ( x =I= y)', is analytic, assu ming one does not allow the empty domain as an interpreting domain in validity theory. This may lead one to suggest that the assertion that there is at least one thing is not a fact but a matter of logic, and hence no existential property is involved in the limiting case of a unit domain. To put it another way the very notion of model, domain or universe involves there being at least one thing. Hence, the fact that separates the unit domain from all others is that there is at most one thing. To put it still differently, the idea of an empty universe makes no sense. Thus the limiting case becomes one in which there is no property of diversity and not one in which the relation becomes a one term property. In this way the logical equivalence of the quantifiers in the unit domain is reflected in there being only one general fact and one general property in such a universe. To speak of a property of existence, as I did above, is thus misleading, if one holds that ' (3x) (x = x)' is analytic and does not state a fact. (This assumes that one takes '(3x) (x = x) ' to state that at least one thing exists. ) Since the relations of generality and diversity ground the truth of a statement as to the number of things that there are in a given universe or model, one might be tempted to think of the ontological ground of elementary arithmetic in terms of such relations rather than in terms of classes of classes. N atural numbers could then be taken to be relational properties holding in possible universes or models. Thus, the facts, or possible facts, indicated by ' (3x) (3y) [ x =I= y & (z) (z = x v z = y)] ' and ' (3 x) (3y) (3z) [ x =I= y & x =I= z & y =I= z & (w) (w = x v w = y vw = z)] ' would constitute the ontological ground of 2 and 3. Natural numbers would then have a distinct ontological status rather than being construed as the logical fictions (classes) of the logistic reconstruction. To explore this in detail is beyond the scope of this paper, though we may note the bearing of such a possibility on the status of '(3x) (x = x) ' as analytic, and a su bsequent question about the number 0. The above discussion of universal and existential facts has been confined to finite models. Withou t going into the question of whether the notion of an infinite universe is ultimately a sensible one, we can note 309

certain fundamental differences about infinite and finite model universes . The existential sentence applicable to an infinite world would be of a different kind since , in principle , no set of existential quantifiers over individual variables would suffice . Thus , something like Russell 's axiom of infinity would be required . Moreover , one universal fact would no longer suffice . For, together with the sentence used to state that there were at most a denumberably infinite number of obj ects , no set of atomic ( and negations of atomic) sentences would suffice as premises to derive a generality of the form ' (x) (Fx :) Gx)' . This is not merely a matter of there possibly being an infinite number of F's but of the possibility of there being an infinite number of F's which are G's without all F's being G's. The point is a simple consequence of the arithmetic of the infinite . Each true generalization would, in effect , require its own fact . Thus what sort of existential and universal facts there are would seem to depend on whether the domain one speaks about is finite or not . One fundamental obj ection may be raised concerning the criterion used to dispense with the need for admitting conj unctive and general facts. It can be pointed out that j ust as one may derive the conj unction 'Wa & Wb' from 'Wa' together with 'Bb' and derive ' (x)Sx' from 'Sa' , 'Sb' , and '(x)(x = a v x = b) ' , we can also derive ' Wa' from the conj uncdon and we could derive a sentence indicating an atomic fact from a generalization , as in the case of 'Sa' and '(x)Sx ' , or from a generalization and another atomic sentence , as in the case of ' (x)(Wx :) Sx) ' , 'Wa' and 'Sa' . Hence , it would appear as if one could hold , in terms of the criterion employed above , either that there is a conj unctive fact but no atomic fact or a general fact but not an atomic fact indicated by the atomic sentence derived by use of the general sentence . But this objection would overlook a crucial point . One could not consistently claim that there was a conj unctive fact without there being an atomic fact . For , one would be forced either to consider the conj unctive fact to be such that it contains atomic facts as constituents or to analyze it into a relation among the constituents of the various atomic facts . In the former case one explicitly acknowledges atomic facts . In the latter case a problem arises . Let the structure of the conj unctive fact indicated by ·wa & Bb' be depicted by 'R(W, B a, b) ' . One could not then derive that a is W while b is B . One would have to link the appropriate particulars and universals in the structure of the conj unctive fact . But this would be to acknowledge atomic facts . Thus , the existence of the conj unctive fact would presuppose the existence of the atomic fact it supposedly dispenses with . A similar point can be made against the purported use of a general fact to ground the truth of atomic sentences and hence enable us to dispense with atomic facts . Here , unlike the case of conj unction , one will require atomic sentences in the derivations in those cases where the generalities 310

are like ' ( x) (Fx ::::> Gx)' , rather than of the form'(x)Fx' . Hence, one coul d not seek to avoid atomic facts altogether. Aside from this, a question arises about the structure of general facts. If one anal yzes them in terms of a relation among all the constituent obj ects and properties involved , the same probl em arises that we noted in the case of conj unction. But there is an alternative for generalizations like ' (x)(Fx ::::> Gx)' . One could, for example, take the general fact to involve a second order relation between the relevant properties. 3 Yet, in both this case and the case of · ( x)Sx', the existence of the general fact may still be said to imply the existence of atomic facts ind icated by the atomic sentences whose truth the general fact is supposed to ground . For. given th e distinction between particulars and properties and the consequent tie of exemplification, it would be incoherent to deny that th e particular and universal , referred to by the name and predicate of the derived atomic sentence, are tied by exemplification. By contrast, one could deny that the constituents of the supposed general fact do stand in a primitive relation corresponding to the universal operator even though a relevant set of atomic sentences, guaranteeing the truth of the general ization, held 4 • This simply reflects the point that we use the universal quantifier to assert that each particular is tied to a certain property. To put it another way, the basis for the inference from ' (x)Sx' to " Sa' is that the universal quantifier is taken to express that every particular, incl uding a, exempl ifies S. If ' (x)Sx' , when true, is to be taken as indicating some general fact, a question arises as to the connection between '(x)Sx' and 'Sa' whereby the latter is derived from the former. To put it still differently to take ' (x)Sx' as indicating a general fact would make no sense unless the existence of that general fact involved or implied that there were atomic facts like that indicated by 'Sa' . But there is nothing comparabl e to this forcing us to h old that general facts are presupposed by acknowledging atomic facts and the one general fact about the number of individuals in th e domain. In dispensing with certain categories of facts the criterion is thus not simply the derivability of sentences purportedly referring to those facts. What one must hold to exist, in order to coherently analyze the facts acknowl edged as existents, is also involved.

Notes 1

2 3

B . Russell, 'The Philosophy of Logical Atomism' , in Logic and Kno wledge, ed . by R. C . Marsh (Allen and Unwin , 1956) , pp. 212-13. B . Russell , An Inquiry into Meaning and Truth (Penguin Books, 1962) , pp . 8tr-87 . Recently a few authors (Armstrong ( 1978) , Dretske ( 1977) , and Tooley ( 1977)) have advocated taking such a second order relation among first order universals to provide an

31 1

4

adequate analysis of natural laws and causal connections. This reproduces a crucial idea of Frege's: that generality or universal quantification is to be understood in terms of second order functions taking first order functions as arguments . That one construes causal connections in this way is a relatively trivial matter . The basic question is whether or not one can viably treat quantification in terms of second order functions. The obvious obj ection is that one apparently requires different functions so that '(x) (Fx => (Gx v Hx))' can be distinguished from ' (x)((Fx => Gx) v Hx)' ; '(x) (y) (Rxy => Ryx)' from '(x) (y) ( Ryx => Rxx)' ; and so on. This recalls Russell's recognition of forms of propositions in his unpublished manuscript 'Theory of Knowledge' . On this point see 'Logical Form , Existence and Relational Predication' in this volume. For a discussion of attempts to use second order relations to analyze causal connections see my 'Natural Necessity and Laws of Nature' , Philosophy ofScience, 48 , 3, 1981. This is not to say that one can consistently deny that the constituents , along with the other individuals of the domain, stand in a relation to constitute the one general fact that I have argued is necessary in any finite domain. The point is that if a generalization like '(x)Wx' or '(x) (Wx => Sx) ' is true, one can consistently deny that there are general facts corresponding to those particular true sentences.

Arithmetic and Propositional Form in Wittgenstein's Tractatus

One crucial aspect of Wittgenstein ·s later philosophy. the idea that the logical is not restricted to the tautological, is a simple consequence of his philosophy of arithmetic as presented in the Tractatus. According to Wittgenstein, number is a formal concept, as are object, fact, and function. Thus what represents such a concept in a proper symbolism is neither a class sign nor a function sign but a category of sign or kind of variable, just as the notion of an object may be represented by the category of individual signs or the occurrence of individual variables in the symbolism. As Wittgenstein puts it: 4 . 1 272 Thus the variable name •x' is the proper sign for the pse udoconcept object. 1

This immediately rules out the Russellian reduction of arithmetic to logic, since ( 1) there can be no class of natural numbers , (2) a number of statements that can be made on the Russell reconstruction such as '1 is a number' are declared nonsensical ( 3) a number cannot be a class , for number expressions must be of a logically different kind than class expressions ( just as signs for objects must be of a logically different kind than propositional signs or function signs) , and , hence, ( 4) an adequate symbolism for ari thmeti c must show the logical or internal relations among arithmetical signs , while Russell' s symbolism purportedly does not do so. In short, that a sign stands for a number must 'show itself', and what must show itself cannot be said : 4 . 1 26 A name shows that it signifies an obj ect , a sign for a number that it signifies a numbe r , etc .

But then, does the ontology of the Tractatus encompass numbers as well as objects and facts? If so, what are numbers? If not, what is intended by the assertion that a sign for a number signifies a number? At places Wittgenstein suggests that a number is what is in the sequence of successors of O or reached as the result of some finite number of applications of the successor function; it is , in the familiar way, a member of the posterity of O with respect to the successor function. This is what is involved in

3 13

6 . 03

The general form of an integer is [O,

s, s + 1 ] .

and i n the series of definitions i n 6.02, and i n 4. 1273 If we want to express i n conceptual notation the general proposition , 'b is a successor of a' , then we require an expression for the general term of the series of forms aRb , (3x) : aRx · xRb , (3x ,y) : aRx · xRy · yRb ,

In order to express the general term of a series of forms, we must use a variable , because the concept 'term of that series of forms' is a formal concept . (This is what Frege and Russell overlooked . . . ) We can determine the general term of a series of forms by giving its first term and the general form of the operation that produces the next term out of the proposition that precedes it .

If 'R' is taken as a variable, then it is clear in what sense the series of forms is 'vari able' . But even if 'R' expresses the relation between any number and its immediate successor in the number series , Wittgenstein holds that we must use a variable to stand for the concept "being a member of that series' . Only such a sign will reveal the internal or logical connection among the terms: that they are generated from O by the successor function. This is 'shown' by the sign ' [O,s,� + 1] ' and not by Russell's definition of n atural number. Yet , aside from the somewhat specious declaration that certain things must be shown and not said , Wittgenstein at best seems to be simply offering the Peano postulates as a philosophy of arithmetic. For it seems as if he takes 'O' and ' + 1' as primitive notions and holds that 'number' is shown in the symbols for numbers . That is, the sign ' [O ,s,s + 1 ] ' suggests that correspondents of the Peano primitives are taken as primitive by Wittge nstein . The only apparent difference is Wittgenstein's 'insight' that the general form of expression , and not another primitive term , stands for the formal concept number. 2 Moreover, such a point is rather minor by contrast with a dilemma one now faces . Given that 'O' and ' + 1' are primitive , one either takes them as uninterpreted signs in an abstract axiomatic system or as having a definite interpretation . If the former, then we are , so to speak , taken back to the point where Russell and Frege took issue with the "formalists' . If the latter , then we face the query regarding what such signs 'signify' . This is especially pointed for Wittgenstein since , in some sense , arithmetic is a matter of logic: 3 14

6.2

Mathematics i s a logical method . The propositions of mathematics are equations , and therefore pseudo­ propositions . 6 . 2 1 A proposition of mathematics does not express a thought . 6 . 22 The logic of the world , which is shown in tautologies by the propositions of logic , is shown in equations by mathematics .

It would thus appear that the basic arithmetical signs are in some way to be construed as logical sign s, and logical signs do not signify as names signify obj ects. Wittgenstein must then present an analysis which, first, shows in what sen se arithmetical signs are logical signs, second , avoids the formalist treatment of arithmetic as uninterpreted axiomatics, and , third , avoids the introduction of arithmetical entities. We can see how Wittgenstein goes about such an analysis by considering his discussion of the general form of a truth function and of a proposition. This he writes as ' [ p,tN(�) ] ', which he explains as: 6 . 00 1 What this says is just that every proposition is a result of successive applications to elementary proposi tions of the operation N (�) .

The expression for the general form of a proposition stands for (or one may say 'shows that') any proposition may be arrived at by applying j oint denial to a selection of elementary propositions, and then applying it again to any selection made from the set of elementary propositions or propositions so obtained , and so on. One also gets a sequence of propositions generated by N(�) which bears some analogy to the series of propositions generated by R in 4. 1 273. The elementary propositions can be considered the limiting case where the operation is performed zero times, or as the result of not performing it. Thus, 5 . A proposition is a truth-function of elementary propositions . (An elementary proposi tion is a truth-function of itself. )

The analogy with the set of statements in 4. 1 273 or with the number series itself is weak. A nscombe seeks to strengthen it by offering a procedure for ordering the obtainable truth functions of elementary propositions. 3 But that does not really help. What is involved can be seen in terms of the application of j oint denial (' ! ') to a single proposition 'p 1 ' . The first result is 'p1 ! p1 ' . If we follow the analogy with either the number series or the propositions of 4. 1 273, we should then apply the operation to this result. But that yields ' (p1 ! p1 ) ! (p 1 ! p 1 ) ' , which is equivalent to 'p 1 ' . Hence we d o n ot produce a new member of the series. Moreover, there are two other function s of the proposition, 'p1 v -p 1 ' an d ' p1 · - p1 ' , which must be generated. We get one of them if we apply the stroke to the two

315

arguments 'p 1 ' and 'p 1 i p 1 ' with the result 'p 1 i (p 1 i p 1 ) ' . We get the other, if we apply the stroke to that result and obtain '(p 1 i (p 1 i p 1 ) ) i (p 1 i (p 1 i p 1 ) ) ' . We then have the sequence;

There is an analogy to the use of the successor function in the case of the generation of the second line from the first and of the fourth line from the third. But the getting of the third line from the first and second is a different kind of move.Of course one can stipulate a stepwise procedure for getting to all the functions of a set of propositional variables and dropping redundant ones.This Anscombe does, and, in effect, we just did for the case of one variable. But two things are clear. First, the members of such a series could be arranged in several ways, depending on the procedure chosen.Second, one does not, in all cases, get to the n + l st member by applying the operation to the nth member. The natural numbers, by contrast, cannot be 'arranged' in several ways via the application of the successor function to O and the successors of 0.And, of course, one is getting to each term of the series by applying the operation to the previous one, or, perhaps, applying it a successive number of times to 0.Moreover, there is another distinct and crucial disparity.With one variable the base, in this case'p 1 ' , can be considered a term of the series and yet every term is generated from previous terms in the series.With two or more variables either the base propositions are not members of the series or, if they are, some members will not be generated from previous members. In the arithmetical case O is a member of the series (or mentioned in the first proposition of the series in 4 . 1273 or 6.02), and every other term comes from applying the operation to a prior term.4 Such differences aside , there is a sort of analogy that appeals to Wittgenstein. Thus arithmetic appears as a matter of form just as propositional logic, based on truth functions, is a matter of form. The relation of 1 to O is something like the relation of "p 1 i p 1 ' to "p 1 ' , and hence a matter of logic. One can then reasonably hold that a postulate (for Peano) like 'O is a number' shows itself just as it shows itself that "p 1 i p 1 ' is a truth function of'p 1 ' . Just as a propositional logic constitutes a · method of logic', so arithmetic may be thought to be a 'method of logic'.Hence , while arithmetic is not reducible to logic in Russell's sense, it is a kind of logic or a disti nct part of logic, somewhat as one may nowadays hold that quantificational logic is a part of logic in addition to propositional logic. But there is more than the analogy between the operations N(s) and + 1, 316

and their role in generating numbers and truth functions or propositions , behind Wittgenstein' s characterization of arithmetic as a method of logic . He seems to see another analogy between the way arithmetical equations function in their empirical applications and the way in which argument patterns and logical truths function . Consider the argument pattern (x) (Fx :J Gx) (3x)Fx : . (3x)Gx with "F' and "G' as variables . Suppose I know that all men are mortal and that there are men . I may then use the above pattern to conclude that there are mortals . Wittgenstein thinks that we apply arithmetical truths in the same way . 6 . 2 1 1 Indeed in real life a mathematical proposition is never what we wan t . Rathe r, w e make use o f mathematical propositions only i n inferences from propositions that do not belong to mathematics to others that likewise do not belong to mathematics .

Thus , j ust as the above pattern corresponds to the logical truth '[ (x) (Fx :J Gx) · (3x)Fx] :J (3x) Gx' , the arithmetical truth '7 + 5 = 12' corresponds to the addition pattern

+

7 5 12

which may be applied , i n a given case , as 7 apples here 5 apples there 1 2 apples altogether. This not only links up arithmetic to logic as another method of logic, but also points to a way of holding that arithmetic is a matter of logic, that arithmetical signs do not stand for entities , and that, nevertheless , both Russell's reduction and the abstract axiomatics of the formalist can be avoided. For one does not require the logical signs in the application of tautological patterns to stand for anything. Yet logic is not a question of

317

abstract axiomatics. What logic reflects is the form of the world. Hence, so does arithmetic. Recall here 6. 22. Wittgenstein escapes from the crucial questions regarding the ontological ground of arithmetic exactly as he does those of logic itself. 5 There is yet another link of the arithmetical and the logical. He tells us that 6 . 021 A number is the exponent of an operation .

Whether we think in terms of some variable operation or in terms of the specific operation N (�), the idea is that we can construe numbers recursively in terms of exponents of operations. Taking the operation to be N (;) gives us another link between 'ordinary' logic and arithmetic. The main point, h owever, is that arithmetical signs do not stand for anything, since they are all defined in terms of 'O' and ' + 1 ', and 'O' and ' + 1 ' are recursively defined by the device of taking them as exponential signs. Of course what Wittgenstein defines is what we might call the null application of an operation. Similarly, he defines the n + l st application of the operation, for, recall, having disposed of the notion of number as a pseudo-concept, he need only define Peano' s remaining two primitives, 0 and successor.The n + 1 st application of an operation yields the same result as the application of the operation to the result of the nth application. The two definitions in 6. 02 0

' x Def. , f! ' f!Y' x = f! Y + 1 ' x Def.

X = f!

then yield the series X,

fl ' X , f!'f!' X , f!' f!' f!'x, . . .

Here I am not concerned with questions regarding the adequacy of Wittgenstein' s approach: with whether or not his use of the iteration of the operation sign is question-begging, or whether his definition of 'O' is so. 6 All we need note is that , as expone nts of logical operations, the basic arithmetical notions expressed by ·o· and ' + 1' indicate that arith metic constitutes a system of logic. 7 A s arithmetical truths are not tautologies, but nevertheless logical or formal truths, the stage is set for the idea that the logical is a q uestion of the rules of the 'language game' and not to be restricted to the analytic truths of formal logic. Russell accommodated the felt difference between arithmetical truths and empirical ones by attempting to reduce the former to logical truths. Wittgenstein, not holding that all logical or formal truths can be construed as truths of 318

standard logic, will end up by holding that what is logical is a question of a language rule. In a way, then, he wi ll come to hold that the tautologies are merely another kind of reflecti on of rules of language. One who holds that x is F and not F doesn, t understand the language, just as one who holds that 3 equals 2 + 5 does not understand the language of arithmetic. Wittgenstein thus starts out by objecti ng to Russell's reducti on of one kind of statement to another, the arithmetical to the analytic, and he ends by treating the analytical i n the same way he treats the arithmetical. What he does i s end wi th the formalist view, which Russell and Frege attacked in arithmetic, in both arithmetic and logic. This is only partly disguised by grafting a contextualist theory of meaning onto the structure of his positi on under the guise of ' the use· replacing ' the meaning'. In short, the doctrine that mathematics constitutes its own language game and that understanding mathematics is understanding the connection of various statements and procedures within the game, as well as their application in nonmathematical contexts, expresses the core of the formalist view attacked by Russell and Frege. That such a view was already essentially present in the Tractatus, we have seen. To clinch the matter, all we need to note is that the construction of arithmetic in the Tractatus depends on taking the exponent (or iteration) ofan operation as a primitive notion. 8 In effect, what Wittgenstein did was cryptically lay down rules for such a notion, as Peano had set down stipulations about number and successor. Later he came to take the contexts of pure mathematics and the application of mathematics to constitute the rules. Notes 1

2

3 4

5

Quotations from the Tractatus are taken from the D. F. Pears and B. F. McGuinness translation ( London: Routledge & Kegan Paul , 196 1 ) . The stress on the apparent difference here i s twofold. As we shall see, Wittgenstein offers definitions of 'O' and ' + 1 ' . and it is not clear that his version of arithmetic is adequate. There are problems regarding the arrangement of parentheses in identity statements , which Rhees reports he later recognized ( Rush Rhees , Discussions of Wittgenstein [New York: Schocken Books, 1970) , p . 35) , though I am not sure that Wittgenstein recognizes that '( l + 1 + 1 + 1) = ( 1 + 1 ) + (1 + l ) ' must be proven or, perhaps better, '[ ( 1 + 1 ) + 1) + 1 = ( 1 + 1) + ( 1 + l ) ' . G . E. M . Anscombe , An Introduction to Wittgenstein 's ' Tractatus ' ( New York: Harper & Row, 1965) , Ch. 10. Sometimes Wittgenstein seems to think of the series of forms generated by the successor operation as the series of propositions reflecting what is the successor of what ; sometimes he considers the number series itself. The former case strengthens the analogy with the truth­ functional sequence , since we deal with propositions in both sequences. In a way he doesn't, for all matters of form are grounded in the essences or forms of obj ects (and perhaps facts). See 'Facts, Possibilities, and Essences in the Tractatus' , in Essays on Wittgenstein.

3 19

6 7

8

There is a problem corresponding to the one leading to Russell's axiom of infinity , if we are dealing with an operation (ultimately) on eleme ntary propositions . The phrase 'logical operation' has two senses here . First, there is the idea that we are dealing with what is common to any series of the kind Wittgenstein is concerned with ; that is , where we have a null application that yields the same member that it operated on and where the n + 1 st application yields the same result as operating on the result of the nth application. In this sense we deal with something logical in that we are concerned with the abstract pattern common to all such series . Second , there is the idea that we have a logical operation , j oint denial , which generates such a series . The first sense reinforces the idea that arithmetic, like logic, is a matter ofform applicable to various content. I t also shows Wittgenstein to hold that we see what arithmetic is when we see what is common to all series of the relevant kind . He thus advocates a mixture of formalist and intuitionist lines . His view is intuitionistic in that everything is based on the fundamental operation genera ting a series . It is formalistic in that it appears as if we are dealing with an abstract model for such a series with abstract , uninterpreted signs . The twist Wittgenstein introduces is to 'define' the operation . But he presupposes an iteration notion as primitive. If the notion of iteration is itself formalized in an abstract system, we h ave a formalist view. If it is implicitly taken as a basic notion and not treated in terms of abstract sign patterns , then we have a kind of intuitionist view . As Wittgenstein puts it, we have a bit of both . In either case he doesn 't really ' define' the successor operation ; he simply rewrites it. (See n. 8 . ) A furt her point may lie behind Wittgenstein's not thinking he is offering a version of the formalism Russell rejected. O ne naturally speaks of an abstract system as indicating a form common to its various interpretations . Thus , one may think of such a system as standing for a formal property of such interpretations and , hence , as not being merely an abstract system of signs. See Quine's discussion of iteration in Set Theory and Its Logic, pp . 79-80 , 95-102 . Perhaps the point to note is Quine's definition of 'iterates' in terms of 'successor' and , ul timately , in terms of the logical apparatus with iterates as powers of relations .

Russell's Reduction of Arithmetic to Logic

Several problems are connected with Russell's ·reduction' of elementary arithmetic to logic . I n this paper I am concerned with only one of them , which , while not technical like those surrounding the theory of types and the axiom of infinity , is a fundamental one . The problem may be put simply . Aside from any technical issues relating to the logicists' reconstruction of the Peano postulates and concepts , why has Russell ·reduced' elementary arithmetic to logic and not merely mapped the Peano system onto Principia Mathematica or, at best , constructed an interpretation of the Peano system? The question is particularly pointed if one does not feel that , via the bridge afforded by analytical geometry , one may also reduce , for example , Euclidean plane geometry to logic. To j ustify this feeling one must point out distinguishing features which make one mapping a reduction and the other merely a mapping . Here I shall attempt to defend the Russell reduction in the case of elementary arithmetic and argue that there are distinguishing features which , first , substantiate his claims about arithmetic and , second , need not force the logicist to claim that the same type of analysis applies in the case of a geometry . Russell's program must be understood as embodying a philosophical thesis in response to distinct philosophical or metaphysical questions . In the case of arithmetic there are two questions . First , what entities , if any , are the arithmetical propositions about? Or, to put it differently, what constitutes the ontological ground for the truths of arithmetic? Second , what kinds of truths are the true propositions of arithmetic? To get at these questions we must first note that three distinct sets of propositions and signs are involved . First , there is the set of propositions and signs that constitute ordinary arithmetic. Among them are signs like '7' , ' 5 ' , · + ' , etc . and truths like '7 + 5 = 12' . These signs and statements are , in one definite sense , unproblematic with respect to their meaning. Questions about their meaning can be answered in terms of their ordinary use and context . Part of this context consists of their role in mathematics proper and part in the applications of mathematics - from the esoteric uses in science to the most mundane actuarial functions . One crucial feature stands out - as Russell noted - namely , in applying numbers to things , we attribute them.to groups or collections of things . Thus, we may be said to ordinarily tre._i t numbers as attributes of collections . We will 321

return to this later; for the moment we need only note that it is a feature of our ordinary use of numerical concepts. A second system of signs and sign sequences involved is the Peano system of postulates and theorems. This system can be considered in two ways. To get at what is essential it will be harmless if we take liberties with both the Peano system and history. Assume that, at the time of Euclid, a number of geometrical truths about triangles, circles, etc. , were known but not organized into an axiomatic system. Assume next that Euclid supplied such an axiomatics for such propositions, and, in so doing, added some new ones and selected some as axioms or basic propositions so that a number of the known truths turned out to be theorems. In one sense one may say that the propositions, known prior to Euclid's work, are exactly the same as those incorporated into his system. The same terms are used, the applications are the same, etc. However, in another sense, onemaysaythatEuclid, in changing the contextual setting for such terms and sentences involving them, changed the significance and hence the meaning of the signs and propositions. In Euclid's axiomatic system, the same signs and sentences may occur but with a different meaning. So long as were are clear how we use ' meaning' and 'same' here, there is no problem and nothing to argue about. Similarly, in one sense, we may say that Peano supplied an organized axiomatics for our ordinary arithmetical propositions. So taken, he enriched the context for arithmetical concepts and propositions. But there is a second way of looking at the Peano system. We may take it as an abstract axiomatic system of which ordinary arithmetic, organized in a certain way, is an interpretation. We then have an abstract axiomatic system, whose primitive signs are uninterpreted marks, and a system of ordinary arithmetic, whose signs need no interpretation due to the meaning supplied by the rich context of the ordinary use and application of such signs. 1 Noting the two ways in which the Peano system may be taken is important for the issues at hand, but, once noted, we may forget, for our discussion, its role as an organized version of ordinary arithmetic and consider it to be an abstract axiomatic system. The third relevant system of signs is that portion of Russell and Whitehead's Principia Mathematica which is involved in the interpretation of the Peano system that they proposed. Just as the signs in the Peano system differ from those of ordinary arithmetic, the '"number signs" of PM differ from both of these . The '"meanings" of the number signs of PM, in a clear sense of ·meaning' , are precisely specified in terms of their definitions by means of the logical primitives of PM. Such signs are thus neither abstract marks nor those of ordinary arithmetic. Let us call the three systems 'O A' , 'PA' , and 'RA' , respectively, and use a lower case 'p' and "r' as subscripts for the respective arithmetical 322

signs of the Peano and Russellian systems, that is of PA and RA. Thus, '7' , '7/ , and ' 7/ , are three signs and not one. Consequently, to say either that Peano defined ' 7' or that Russell defined "7' or '7 ' is at best false. What one should say, if he seeks to defend Russell's philcisophy of arithmetic, is that by means of RA Russell solved the philosophical problems raised about arithmetic that we mentioned above. In terms of our example, Russell used ·7/ and RA to interpret ·7 ' and PA and answer metaphysical questions raised in connection with ·7Pand OA. Just what this means and involves must be spelled out and defended. Since the key to Russell' s solution is his handling of the ontological questions, we will get into them by taking up some points about ontological issues in general. C onsider an object, o, that has the property white and a philosopher who, in providing an ontological analysis of the obj ect and specifying the ground or basis for the truth of the sentence 'o is white', recognizes in his ontology, first, a universal referred to by the term'W, a particular or substratum referred to by · 0 1 ' , a nexus of exemplification which connects universals to particulars and which is reflected in a clarified language by j uxtaposition of subj ect and predicate terms, and a fact, which is indicated by the sentence 'Wo 1 ' and which contains W, o 1 , and the nexus. The ordinary claim that it is the case that o is white is thus reflected in such an ontological position by adhering to several different ontological kinds. (One could say " kinds of entity', but som e philosophers distinguish universals and particulars as entities from the nexus which, while not an entity, has some sort of ontological status. ) A different philosopher will contend that the ordinary fact can be reflected and the ordinary obj ect analyzed without recourse to such an extensive ontology. We thus arrive at a metaphysical conflict between alternative ontologies. How such a conflict may be resolved , or if it is resolvable at all, are not questions we need take up here in any detail. We need only note, first, that Occam ' s razor functions as a basic principle in constructing an ontology, and, second, that an adequate ontology must fit the ordinary facts and objects. A mapping or coordination helps explai n what is meant by 'fitting' ontological analyses to ordinary facts. A philosopher, advocating the ontological analysis we j ust consi dered , coordinates to the notion of the ordinary object o that of a complex entity of which 0 1 and W are constituents. Likewise, he coordinates the universal W to the property white and the fact that o 1 exemplifies Wto the ordinary fact, or, if you will, truth that o is white. 2 Thus, an ontological position consists of a set of interrelated concepts and statements together with a coordination procedure which maps these onto ordinary notions and statements. What is required , of course, is that the philosophical statements mapped onto the ordinary ones be both equivalent in truth value and relevant to the philosophical issues raised in connection with the ordinary statements. 323

Truth functional equivalence does not pose a problem ; the relevance of proposed analyses does. But this latter question is not fruitful ly discussed in general. Hence, we shall return to it later in a specific context for the issues at hand. By offering an ontological analysis of the object o and the specifying of a ground of truth for the sentence 'o is white' , one does not hold that the object is not real, since only its constituents are real. Nor, does one hold that the object really is composed of universals and substrata instead of chalk, wood, atoms, or what have you. The sort of constituents of the object that we have been talking about are mentioned in connection with a specific set of questions, which are not requests for a compositional analysis in any ordinary sense. Similarly, if one was concerned with other philosophical problems, one might talk about a chair as a pattern of sensa or a history of momentary physical time slices, about a self as a set of experiences related in certain ways , and so on. Again, if one is careful, he does not thereby deny that, in one sense, there are chairs, selves, continuants, and so forth, since, in another context, they are construed or analyzed according to a philosophical position. In short, one must not confuse the two distinct contexts, one philosophical, the other ordinary, since one speaks of chairs, selves, numbers, etc. in both and uses the one to answer questions and problems raised about the other. Keeping the two contexts distinct should also prevent one from making the misleading claim that the one context, the philosophical one , provides the real meaning for the concepts of the other, the ordinary one. This question of 'meaning' is central and we shall return to it later. For the moment an analogy may be helpful. A common use of the term 'reduction' occurs when one speaks of macro-theories being reduced to micro-theories . One example would be in the case of a macro-behavioral theory being reduced to a physiological one ; another classical example is the reduction of a macro-theory of gases to a particle theory. One also speaks of states of the respective systems being reduced to corresponding micro-states. Thus , the fact that a gas has a certain temperature is reduced to the fact that the particles which constitute the gas are in certain state. In a straight-forward sense of 'formal', such reductions are formally the same as the philosophical or ontological analyses we considered. Two sets of statements and terms are coordinated in such a way that the one set is mapped onto the other. In the case of the scientific theories there are several relevant features that lead one to speak of ·reduction'. For example, the particles are taken to be literally physical parts of the macro-objects. Further , the coordinating statements are either discovered or hypothetical physical laws, and the basic laws of one theory become deductive consequences of those of the other, taken together with the coordinating statements. But even in such

324

cases, it is misleading to hold that the gas does not have a temperature, really , since there is ultimately only the particle state or that for a gas to have a certain temperature really means that it is in a certain particle state. Of course one can so legislate his use of such terms that the claims are stipulated to be true. But such legislation provides no resolution for the philosophical puzzles generated by our use of such terms. A first step at a solution involves noting , and not confusing, the various senses of terms like · real' and · mean' that are germane to the issues. 3 The statements that there is one and only one even natural number between 7 and 9 and that 7 plus 5 equals 1 2 are truths that belong to the ordinary arithmetical context. Thus, we ordinarily claim that there are numbers with certain characteristics. If one then raises the questions, "What do we mean when we say a number exists?' and "Why is'7 + 5 = 1 2' true? " . an appropriate response might be along the lines Wittgenstein goes through in Remarks on the Foundations of Mathematics and earlier in the Tractatus. One sort of answer to these questions lies in understanding the context of ordinary mathematical usage; learning the language game of arithmetic and its applications to physical objects and situations. Similarly, if one asks what an object is or what a property is and how objects and properties are connected , one could explain the ordinary context for the use of such concepts. But none of these explanations constitute an ontological or philosophical analysis. 4 To provide such an analysis one must take such questions in another way. Taken in this other way, to ask what is a number or what number signs indicate or why '7 + 5 = 1 2· is true is to ask for the location, in one' s ontology, of the things and relations which ground the truths of arithmetic. Russell located such a ground in terms of classes and logical truths about classes. What ontological status classes and logical truths have, and what philosophical problems may be raised about such notions, is not relevant here, though it is, of course, a basic part of the whole story. The point that is relevant here is that the way one deals with such questions about classes and logic provides, on the Russellian approach, a solution to the ontological problems about arithmetic. But proposing the Russellian solution invites the raising of our original question in another form: ' Are numbers really classes? ' Consider the problem of universals. A philosopher argues that there are universals on the basis of holding that no nominalistic account can adequately ground the true ascription of one and the same predicate to two objects or the fact that two objects have the same property. If this line is cogent, one seems forced to accept universals, just as one seems forced to accept a special tie or nexus of exemplification, or something similar, in order to avoid Bradley's regress. By contrast, one seems to argue for the construing of the ground of arithmetical truths in terms of relations

325

among classes on the basis of parsimony. One does not seem forced to accept Russell's thesis as one does seem forced to accept universals and the nexus of exemplification. Yet, the appearance is deceptive. For , given the adherence to principles like Occam's razor, in the absence of any equally cogent and parsimonious alternative, one is forced to Russell's solution.Hence, if we ask 'Why is Russell's achievement more than a mere mapping?' a reply is forthcoming, in part, in that the logistic thesis provides the most parsimonious ontological grounding for the truths of arithmetic. But this needs to be spelled out. To do so I shall consider some objections and alternatives to the logistic program, which return us to the question 'Are numbers really classes' ? Suppose one objects to the Russell program on the basis of the old adage that a thing is what it is and not another thing. In short, numbers are numbers, not classes.Russell is claiming that one sort of thing, a number, is another sort of thing, a class.This may result in a more parsimonious ontology, but it involves a grotesque procedure of 'reducing' something to something else. Just as a phenomenalist is involved in the absurd claim that a physical object, such as a train, is a collection of sense data or a physicalist makes the ridiculous assertion that awarenesses, experiences , and all mental states are really bodily states, the logicist claims that numbers are classes and arithmetical truths really logical truths. To avoid such illicit reduction one must recognize numbers as a category of entity and ground the truths of arithmetic in terms of relations holding among such entities. One thus recognizes arithmetical entities and, as it were , arithmetical facts. Such a line assumes that we start from the implicit or explicit acceptance of the fact that there are numbers. But, in the sense in which this is true, it is irrelevant. We all admit that there is an even natural number between 7 and 9 .This does not mean that we hold, either implicitly or explicitly, to a philosophical position recognizing an entity which stands in a three-term relation to two other entities denoted by the signs '7' and '9' . Nor does it mean that we are forced to adopt such a position on that basis alone . This would be like holding that, since one can buy two shirts of the same color, some form of Platonism must be adopted and all forms of nominalism rejected, on the basis of that fact alone . Here we should not be misled by its being the case that we might be led to an ontology recognizing numbers as basic entities or some form of Platonism on the grounds that such views alone serve to adequately analyze and ground the ordinary truths like '7 + 5 = 12' and 'Those shirts have the same color'. Clearly what is involved here is a very complicated procedure of fitting one alternative to the ordinary facts and successfully refuting the competing views. But this is not done in the simpleminded way of either pointing out that we say there are numbers and that objects 326

have the same color in our ordinary use of number signs and color words or that we accept such claims as obvious truths. To think otherwise is, in effect, to fuse ordinary language analysis with ontological analysis and, in so doing, trivialize both. Whether numbers are entities and if so what kind, are precisely the philosophical questions which we answer by fitting a metaphysical position to the ordinary facts. But this involves much more than reciting commonplace truths. The truths of arithmetic, which we start from, do not, by themselves, imply that numbers are entities, or that they are classes, or that they are not classes. To rej ect the Russellian thesis on the ground that arithmetic is about numbers, not classes, is to be overly simplistic about philosophical questions in general and not merely about those relevant to arithmetic. The obj ection we have j ust considered may be raised in an apparently more cogent form. Instead of merely asserting that numbers are numbers and being indignant about Russell's presuming to reduce one thing to another, one might hold that no statements about classes may be taken as · reconstructions' of arithmetical statements, since the two kinds of statements do not · mean' the same thing. To dissolve this form of the obj ection we must separate three aspects of it or, perhaps better, three different senses of ' meaning' that are involved in it. First, there is th e simple point that the intentions and mental states relevant to the one set of statements differ from those relevant to the other. Thus, what we have in mind on occasions when we ordinarily use arithmetical notions and assert arithmetical truths will not, generally, be what one has in mind when he uses the correlates of such notions and statements, which are offered by a particular philosophical position, in the context of a philosophical analysis. While this is unquestionably true, it is also, I submit, irrelevant. Consistently pursued, such an approach will merely lead one to a version of the so-called paradox of analysis and to the self-defeating rej ection of every philosophical and ontological analysis. If the fact that such an approach is self-defeating and paradoxical is not taken as sufficient grounds for rej ecting it, we may part company from its advocates on the basis that such a requirement (intentional equivalence, if I may so put it) for a ph ilosophical analysis being sufficient for or fitting the ordinary facts is overly puritanical and stringent. B eing so, it does not provide a reasonable ground for the rej ection of Russell's reconstruction of arithmetic or of any other proposed philosophical analysis. 5 A second strand of the obj ection is closely related to the first. One may rej ect a thesis like Russell' s by holding th at the proper answers to the philosophical questions surrounding arithmetic are to be arrived at by exploring the subtle.ties of the ordinary conceptual context of arithmetical propositions and notions. To understand what numbers are

327

it suffices to understand how to use arithmetical concepts as they relate to each other, as well as in measuring, counting, etc. The meaning of arithmetical concepts is given by such a context, rather than by definitions or reconstructions in Russellian fashion within the system of PM. Statements belonging to the one context cannot, therefore, be construed in terms of statements belonging to the other. On the basis of such an argument, one not only recognizes that '7' and '7 / are different terms belonging to different contextual settings but also claims that in no sense is '7 / relevant to the 'meaning' of '7'. One does more than insist on the distinction between the ordinary and the philosophical contexts and claims more than that the two give distinct meanings to their respective concepts. For, in effect, such a line of argument claims that philosophical puzzles are answered or dissolved by merely exploring the details and subtleties of the ordinary contextual setting. What one does is hold that only the ordinary context is significant or, to put it another way, that there are no ontological problems to be answered by specifying a ground for the arithmetical truths. One thus refuses to raise certain questions or, putting the same thing still differently, one takes ordinary arithmetic to constitute its own ontological ground. The same approach is naturally applied to other philosophical problems. Using it, one does not solve the problems of universals and predication by recognizing entities of a certain kind but dissolves them by exploring the ways in which we use predicates and verbs. It is, perhaps, ironic that the first way of rejecting the Russellian analysis, on the basis that arithmetical statements do not mean the same thing as statements about classes, leads to an abundant ontology with the recognition of special arithmetical entities ; while the second objection, on the ground that arithmetical concepts are unique and to be understood only in their own context , leads to the rejection of any ontological questions about arithmetic. The first has affinities to an extreme form of Platonism and is reminiscent of Meinong ; the second has similarities to the formalistic standpoint that Russell attacked and is generally the approach implicit in Wittgenstein's comments about arithmetic. Both share certain features with the so-called intuitionist philosophy of mathematics. But, historical connections aside, both gambits may be rejected as critiques of Russell's approach. The first may be rejected in that it relies on a requirement that is too stringent and, in effect, rejects any analysis ; the second in that it refuses to consider the questions Russell sought to answer. This brings us to the third aspect or sense of 'meaning' that is relevant to these objections and to the question of whether or not Russell showed that numbers are really classes. Recall two statements that we discussed earlier :

328

( 1 ) o is white. (2) o contains a particular o 1 which exemplifies the universal W . It is clear that ( 1) and (2) will differ in intentional contexts. It is also clear that (2) does not belong to any ordinary , nonphilosophical context , and that if we restrict ourselves to such contexts we shall never utter a sentence like (2), let alone raise a question as to whether it provides an analysis for, or meaning to, ( 1 ). In these senses ( 1 ) and (2) mean quite different things. But, if we do raise problems about universals , particulars, and predication, then a further sense of 'meaning' becomes relevant. The analysis a philosopher gives of ( 1), which is schematically expressed by (2), may be taken to provide such a further sense of 'meaning'. In this sense one may hold that ( 1) and (2) mean the same thing , according to one specific metaphysical position. Hence, it should also be clear that the difference in meaning between ( 1) and (2) and the fact that ( 1) , but not (2) , has an ordinary nonphilosophical use are irrelevant to the adequacy of (2) as a proposed analysis of ( 1). What goes for the problem of universals goes for Russell's analysis of arithmetic. Certain statements about classes in the system of PM can be taken to mean the same thing as the arithmetical statements of OA with which they are correlated, in the sense that they are offered as purported ontological analyses in response to ontological questions about OA . This is why the two objections we considered above are not relevant to the question of the success . or lack of it , of Russell's analysis. What is relevant is, first , whether the mapping is technically adequate , and, second , whether it can be used to cogently answer the philosophical questions. The question of the technical adequacy is not our concern here. Defending the philosophical cogency of the Russell analysis amounts to pointing out how it jibes with or fits the ordinary context of arithmetic, arguing that it is the ontologically most parsimonious alternative, refuting philosophical objections to it , and establishing the inadequacies of proffered alternatives. The technical adequacy and on­ tological parsimony I have taken for granted here. I have tried, how­ ever, to rebut some objections , and, in so doing, argue against some alter­ natives. We shall return to those alternatives and to another as well, short­ ly. Now , I should like to point to those features of the ordinary context that show Russell's analysis to both jibe with it and be relevant to it . One outstanding fit with the ordinary context is that feature we noted earlier and which Russell stressed. We ordinarily apply numbers to groups or classes of objects. Construing numbers as classes of classes accurately reflects such usage. A second fit is with our preanalytic notion that arithmetical truths are somehow necessary or conceptual truths rather than empirical ones, for, on Russell's analysis, they are correlated

329

to logical truths. If one recalls that phrases like 'necessary truth' and 'conceptual truth' are themselves analyzable, along lines quite compatible with Russell's approach, in terms of the notion of logical truth, the point becomes even more forceful. Russell's analysis also preserves, as he noted, the ordinary notion that arithmetical signs are not mere abstract marks or symbols but meaningful signs associated with definite concepts. Russell , however , seems to have thought that he was supplying the meaning to our ordinary arithmetical signs by a set of definitions in RA just as, in another context, he thought he was supplying the meaning to signs like 'Pegasus' and 'the present king of France'. In both cases he puts matters in a misleading way and invites the Wittgensteinian response about arithmetic and Strawson's criticism of the Theory of Descriptions. B ut, in both cases Russell is essentially correct in that the philosophical analyses he gives of the respective problems can be cogently put. 6 Here, the point is not that he supplies the meaning to ordinary arithmetical expressions and statements , but that according to his analysis, as well as according to our ordinary usage, arithmetical expressions are not mere abstract sign designs. No attempt to solve the philosophical problems surrounding arithmetic which takes arithmetical expressions to be uninterpreted, abstract sign designs can jibe with this obvious feature of the ordinary context. This disparity goes along with another. Taking arithmetical statements as uninterpreted formal patterns precludes taking them as true statements. Russell also made this point. But, thinking that he was supplying the meaning to the ordinary arithmetical terms , he put both claims a bit differently: . . . it fails to give an adequate basis for arithmetic. In the first place , it does not enable us to know whether there are any sets of terms verifying Peano ' s axioms ; it does not even give the faintest suggestion of any way of discovering whether there are such sets . In the second place , as already observed , we want our numbers to be such as can be used for counting common obj ects , and this requires that our numbers should have a definite meaning, not merely that they should have certain formal properties . This definite meaning is defined by the logical theory of arithmetic . 7

By contrast , the Russellian analysis not only jibes with the more specific ordinary notion that arithmetic truths are a special kind of truth , but with the more obvious idea that some arithmetical statements, such as "7 + 5 = 1 2' ,are true, while others , such as '7 + 5 = 13' , are false. There is another, and decisive , respect in which the Russell re­ construction fits the ordinary context. Perhaps it is no more than an elaboration of the last mentioned point. Measuring and counting are the two basic applications of arithmetic to empirical situations. Our ordinary 330

use of arithmetical concepts in measuring and coun ting tak e for granted that such concepts have a definite meaning and that certain arithmetical statements are true. Counting is a rather obvious case. What we do is establish a 1- 1 correlation between a set of numbers and, say, a set of people. It is understood that the numbers are arranged in the standard way; that is, that certain arithmetical truths hold . We do not treat the arithmetical terms and statements as uninterpreted sign designs which we interpret in terms of signs referring to people and relations among them. When we count a group of people consisting of Jones, Smith, and Brown, we do not interpret the sign · 1 ' in terms of the sign •Jones' or hold that • 1 ' refers to Jones or · means' Jones ( or "Jones' for that matter) . The same is true in cases of measurement. Consider a simple case of measuring, where we merely establish a rank ordering among minerals, in terms of the relations harder than and equally hard, and where these relations are specified in terms of a scratch test. What we do is discover that a set of numbers and the relations > and =, taken in the standard arithmetical way, share certain logical properties with the minerals and the relations harder than and equally hard. 8 This enables us to coordinate to each statement about a mineral, or kind of mineral, being harder than another ( or equally as hard as another) a statement about a number being greater than another ( or a statement of numerical equality) . We do not treat the arithmetical statements as abstract, uninterpreted sign sequences and the empirical statements about the minerals as interpretations of them. If we speak of interpretations at all in such a context, we might think of the arithmetical statements and the empirical ones about the minerals as both being interpretations of an abstract set of sentences. Thus, we discover that the two systems of statements share 'certain formal properties'. Such a set of abstract sentences, taken in a certain context, would specify what it is to be a rank ordering, just as a set of formalized or abstract axioms can be taken to specify what it is for an interpreting system to be a group in mathematics. The use of arithmetic in both counting and measuring dramatically contrasts with the application of a geometry in empirical matters. In applying a geometry, we interpret it. In other words, the ordinary context for geometrical statements reveals that we take them to be either empirical statements or sentential forms belonging to abstract axiomatic systems, which have certain logical and mathematical features. Thus, to apply a geometry is to give an interpretation to an abstract axiomatic system, which is quite unlike what we do in counting and measuring. This shows that the ordinary context for geometrical statements would be ignored or distorted if we took a mapping of, say, Euclidean plane geometry onto PM to be an analysis on a par with Russell's logistic reduction of arithmetic. 9

33 1

The philosophical problems that arise in connection with geometry 'What are geometrical statements about? What kinds of truths are the axioms and theorems of a geometry? How do we reconcile alternative geometries ? ' - can be answered, and must be answered, in terms of the distinction between a geometry as an abstract axiomatic system and a geometry as a set of physically interpreted laws or hypotheses. 1 0 Taken in the one sense, no ground of truth is required for the statements of a geometry, since one deals with uninterpreted sign sequences and not statements that are true or false ; taken in the other sense, we deal with merely another set of empirical laws or purported laws. Such a resolution of the philosophical problems raised by geometries fits with the ordinary context for geometrical propositions. We can even put it more emphatically. The fact that there are alternative geometries in the ordinary mathematical context requires a solution along the lines we j ust considered. There is no other way of resolving the apparent conflict between Euclidean and non-Euclidean geometries (as well as the same type of conflict among the various non-Euclidean geometries). To suggest that each geometry is about a special type of idealized set of entities (Euclidean triangles, Riemannian triangles, etc. ) as arithmetic is about a unique sort of entity-numbers-transforms an extreme form of Platonism into a kind of philosophical lotus-eating. This gambit, I suspect, we may reject without further discussion. It does, however, reflect detrimentally on the view we considered earlier which rej ected the Russellian reduction of arithmetic to logic on the ground that 'numbers are numbers and not anything else' . Moreover, note, again, the similarity between such an extreme form of Platonism and the Wittgensteinian advocacy of the logical independence of different language games (and the implicit holistic theory of meaning he advocates). Is there ·really' a difference between inventing special entities for every language game to be about and insisting on the unanalyzability of any system in terms of any other? More on this point shortly . In the ordinary context there are alternative geometries. There are not, in any similar or reasonable sense, alternative arithmetics. 1 1 Thus, to treat arithmetic along the lines that one may use to resolve the philosophical problems raised by geometries would be mistaken. This fact not only strongly supports the Russellian ·reduction' of arithmetic to logic,but it provides the careful logicist with a satisfactory basis for rejecting any similar reduction of a geometry to logic . 1 2 We have considered some ways in which the Russellian analysis is supported by features of the ordinary context of arithmetic. There is yet another point in favor of Russell's analysis, but it is, perhaps, more problematic. Recall our earlier discussion of the reduction of some scientific theories to others. In some cases of such reduction the axioms of

332

the theory which is red uced will be correlated to theorems of the reducing theory. This, of course, strengthens the sense of the term ' reduction', since there is an obvious sense in which theorems are derived from axioms. On Russell' s program, the axioms of Peano's system, whether taken as an abstract axiomatics of which ordinary arithmetic is an interpretation or as an organized version of ordinary arithmetic, correspond to theorems of PM . This means that his reduction is one in the stronger sense that I j ust mentioned. This can be taken to answer yet another way of taking the question about grounding the truths of ordinary arithmetic. For, one sense of grounding a set of truths is to specify other truths from which they logically follow. With all the qualifications implied by our discussion so far, Russell may be said to have grounded arithmetic in this sense as well. That is, he grounded the truths of arithmetic in a sense in which some of the truths of logic have no ground . Alternatively, logical truths and concepts might have an ontological ground in a sense in which those of arithmetic do not , if the former, not being 'reducible' in turn, are taken to be ontologically grounded by basic entities and relations among them. Here we make use of the idea that the primitive concepts of a metaphysical position reveal its ontology. Some further comments may be made about the alternatives to Russell's analysis. Concerning the gambit that seeks to establish a unique ontological status for numbers, we might note that the style of argument employed against Russell's analysis of the ontological problems of arithmetic would also lead one, if he were consistent , to rej ect the 'construction' of other numbers on the basis of the natural numbers as well as Russell's avoidance of possible or fictional entities, like Pegasus and the king of France in 1 969, by means of the theory of descriptions. Thus, j ust as one can seemingly rej ect ' 7r + r 5r = 1 2/ as an analysis of '7 + 5 = 1 2' on the ground that they do not mean the same thing, one can rej ect "(3.x) [0x & (y) (0y :) y =x) & '4Jx] ' as an analysis of 'the 0 is '4' ' on the same ground. One might then take expressions like 'Pegasus' and 'the present king of France' as primitive or incapable of definition or analysis, j ust as one may take some arithmetical expressions to be primitive. If one philosophizes within the restrictions Russell imposes on himself, one can be led to recognize, in the fashion of Meinong, entities corresponding to such primitive notions, assuming one treats sentences involving them as meaningful assertions, which are true or false on the basi s of certain facts. 1 3 If one does not so philosophize, he, in effect, gives up his gambit and transforms it into the Wittgensteinian alternative we also considered . This comment requires some unpacking. The key to dealing with ontological or metaphysical issues is to be concerned with language in its referential use. Hence, we are concerned 333

with how language is connected to the world so that statements are descriptive or true of it. One obvious, simple, and cogent way of establishing the link between language and the world is to have the primitive terms and categories of a language refer to objects in and indicate features of the world. Here, one may fruitfully think in terms of a schematic or idealized language constructed solely for the purpose of reproducing or reconstructing or, if you will, containing correlates of the ordinary truths of fact, logic, and mathematics. An ontology or metaphysical position is specified by the things one takes one's terms to be linked to and the ways in which one connects the terms. This is one crucial point that is behind the so-called reference theory of meaning.To accept primitive terms but refuse to recognize a corresponding entity or feature of reality is to refuse to play the philosophical game. 14 One way of attempting to make such a refusal intellectually respectable is to attack the reference theory of meaning and advocate a holistic or contextualist theory of meaning as an alternative. This is the sum and substance of the Wittgensteinian gambit. But it is one thing, and a legitimate one, to hold that the ordinary context and use of terms provide their 'meaning'. It is another thing, and an illegitimate one, to confuse this sense of · meaning' with the advocacy of a holistic theory of meaning as a solution to the ontological problems. Actually, this latter move has different variants. One can take it, as we noted earlier, to be an explicit refusal to consider philosophical probiems.Or, a philosopher can, in the style of something like Bradley's absolute idealism, consciously accept the metaphysical pattern of holism and introduce some sort of Absolute as the ontological ground of 'everything'.Alternatively, one can speak, like Quine, of tying language as a whole to "experiences' . Here, it is characteristic to speak of meaning only in terms of the whole system, but, nevertheless, Quine also speaks of anchoring parts of the language to experience. If we probe into this notion of tying or anchoring, it becomes apparent that the holistic talk about meaning merely avoids the issues, since we return to the question of the referenee of language . Whether we talk about reference as meaning or not is really beside the point and not worth arguing about. Thus, the Quinean gambit can be taken to implicitly refuse to take up the issues, even though Quine seems concerned with "ontological commitments'. The point is really elementary: to raise ontological issues in connection with 'the meaning' of language is to ask about the connection of language to what it is about. One who insists on speaking of "meaning' in such a way as to explicitly or implicitly avoid such questions about the referential use of language avoids, not dissolves, the issues. These comments should help to explain why the advocate of "numbers are numbers' should be led either to accept possible entities like Pegasus or reject ontology altogether, via the Wittgensteinian route; they also further explain what 334

is meant by saying that the Wittgensteinian critique of Russell is irrelevant. There is yet another feature worth noting about what I have characterized as the Wittgensteinian critique . In structure and style of argument it is basically the same as the extreme Platonistic critique. Yet, an advocate of it rejects ontological issues and positions. In spite of this, if we project an ontology onto it, by considering its primitive concepts and features to be ontologically significant , it will not only share a style of argument but an ontology with the extreme form of Platonism. Need one bother to paraphrase the comment to the effect that he who ignores history is forced to relive it, or, in this case, reproduce it? We may also note here that what I have called an extreme form of Platonism need not preclude taking terms for some natural numbers as construed or defined in terms of others. Thus, '2' might be held to be definable in terms of " 1', 'successor' , and the notion of natural number. In other words, a modification of the view might involve taking, in effect, only Peano's primitives (or an alternative set of primitive concepts) as ontologically significant. Along this line, one might hold that '2' can be taken to mean "the successor of 1', even though ' 1' cannot be taken to mean "the class of all first level unit classes' and' + 2' cannot be taken as indicating a class of pairs of natural numbers. Another alternative to Russell may be briefly dismissed . Ayer, 1 5 following a line elegantly stated by Hahn, 1 6 has argued that arithmetical truths are true by definition. While attractive, the line is incoherent. � Either all arithmetical terms will be defined or not. If they are, then the thesis reduces to Russell's or something like what Russell did. For, every arithmetical truth will then be stateable without using any arithmetical terms. It is then completely misleading to speak of definitions alone as the ground of truth for arithmetical statements. //they are not, then one is left with some statements employing arithmetical terms as primitives. He must then take such statements to 'define' such terms. But this is clearly "definition' in the sense of 'implicit definition', and is, in effect, no different from the advocacy of the Peano postulates, taken as an abstract axiomatic system, as a "philosophy of arithmetic'. This, as Russell showed, is unsatisfactory. One may be dissatisfied with my purported defense of Russell, since he may feel that all that I have done is stipulate that the sort of mapping Russell constructed provides a sense of meaning for arithmetical statements and, since it does, he has'reduced' arithmetic to logic. Thus, one may feel that I have done no more than repeat, in a roundabout fashion, the claim of Hempel's which I criticized above. 17 I do not think that such a criticism is justified. 1 8 Nevertheless, I understand the sense of dissatisfaction. For, taking the Russell 'reduction' to be a mapping that is

335

cogent, on the grounds I appealed to, is not nearly as strong a claim as Russell appeared to make. But we must note that what goes for Russells's reconstruction of arithmetic goes for all ontological analyses. It is then in a very limited sense that a metaphysical position purports to tell us what is 'really' what. The logistic reduction of Russell differed from that of Frege in a significant way. Frege took numbers to be classes of concepts. On Russell's no class theory, numbers do not exist, since classes do not exist and numbers are, so to speak, construed as classes of classes. Thus, Russell proposed an analysis of arithmetic truths within an ontology that conforms to an empiricist principle of acquaintance and that does not acknowledge arithmetical entities. In my 1956 paper'Peano, Russell, and Logicism' I noted the obvious fact that Russell provided one, among many, interpretation of the Peano postulates, taken as an abstract system. In his 1965 paper 'What Numbers Could Not Be', Benacerraf sought to use such a simple point, which Russell and Wittgenstein recognized, to argue that numbers cannot be Russellian classes. His argument makes two errors. First, it fails to recognize that Russell did not take numbers to be classes or anything else: he eliminated numbers as entities. Second, the fact that there are alternative ways to introduce arithmetical statements does not show, as I noted in 1956, that one cannot use one purported analysis rather than another. What it does show, as I had argued, is that the construction of an interpretation of the Peano postulates in a logistic system does not suffice as an analysis. One must provide additional philosophical arguments. Thus, an analysis that makes use of the null class as a basic entity is suspect on philosophical grounds, however formally adequate it may be . Russell's analysis has the merit of accomodating arithmetical truths, as a form of necessary truth, within an empiricist framework at no ontological expense. It thus conforms to Occam's razor as a principle guiding ontological analyses. That there may be other equally viable alternatives is not an argument that no alternative is viable. All that is established is that there are numerous ways to avoid introducing arithmetical entities while accomodating the truths of arithmetic in an adequate ideal language. Notes 1

2

For a detai led discussion of abstract axiomatic systems and interpretations of them see my 'Axiomatic Systems , Formalization, and Scientific Theories', in Symposium on Sociological Theory, ed . L. Gross ( New York: Harper & Row . 1959) , pp. 407-436 . To speak of coordinating the universal white to the property white may well be misleading

336

here. One may hold that the property is a universal. Thus, one coordinates the metaphysical concept of a universal to the ordinary concept of a property. He does not correlate one thing with another. That is , one coordinates different concepts (notions , meaningful expressions) which are embedded in different contexts, j ustas in the case of the obj ect o and its analysis. However , some philosophical gambits involve the coordination of different things, a bare substratum , o 1 , to an object , o. for example. Such positions literally introduce special things in order to analyze ordinary things. There is much to be clarified here , but that would involve an essay in itself. 3 For a further discussion of these issues see my ' Intervening Variables, Hypothetical Constructs , and Metaphysics· , in Current Issues in the Philosophy of Science (New York : Holt, Rinehart & Winston , 196 1), pp. 448-460 ; 'Physicalism, Behaviorism, and Phenomena' . in this volume , pp. 37 4 f. and G. Bergmann , Philosophy ofScience (Madison : UniversitY. of Wisconsin Press . 1957) . 4 This is not to say that such explanations and explorations of the ordinary context are not relevant to proposed ontological analyses. For surely a presupposition of an adequate analysis of the ordinary context is an understanding of its subtleties. Moreover, some philosophical disputes are certainly about what belongs to the ordinary context. For example, see the dispute between D.A. T. Gasking, ' Mathematics and the World' , and H .N. Castaneda . 'Arithmetic and Reality' , as to whether or not there are alternative arithmetics as there are alternative geometries. Both papers are reproduced in Philosophy of Mathematics. ed . P. Benacerraf and H. Putnam (Englewood Cliffs , N . J . : Prentice-Hall , 1964) , pp. 390-4 17. For some related comments on the paradox of analysis see G. Bergmann and H. Hochberg , 'Concepts', Philosophical Studies, 8 ( 1957) : 19-27 . 0 I have attempted to do so for the theory of descriptions in 'Strawson and Russell on Reference and Description· in this volume , pp. 105 f. 7 B. Russell, Introduction to Mathematical Philosophy (London : Allen & Unwin, 1953), p . 10. 8 See G. Bergmann and K.W. Spence , 'The Logic of Psychophysical Measurement' , in H. Feigl and M. Brodbeck , Readings in the Philosophy of Science (New York : Appleton­ Century-Crofts , 1953), and C . G . Hempel, Fundamentals of Concept Formation in Empirical Science, International Encyclopedia of Unified Science (Chicago : University of Chicago Press, 1952), vol. 2, p. 7. 9 Unfortunately, some logicists of a formalistic bent have thought that just because Russell constructed a mapping he solved the philosophical problems of arithmetic. The embarrassing gap is sometimes covered by saying that Russell reconstructed the meaning of the arithmetical terms and statements in "the logical, not the psychological sense of the term 'meaning'''. C. G. Hempel , 'On the Nature of Mathematical Truth ' , in Benacerrafand Putnam , Philosophy of Mathematics, p. 375. to This is spelled out in Hochberg. 'Axiomatic Systems, Formalization , and Scientific Theories', in Symposium on Sociological Theory. 1 1 Gasking has attempted to argue that there are alternative arithmetics and Castaneda has neatly pointed out how silly some of his arguments are. Gasking and Castaneda , in Philosophy of Mathematics. 12 Some logicists, like Quine , are so entranced with th e formal possibilities of a mapping that they put arithmetic and geometry in the same category and 'reduce' both to logic. W . V. Quine , 'Truth by Convention' , in Benacerraf and Putnam, Philosophy of Mathematics, p. 339. If one becomes truly intoxicated by such mappings, he may conclude that all statements are reducible to logic and hence that there is no real difference between logical truths and empirical truths. This is probably one underlying cause of Quine's celebrated attack on the synthetic-analytic distinction, proxy functions notwithstanding.

337

13 14

15 16

17 18

This is a long story in itself, and hence what I have said is cryptically, rather than accurately, put . Just as Quine refuses to play when he holds that a nominalist may admit primitive predicates so long as he does not take them as substitutable for variables or at the level of language where one has quantifiers. Thus, his celebrated ontological criterion 'to be is to be the value of a variable' is merely a way of verbally , as opposed to 'really' , avoiding a commitment to universals. A.J . Ayer , Language, Truth, and Logic (New York : Dover, 1952), p. 82. H. Hahn, ' Logic, Mathematics and Knowledge of Nature', in 20th-Century Philosophy: The Analytic Tradition, ed . M. Weitz (New York : The Free Press, 1 966), pp. 222-235 . Perhaps, however, Hahn only means to advocate the Russellian analysis and is not stating a view similar to Ayer's. See n . 9 above. If nothing more, being clear about just what Russell has done and why it is philosophically relevant is one thing ; to cover up a philosophical problem, or pass it by, with the use of a phrase like 'the logical meaning' or 'a logical reconstruction' is another .

Properties , Abstracts , and the Axiom of Infinity

In Meaning and Necessity Carnap considered a schema where we can purportedly take the predicates to refer to either classes or properties. His system is misleading and his claim incorrect, but in a very instructive way. Carnap introduces lambda abstraction for forming defined predicate expressions so that ' (h )( . . . x...)' stands for 'the property of x such that. . . ' . In short, he uses definite descriptions so that, for example, the property of being R and S, where 'R' and 'S' are predicates of the schema, is indicated by

(1) (Lj)(x)[fx = (Rx · Sx)].1 However, by so using such an abstract he precludes our taking predicates to stand for properties. The use of (1) implies that there is one and only one property such that it is had by those things having R and S. Thus, we could not allow for coextensive but different properties in such a case. This means that the use of (1 ) , and hence of lambda abstraction, forces us to take the predicates defined by such means as referring to classes, not properties. 2 The lesson is simple. To acknowledge defined or 'constructed' predicates as standing for properties we have to introduce other means than that embodied in Carnap' s use of lambda abstraction to form such predicate expressions. One obvious way, paralleling Carnap's symbolism, would be to introduce 'parodies' of the logical connectives so that

would be used and be transcribed, respectively, as being R or S, being R and S, and being not R. While we could then acknowledge such truths as,

= =

(2) (x)[RvS(x) (Rx v Sx)] ( 3 ) (x)[R·S(x ) (Rx·Sx)] (4) (x)[R(x) = -Rx] 339

such statements would not be consequences of definitions, as they would be if we took ( 1 ) to define ' R"S' , and corresponding abstracts for the other defined predicates. Not employing such abstracts, our complex or constructed predicates are not defined signs, and hence not eliminable in (2)-(4). (2)-(4) may then be taken as instances of axiom schemata employing the 'parodies' of the connectives as'predicators', i. e., signs used to form complex predicates out of simpler predicate expressions. One consequence of our rejecting lambda abstraction is that sentences like

(5) 0( R·S)

= 0[(tf) (x) (fx = (Rx· Sx)) ]

would not hold. For , in claiming that R "S has 0 we are not claiming that R "S is the one property had by all and only those things that have both R and S . A question that immediately arises concerns the rules for the identity of properties indicated by complex predicates formed by means of the 'predicator' devices. Given that we can form both 'R"S' and "SR' are we to hold that 'RS = SR'? One could adopt a general rule, (R 1 ) , so that where 'A' and' B' are metalinguistic predicate expressions,

A=B holds if and only if

(x) [ Ax

= Bx]

is a theorem (or L-truth in Carnap's sense) of the schema. Thus, 'RS = SR' will hold since (6) (x) [ R ·S(x)

= S · R(x) ] ,

will be a theorem. (6) will be a theorem as both (3) and ' (x) [ S ·R (x) (Sx·Rx) ] ' will be among the axioms or theorems. But such a general rule conflicts with another requirement of a theory of properties. Properties may be co-extensive but not identical . A weak form of fulfilling such a requirement would be to omit an extensionality axiom from the proposed schema. Yet, one may reasonably hold that to have a genuine schema providing a logic of properties we require a stronger condition, such as:

=

( P 1 ) (3f) (3g)[(x ) (fx 340

= gx) -f=t= g] .

An intui_tively simple way of satisfying (P i ) is to hold that where "f=l=-g' , then "f f=I=- g-g· . Thus , we would have.

and similar claims for other connectives, or, rather, their "parodies' . B u t , given (P-) we cannot consistently adopt the rule (R 1 ) , since by the axiom schema for (3 ) we would have

= (fx -fx) ]

(x ) [f l(x) and (x ) [g 'g(x )

= (gx ·gx)]

and , conseq uently, by the axiom schema for (4) and standard logic, we get

(x ) [(fx]x )

= (fx· -fx) ]

and

(x ) [ (gx·gx)

= (gx · -gx)] .

We would thus arrive, by standard use of '=' and the standard theorem

(x ) [(fx ·-fx ) at

(x ) [ff(x)

= (gx · -gx) ]

= g-g(x)]

and , hence, by R 1 conclude that 'ff = g- 'g' holds irrespective of whether 'f=g' holds. We thu s contradict (Pi ) , if we adopt R 1 in addition to the patterns (2)-( 4) and a standard logical framework for the connectives and quantifiers. Hence, retaining (P 1 ) and (Pi ) requires replacing (R 1 ) . 3 B efore considering a way of doing that we may note a relevant but mistaken argument that Carnap directs against Russell' s views in

Principia Mathematica.

Carnap's argument is directed against a feature of Russell' s 'no class' theory of Principia. According to Principia, but ex pressed in Carnap's more readable symbolism, the no class theory involved the elimination of class expressions by considering them to be contextually defined symbols as follows: 341

' . . . i(fz) . . . ' for '( 3g)[ (x)(g(x)

= f(x)) · . . . g . . . ] '

where the dots indicate any context for the relevant expression. Carnap finds such a definitional schema problematic. 4 His argument depends on taking "as premises the following two sentences . . . (i) (ii)

=

' (x) (Fx · Bx Hx) ' , or briefly, 'F· B 'Fi· Bi =l= Hi' .

= H' .

These sentences say that the property Featherless Biped and the property Human are equivalent but not identical. Hence they are true". 5 On the basis of taking (i) and (ii) to be true Carnap proceeds to consider (iii) 'i(Hz) = Hi' . (iv) 'i(Hz) =I= Hi' .

= =

' (3g)[g H· (gi = Hi)] ' . (vi) '(3g)[(g H· (gi =I= Hi) ] ' . (v)

with (v) and (vi) taken as the expansions of (iii) and (iv) respectively, and concludes that as (v) and (vi) are both true so are (iii) and (iv). 6 While this is not a formal contradiction since (iii) and (iv) are not the negations of each other due to their expansions , it is, as Carnap sees it, paradoxical. Carnap also contends that similar results can be reached for the system of Principia without the use of nonlogical constants. He assumes that (P 1 ) holds for Principia, since it "is, no doubt , true on the basis of the interpretation intended in [P . M. ] ; this work itself mentions the example of the properties Featherless Biped and Human" .7 From (P 1 ) he derives ' ( 3.f) [ i(fz) = fi . (i(fz) =I= fi)]' , which he finds to be "not actually self­ contradictory" but which "looks as if it were". 8 Moreover , this latter sentence "shows again that the way the class expressions are introduced by Russell's definition is not quite in agreement with the intended purpose". 9 My concern here is only with the premises of Carnap's arguments. The claims that (ii) and ( P 1 ) hold for Principia, with the added constants in the case of the former. Consider (ii). It is elliptical for , (ii ' )

(h)(Fx · Bx) =I= H,

ignoring the specialities of Russell's notation. If lambda abstraction is then construed in terms of definite descriptions (ii '), in turn, is elliptical for,

(ii " ) ( tf)(x)[fx 342

= Fx· Bx] =I= H.

But, given that (i) is true, this latter sentence is false, since the property of being a featherless biped is the one property had by all and only those things that are featherless bipeds. Since (i) is assumed to be true, it follows that the property of being a featherless biped is identical with the property of being human. Alternatively, if one does not construe lambda abstraction in terms of definite descriptions along the lines of ( 1) and, in this case, (ii' ' ) , but takes (ii' ) to invblve a basic notational device of the system then he must formulate a theory of properties. He can not just gratuitously assume that sentences like (2)-( 4) hold, since he is using the logical connectives with predicates, to form new predicates, in ways that are not merely elliptical for their use with sentences. The connectives, so used, become the parodies of the logical connectives discussed above. Neither Carnap nor Russell offers such a theory. In fact, Carnap appears to overlook the difference between taking (ii' ) as basic and taking it as elliptical for (ii' ' ) . A similar point holds about (P 1 ). In spite of what Russell may have said in the passages accompanying the formal text, (P 1 ) does not hold for Principia. In fact, in the second edition, there is an explicit extensionality axiom for first order functions that would directly contradict (P 1 ) , with the latter taken as holding for first order functions. 1 0 Carnap' s objections are thus mistaken, but point to the problem at hand . Moreover, he is correct, I believe, in implicity focusing on (Pt) as the basis for a theory of properties. Finally, Carnap' s concept of intentional isomorphism, formulated for other purposes, is suggestive for dealing with the problem posed by the rejection of (R t ) . Consider a schema with a set of primitive predicates; (a)Fl F2 , . . . Fn , · · · ' such that no two predicates stand for the same property. In short, for any two predicates of (a) , say ' F/ and ' F/

FI =I= F.J holds. We will limit the functional connectives of the system to negation, disjunction, and conjunction and use the same logical signs as parodies to form predicates from predicates in the manner of Carnap. With'R' and 'S ' in (2)-( 4) taken as variables, we can adopt (2)-( 4) as axiom schemata. Specifying when we take different predicates to stand for the same property involves making ' arbitrary' choices or choices dependent upon one's intuitions and purposes. How one puts the matter will depend on how favorably one looks upon the notion of a property, as opposed to a class. My choices here are guided by two themes: first, suggestions

343

involved in Carnap's notion of intentional isomorphism that suggest · 'reasonable' ways of taking complex coextensive properties to be different ; second, suggesting a modification of Russell's logistic thesis that involves taking natural numbers to be properties in such a way as to do without an appeal to his Axiom of Infinity. The simplest way to arrive at a calculus for properties would be to hold that where 'A ' and 'B' are predicate expressions 'A = B' holds if and only if 'A ' and 'B' are the same expressions.This would mean that 'f'iF2 ' and 'Pz.F1 ' would stand for different properties as would 'F1 ' and 'F1 ' . Such treatment of predicates and properties might appeal to one on the ground that it provides a prima facie basis for handling certain so-called intensional contexts.One could not infer that having a belief expressed in terms of one predicate implied the having of a belief expressed in terms of a different, but logically equivalent predicate, since the predicates would stand for different properties. But such a strong requirement for identity runs counter to the notion that the mere rearrangement of the same signs in expressions like 'F'{.F/ and'Pz.F1 ' is significant!y different from the use of different sign patterns, as in 'F1 · F/ and 'F'{_F2 ' , or in 'F/ and 'F'{_ F/ . Without trying to justify, but merely attempting to satisfy this'significant difference', one might think of identifying properties along the following lines. Where 'A ' and' B' are appropriate metalinguistic expressions then:

(I) (II) (III) (1) ( 2)

Where the predicate expressions are elementary predicates A = B if and only if the predicates are the same. Where one predicate expression is elementary and the other is not then A -=I= B . Where both predicate expressions are non-elementary, A = B if: The predicates are of the form A I and B1 , and A 1 = B 1 ; the predicates are of the form A �A 2 and B� B 2 or of the form A iA 2 and B 1 • B 2 and A 2 = B 2 and A 1 = B 1 • There is a set of predicates A 1 , A 2 , • . • , A n and A is an arrangement of the set in disJ· unctive form A vA v.. . vA and B is any arrangement of the set in such form. An arrangement is understood to employ all members of the set. If A is a disjunctive predicate and B is a disjunctive predicate of the form A vA n , where A n is a disjunct of A or a disjunction of disjuncts of A , then A = B . Predicates like A and B will be called sum predicates, and the property they indicate a sum property. In lieu of speaking of predicates as we have, one may speak of there being a sum property for a set of properties. X

344

V

Z

(3) (4)

(5) (IV)

, We repeat the condition (2) with • · ' replacing ,v . Such predicates and properties will be union predicates and properties. One predicate is of the form A 1 where the other is A 1 • Also, as a consequence of 2 and 3, where B is either A�A�A 1 or A 1 A 1 A 1 the iterated predicates reduce in that where A isA�A 1 or A 1 A 1 , r�spectively, A = B . But, where A is elementary and B is A v A, A. or AA . then A -=I= B. Thus , while no elementary and non-elementary predicate stand for the same property, repeated iteration does not 'generate' a new property. One predicate is of the form A�(A;A) or of the form (A�AY )YAz and the other is A�A�A2 ; similarly for • ·'. In all other cases A -=I= B .

The above is merely an outline for a schema where the predicates would be taken to stand for properties. As some speak , it thus constitutes a sketch for a 'theory of properties'. Even as a mere sketch, the 'theory' has an interesting consequence. We can construct a logistic interpretation of elementary arithmetic along Russellian lines without appealing to an axiom of infinity. Frege was led to take numbers as classes, rather than as concepts , since classes guaranteed uniqueness. 1 1 Carnap holds that one can take numbers as properties so long as the properties are extensional in the sense that if numbers are properties of properties they apply to a property if and only if they apply to all properties equivalent to the given property. Thus, he construes cardinal numbers as 'properties of properties which are extensional'. 1 2 It appears as if Carnap overlooks the requirement of uniqueness in that if, for example, the number 2 is to be the property had by all properties applying to two objects , then we must establish that there is only one such property. Carnap does not overlook the need to do so. He handles it in a problematic way by employing his version of lambda abstraction and holding that the difficulty of having two properties identified as the number 2 disappears if we construe equality of numbers as equivalence rather than as identity. 1 3 But, while such a reading of '=' for numbers precludes problems regarding the truth of sentences like 'The number of Fis the n umber of G' , where Fand G are equinumerous, it does not solve the problem about specifying the property which is the number 2. He merely stipulates that if there are two properties satisfying the condition for being the n umber 2, the claim that they are identical is taken as true, since , in such a claim, we treat 'identical' as 'is equivalent'. The problem we noted at the beginning of the paper thus arises in his construal of n umbers as properties and is 'solved' by fiat. His doing so not only reflects an inadequate consideration of properties but is based on a 345

further mistake, earlier in his book, concerning the 'naming' of properties by predicates. Carnap purported to show an ambiguity in the 'name-relation', i. e. , that 'is a name of' is ambiguous. 1 4 His argument proceeded as follows. Consider either the predicates ' H' and 'B' (read as 'is human' and 'is a biped') or the abstracts 'i(Hx)' and 'i(Bx) ' and the sentence, S I , '(x)[Hx :J Bx ] ' . Acknowledging S I to be true we can take the predicates and abstracts to stand for either properties or classes. No set of true sentences employing such predicates in a standard system of logic will provide a ground for holding that properties, rather than classes, are the 'nominata' of predicates ( or vice versa). To say the predicates stand for classes is then to adopt one 'mode of speech' which can be consistently carried out ; but so can the 'mode of speech' which involves talk of properties. What Carnap has in mind is that if we consider a list of true sentences like l. Hs 2. sEi( Hx) 3. (x) [ Hx :J Bx] 4. i( Hx) C i(Bx) etc. we can treat them as, respectively ,

P

1 ' . Socrates has the property of being human. 2'. Socrates has the property of being human. 3 '. Anything which has the property of being human has the property of being a biped. 4 '. The property is human materially implies the property is a biped.

or

C

1 ". Socrates belongs to the class of humans. 2 ". Socrates belongs to the class of humans. 3". Anything which belongs to the class of humans belongs to the class of bipeds. 4". The class of humans is included in the class of bipeds.

Consequently. no sentence or set of such on the list of true sentences of the schema, 8 , will determine that we use P or C as the semantical analysis, assuming that the meanings of the connectives and quantifiers are fixed in the standard way. Carnap concludes that this reveals an 346

ambiguity i n the name relation as applied to predicates and in the predicative relation , whether expressed by j uxtaposition alone or by the use of ' E ' . Carnap further rej ects the apparent solution of the addition of an axiom of extensionality to the list 8. We would then have ; (a) (x) [fx = hx] :) f = h and (b) (x) [x E y(fy)

= x E y(hy)] :) y(fy) = y(hy) .

With (a) and (b) we cannot take ' H' or 'x(Hx)' as signs for properties , provided we read ' = ' as ' identical ' , since we assume we can have coextensive but different properties . Hence , one can hold that with (a) and (b) , we are forced to take predicate signs and the corresponding abstracts as signs for classes , not properties . This does not help , according to Carnap , since it presupposes that we take ' = ' as the identity sign . He sees nothing wrong with taking ' = ' as a sign for the coextensiveness of properties , when it occurs between predicates as in (a) and (b) . Then , with ' F read as 'featherless biped' a sentence like

where we assume the left side of the conditional to be true , is true , while it would be false on the standard view about properties , along with the normal interpretation of ' = ' . Carnap thus refuses to fix the sense of ' = ' in formulating his argument . I will return to that point shortly . First, we may note that even on his own terms , Carnap must restrict 8 to the lower functional calculus . If we take 8 to be a higher functional calculus , his argument fails . This is easily seen . The sentences, (d) (x) [Hx

= Fx] = ( 2f) [ 2f( H) = f( F)] 2

and (e) H = F = ( 2f) [ 2f( H)

= f( F)] , 2

with the superscripts indicating variables and quantifiers of the second level , will turn out to be true when classes are taken as the 'nominata' of the predicates ' H' and ' F , but not when properties are so taken . This is so irrespective of Carnap's reading of ' = ' in terms of the coextensiveness of properties as it occurs in ( e ) . For, so reading ' = ' in ( e) merely means that ( e) is read as ( d) , which is false when the predicates stand for properties . 347

Thus, the purported ambiguity disappears if we consider an extended system of logic. This suffices to rebut his attack on the name relation and remove the ground for his subsequent advocacy of the 'intension-exten­ sion' dichotomy. There is a further point. Just as we must take ( d) to be false when we construe predicates as terms standing for properties , if we take (P 1 ) to be true we are forced to construe predicates as terms for properties and we cannot read ' = ' as property equivalence. For, with Carnap's reading of ' = ' as an equivalence sign, (P 1 ) becomes a contradiction. Moreover, even if we confine ourselves to the lower functional calculus with identity, Carnap's refusal to fix the sense of ' = ' reveals the weakness of his argument in two ways. First, he simply rules out by fiat a genuine identity sign in 8 by allowing for the variable reading of '= '. Second ,he virtually forces the definition of 'f = h' as ' (x)[fx hx]' by the two readings he allows for the identity sign. On the basis of such considerations , Carnap apparently feels he is free to read the identity sign, in an extended functional calculus, as the equivalence sign when it occurs between predicates. Thus, we can construe sentences like ' F = G' in terms of the equivalence of properties. What this means is that Carnap acknowledges that we do not have a unique property that is the number 2. Yet , he avoids the problem, in the formal schema, by reading , for example ,

=

expanded as (e2) (3f) [ (g) (g = 2)

= (f = g)]

as (e3) There is a property f such that any property g is equivalent to 2 if and only if/ is equivalent to g. Thus , even though there is not a unique property which is the number 2, the variable reading of · = ' of the schema, as "identity' or ·equivalence' in the metalanguage, preserves the truth of (e 1 ). In short, the uniqueness clause associated with (e ,,) is no longer a uniqueness clause. This, I submit, amounts to treating numbers as classes, only in a roundabout fashion - through the metalanguage, as it were . Carnap no more solves the problem about the uniqueness of numbers , construed as properties, than he solves the problem of how to construct a schema where complex predicates indicate properties. The so-called method of ·extension-inten­ sion' is merely a method of covering up problems by the permitting of

348

vari able readings of schematic formulae . I conclude that Carnap's construal of numbers as properties either is mistaken in that , in spite of what he says , his numbers are still classes or is inadequate in that he does not provide an adequate theory of properties. The e lusive uniqueness condition is readily satisfied by the notion of a union property . Suppose there is more than one property had by and only by all first level properties that are exemplified by two objects . Then the union property of such properties will also be had by all such first level properties and only by such properties . We can construe the number 2 to be the union property of all second level properties that apply to all and only first level properties exemplified by two objects . The natural numbers are thus construed as union properties of first level properties and hence as second level properties of properties. 0 will then be the union property of all properties had by and only by empty first level properties . The problem of uniqueness is thus resolved, but the question of the need for an infinite number of objects still remains. Recall Russell's case of a finite domain of 10 obj ects . 15 Even if we speak of properties rather than of classes , we can not take 1 1 to be the union property of all properties applying to 1 1 obj ects without running into Russell's problem of having 1 1 and 12 be identical . But there is something we can do with properties that Russell could not do with classes . With a finite number of objects , and even with a finite number of elementary properties of the first type , we have an infinite number of non-elementary properties of the first type . Take F1 to be a l!_elementary property of the first type , then we have the predicates , "(F1 • F1) ' , '(F1 • F1 ) • (F1 • F1 )' , etc . , indicating different properties . Having an infinite number of tirst level properties , we can construct the numbers at the third level in Russellian fashion , rather than at the second , without appealing to an axiom of infinity. That is , without needing to make a claim about the number of obj ects that there are , or even about the number of elementary first level properties, except in so far as we need one such property , we have an infinite number of first level properties . One can then construe the notions of successor,posterity, etc . , along Russellian lines . The only remaining problem is the application of number properties to properties of the first level , properties of obj ects . B ut this is easily solved . We can unproblematically allow for a cross-type equivalence relation so that a second level and a first level property are held to be numerically equivalent if and only if there is a one-one correspondence between the set of properties exemplifying the one and the set of objects exemplifying the other. Then , the first level property will be said to be n-membered2 ( a second level property) if and only if the second level property it is numerically equivalent to is n-membered3 (a third level property) . In short , number 349

predicates will be introduced as second level predicates by means of the number predicates of the third level and the cross-type equivalence relation. Schematically,

where 'S2 • ' stands for the cross-type equivalence relation, ' 0 0 2 ' is the second level numerical predicate being defined, and ' 0 0 3 ' is the third level numerical predicate being used in the definition. One cannot, I believe, cogently object that we have assumed an infinity of first level properties and, hence, that we still appeal to a 'non-logical' assumption in such a Russell-style construction. For, that there are an infinite number of non-elementary first level properties is a consequence of a theory of properties: what one may reasonably call a logic of properties. Hence, I do not think I stretch the term 'logical' too far when I claim that it is a logical truth that there are an infinite number of non­ elementary properties of the first type. But, one may object that this is no different from the claim, which Russell requires, that there are an infinite number of first level classes. Is not the latter a consequence of an adequate set theory? In such matters one can only point to the differences and similarities, and there is a significant difference.For a Principia type schema there are a denumerably infinite number of first-level classes if and only if there are a denumerably infinite number of individuals. Thus, assuming (or establishing) that there are a suitable number of first-level classes for a Russell-type treatment of elementary arithmetic involves a corresponding claim about the number of individuals that there are. This latter claim is highly problematic , especially when we recall (1) that a Principia type schema is to provide the logical skeleton for a perspicuous language in terms of which one purportedly resolves a host of traditional philosophical problems, and (2) the interpretation of the undefined signs of such a schema is to be guided by a Russel!ian principal of acquaintance. By contrast, the schema outlined here permits a Russell-type treatment of arithmetic without having to make any claim about the number of individuals that there are. This difference, I submit, gives substance to the notion that afactual claim is not being made. But, whether one agrees or disagrees with that way of putting it, the difference remains. If we introduce signs for classes into the schema, we could also treat numbers as classes of second level properties, rather than as properties. The points to note are that , given the consideration about property identity, we have, first , an infinite set of first level properties, and, second, uniqueness of the requisite properties, the candidates for numbers, as well as of suitable classes. In this connection it is worth noting that so-called 'finitist' constructions, treating numbers as finite classes, 1

350

have no relevance to our concerns, as they too require an infinite number of classes, albeit only of finite classes to deal with the generation of the natural numbers1 6 . The question is how to guarantee 'enough' properties (or classes) in a Principia type schema without assu ming an infinite class of individuals (zero level ' obj ects') . What is interesting in this connection is that on the present construction while we may have only a finite number of such obj ects, the numbers, as properties of the third level, will be infinite properties, i. e. , be exemplified by an infinite number of second level properties. Notes 1

2

3

R. Carnap, Meaning and Necessity, (University of Chicago Press , Chicago : 1947) , p . 3 . Actually . Carnap first introduces lambda abstraction for predicates in t h e lower functional calculus ; thus he does not literally use something like ( 1 ) . Later , he treats lambda abstraction for predicates as primitive and introduces definite descriptions for predicates by means of lambda abstraction . So , in effect ,he employs ( 1) . Ibid. , pp. 38-39. Carnap would hold , in effect , that there i s no problem here since h e will allow for a variable reading of ' = ' , i. e. , it is read as 'equivalence' when we speak of properties . This means that 'the' no longer involves uniqueness. We will return to this theme later. A referee has raised an interesting point by suggesting that (P2 ) is not plausible while (P1 ) is and that (P2 ) would be controversial in that it would conflict with the view that properties may be construed in terms of functions from objects and possible worlds onto truth values. The conflict with such a view , given its problematic nature, is not disturbing. While that issue is not my concern here it provides an opportunity to make a relevant point . On the kind of view he has in mind one may take properties to be classes of ordered pairs , where one element of each pair is the extension of a predicate and the other is a world of the model. To speak of properties being coextensive , but not identical , is then misleading . In so far as the properties are taken to be the classes of ordered pairs no two would be coextensive . All that one means by holding that coextensive properties are not identical is that two classes of ordered pairs have a common element. There is another point . Such a way of taking properties goes along with providing a semantics for a modal calculus . If one speaks of an extensionality axiom for a Principia-type system then one takes it to be something that holds in all domains of the validity theory employed. In a modal calculus it would correspond to the claim that the extensionality condition was necessary. In a corresponding way , to take ( P 1 ) as an axiom in a Principia-type schema is to take it as holding in all domains . Thus , to take (P 1 ) as holding in a context where one speaks of possible worlds would be to take ( P 1 ) as necessary , as holding in 'all worlds' . This would amount to accepting (P2 ) or something equally unacceptable to the referee , since ( P 1 ) will only be necessary if we have different but coextensive properties in all worlds ; but then the properties cannot be classes of ordered pairs as construed above without imposing arbitrary restrictions . In short , for ( P 1 ) to be necessary we cannot cogently construe properties as functions of the kind required. The question comes down to whether we take ( P 1 ) to deny that the extensionality claim is necessary or to assert the necessity of there being coextensive but different properties ; and, in the latter case, whether the claim is that there are two properties which are coextensive in all worlds or whether it is that in each world

35 1

there are two coextensive properties. The claim that there are two properties coextensive in all domains is clearly incompati ble with taking properties as classes of the kind in question . The claim that for each domain there are two coextensive properties is not , in the same sense , incompatible , but to satisfy it requires an arbitrary limit on the domains of the model . This would be like claiming that we can hold that '(3.x) (3y) (x y) ' is a necessary truth or theorem by limiting the domains of the validity theory to those with at least two elements. In effect , by using (P2 ) , I have taken (P1 ) to make the strongest of the three alternative claims.One who obj ects, as the referee does, must take it in the weaker sense. He cannot , then , treat ( P 1 ) as a candidate for addition to a Principia-type schema without arbitrary restrictions on the model . Alternatively, he would express his view , not by adding ( P 1 ) to a Principia-type schema , but merely by omitting an extensionali ty axiom. ( P 1 ) 's 'pla usi bi Ii ty' is thus not a simple matter. Ibid. , p. 147 . Ibid. , p. 148. Recall , here , that Russell 'caps'the individual variable so that 'Hi' as a signfor the function ( or property) is distinguished from the open sentence 'Hx' as a propositional function. Ibid. , pp. 1 48-49. Ibid. , p. 150. Ibid. , p. 150 . Ibid. , p. 150. B. Russell and A . N. Whitehead, Principia Mathematica, 2ndEd. (Cambridge , 1950), p. xxxix . It should be obvious that in so far as Principia did not have an extensionality axiom in the first edition, Russell and Whitehead, by the use of circumflex caps as an abstraction operator, invite the same problem about properties that is invoived in Carnap's discussion. This is really what is behind their discussions that led Carnap to find class abstraction 'paradoxical' in Principia. Thus, the way they avoid seeing that (a) '(y ) { [fi·gi] y = (fy ·gy) } ' is not a consequence of a definition is that they pack such an 'axiom' into their symbolism in that one does not write (a). For, in place of '[tx·gx]y' we would write 'fy·gy'; as in Carnap's lambda conversion for the lower functional calculus. G. Frege , The Foundations ofArithmetic (Blackwell , Oxford : 1953) , p. 80e . Carnap , op. cit. , p . 1 15 . lbid. , p . l l 7 . Ibid. , pp. 100- 106 . B. Russell , Introduction to Mathematical Philosophy, (Allen and Unwin , London: 1953), p. 24. See W. V. Quine , Set Theory and Its Logic, (Harvard University Press, Cambridge : 1963 ) , pp. 76-77.

*

4

5

6 7

8 9 10

11 12

13 14 15 16

Of Mind and Myth

In recent years Professor Quine has rejected mental entities. Yet, his resulting materialism (physicalism) is a qualified one as he further holds that physical objects are ontological "myths'. This somewhat paradoxical view is a (;\')nsequence of a residual phenomenalism in an eclecti c metaphysics that also embraces elements of pragmati sm, holism, materialism, neo-Kantianism, and analytical philosophy. To see this we will trace Quine· s development from an apparent phenomenalist to a " physical mythologi st' . This development may be divided into three stages. In 1 948 we find Quine holding that phenomena are of the utmost significance for epistemology. The phenomenalist 'conceptual scheme' is "epistemologically fundamental'. Physical objects, on the other hand, are only convenient myths, and the physicalistic conceptual scheme is scientifically fundamental rather than philosophically so. At this stage science and philosophy have yet to be merged. One ground for such a merger is soon to be provided by behaviorist psychology. In 1951, the second stage, we find Quine attacking ph enomenali sm by castigating attempts at phenomenali stic reduction, the thesis that physical object statements can be reconstructed in a phenomenali stic conceptual scheme. It takes only one more year for Quine to advance to the third and final stage by adopting Watsonian behaviori sm as a philosophy of mind. Embraci ng behaviorism, Quine resolves whatever doubt he may still h ave had by expli citly rejecting ph enomena. Yet there are still some ph enomenali stic hangovers. They lurk behind his continued reference to ' mythology' as well as his use of the term ' experience' at critical points. To a detailed account and analysis of these stages I will shortly turn. First, I will briefly consi der some notions that are germane to the analysis.

I The notion of an improved or 'ideal' language is familiar in contemporary philosophical discu ssion. A formalism like Principia Mathematica supplemented by adding certain classes of descri ptive signs (and the so-called axiom of extensi onality to its primitive sentences) can, upon interpretation, serve as an illustration. Some philosophers

353

conceive of philosophical analysis as consisting of informal commonsensical discourse about the structure and interpretation of such ideal languages. One thus reconstructs the traditional philosophical problems as questions about such languages.Not all formalisms are ideal languages. In order for an interpreted formalism to qualify as an ideal language it must fulfill three conditions.(1) It must serve as an adequate tool for the analysis of all philosophical problems. That is, all philosophical propositions can be reconstructed as statements about its structure and interpretation. (2) It must, in principle, contain transcriptions of anything nonphilosophical we might want to say. Thus it must enable us to account for all areas of our experience.For example, it should not only contain (schematically) the way in which scientific behaviorists speak about mental contents but it should also reflect the different way in which one speaks about his own experience and that of others. And, it must then show how these two jibe. (3) It may not contain the transcriptions of any of those problematic propositions and uses which give rise to the philosophical problems. What is involved in and intended by (1), (2), and (3) I hope to make more explicit as we proceed. Quine is not hostile to philosophizing by means of ideal languages, though, as far as I know, he seldom speaks of them as such.1 However, he would not consider (1 ), (2), and (3) as the criteria for an interpreted formalism's being an ideal language. Rather, we see his alternative criteria when he writes: Philosophy is in large part concerned with the theoretical non-genetic underpinnings of scientific theory ; with what science could get along with , could be reconstructed by means of, as distinct from what science has historically made use of. If certain problems of ontology , say , or modality, or causality , or contrary-to­ fact conditionals , which arise in ordinary language , turn out not to arise in science as reconstituted with the help of formal logic , then those philosophical problems have in an important sense been solved: they have been shown not to be implicated in any necessary foundation of science . Such solutions are good to j ust the extent that (a) philosophy of science is philosophy enough and ( b) the refashioned logical underpinnings of science do not engender new philosophical problems of their own . 2

Since we have two conflicting sets of criteria for an ideal language I shall compare and contrast them by pointing out some consequences of each. Since one's philosophy is judged by its fruits, I shall attempt to show that Quine 's position suffers by such comparison.Quine 's (b) seems similar to (3), even though his (a) conflicts with (1 ). However the agreement between (b) and (3) is illusory. Let us see how they differ. Quine 's well known conception of ontology, condensed in the cryptic phrase 'to be is to be the value of a variable', makes the existential 354

operator the key to ontology. Existential statements in his proposed ideal language involve ontological commitments, or , to put it another way, such statements make ontological assertions. Since, first, he has also told us that the existential operator is a more precise or scientific rendering of the "there is' of ordinary language and, second, it is by use of this single sign that one says such radically different things as (p 1 ) There are no physical objects' and (P.:) There are no rocket ships in Glenview', we may conclude that Quine's conception of ontology results in the collapsing of the ordinary and the ontological (philosophical) uses of 'exist' into the existential operator. This point is made quite obvious by his sometimes speaking of the ontological commitments of ordinary language. By contrast . one might observe that (p 1 ) represents a problematic use that gives rise to philosophical puzzlement, while (p7) represents an ordinary use that does not. This distinction is indispensable for an adequate analysis of classical ontology. For its neglect will lead one to merely reassert, rather than analyze, traditional ontological assertions. Thus in his proposed ideal language Quine simply restates , even though in symbols, one of the traditional ontological positions - instead of , by means of a symbolism , explicating all of them without accepting any of them in their traditional form. This shows, first, the difference between (3) and (b) and, second , the inadequacy of both Quine's explication of ontology and his criteria expressed in (b) . For (b) does not prohibit the assertion, within Quine ·s ideal language, of statements that give rise to philosophical problems , since it permits the reproduction of the traditional puzzling uses of language. A purported philosophical analysis that results in the reassertion of traditional metaphysical statements provides neither an explication nor an analysis. 3 In view of (b) 's permitting the assertion of philosophically problematic statements one may well wonder precisely what Quine intends it to exclude. (b) might well be intended to preclude, first, the paradoxes that interest logicians and mathematicians and, second, all philosophical statements and questions that are not reconstructible as ones for, or about, science. This latter would seem to be implicit in his suggestion that 'philosophy of science is philosophy enough'. Hence classical ontological assertions are permissible if, for Quine, they are the ontological 'hypotheses' of the scientist . The scientist thus provides the answer to the classical ontological questions. We can then readily see how the acceptance of behaviorist psychology as science helps set the stage for the philosophical assertion that there are no mental entities. The behaviorist, we recall, always speaks physicalistically. Qui ne's criterion (b) for a language's being an ideal one is thus an important point in his ultimate rejection of phenomena. His explication of ontology(referred to hereafter as (o 1 )) may be considered to have two parts: first,

355

ontological commitments are made via the use of the existential operator and, second , the answer to the question 'What exists?' is given by the ontological commitments of the simplest conceptual scheme that can accomodate all of science : Ontological questions, under this view , are on a par with questions of natural science . . . Our acceptance of an ontology is, I think, similar in principle to our acceptance of a scientific theory , say a system of physics: we adopt , at least insofar as we are reasonable , the simplest conceptual scheme into which the disordered fragments of raw experience can be fitted and arranged . Our ontology is determined once we have fixed upon the over-all conceptual scheme which is to accomodate science in the broadest sense ; . . . 4

Thus it seems that not only is philosophy of science 'philosophy enough' , but for ontological questions, science is enough . As distinct from Quine's reconstruction of ontology, (o 1 ) , one might propose that the undefined descriptive signs of a philosopher's ideal language give his existents in the ontological sense of that term . In connection with this latter explication , (o2 ) , we note three things . First , we are not concerned with ontological com mitments of ordinary l anguages or sign systems in general . Rather , we are concerned with an explication of classical ontology and , as such , speak only of ontological commitment in connection with an ideal language . Thus a philosopher's answer to the ontological question 'What exists?' is reconstructed in terms of the undefined descriptive signs of his ( explicit or implicit) proposed ideal language . Second , ontology is primarily concerned with kinds of entities , rather than with particular existents , and , further , with existents in the narrow sense of that term which connotes simplicity . That is , one of the traditional ontological motifs is the idea that entities which exist in the ontological sense are neither complexes nor patterns of other entities . This is one reason for making the undefined (rather than all of the) descriptive signs carry the ontological burden . Third , one is not involved in making ontological assertions by using the existential operator, since that sign is not , on this view , the key to ontology . Thus one can distinguish between two kinds of uses - philosophical and ordinary ­ of 'exist' . The former is explicated in terms of the undefined descriptive signs while the latter is reflected in the existential operator. This is not to say either that there are not other metaphysical uses , i . e . , ·subsistence' , which also require explication or that one cannot make further distinctions among the various ordi nary uses . The key point is that all the former will be explicated as questions and answers about the structure and i nterpretation of an ideal language , while the latter will be reflected in the various existential statements made in the language . Hence one 356

does not reproduce within the language the ontological statements that give rise to the philosophical puzzles . We may then consider a difference between a phenomenalist and a realist to be partially explicated as follows . The former would hold that the undefined descriptive signs of his ideal language refer only to phenomena and properties of such ; the latter would insist that such signs refer to physical obj ects and their properties . The traditional question can then be partially reconstructed as a question about the i nterpretation of the ideal language . Its answer is given by a consideration of whether or not the proposed alternative ideal languages do in fact fulfill ( 1 ) , (2) , and (3) . That is , if it can be argued that a proposed ideal language that is physicalistically interpreted cannot fulfill these cri teria but , on the other hand , a phenomenalistically interpreted one can , then the reconstructed question is answered . But notice , in saying all this one neither adopts nor rejects either position in its traditional form . In fact we no longer find ourselves in the predicament of being forced to choose between the m . For , upon reconstruction , the common sense cores of the various traditional positions need no longer be incompatible . Thus one can recover these various ·cores' without either asserting the puzzling propositions or accepting the extravagances of the classical positions . 5

II I n his paper ' O n What The re Is' Quine sets forth the position of the first stage . But simplicity , as a guiding principle in constructing conceptual schemes , is not a clear and unambiguous idea ; and it is quite capable of presenting a double or multiple standard . I magine , for example , that we have devised the most economical set of concepts adequate to the play-by-play reporting of immediate experience . The entities under this sche me- the values of bound variables- are , let us suppose , individual subj ective events of sensation or reflection . We should still find , no doubt , that a physicalistic conceptual scheme , purporting to talk about external objects, offers great advantages in simplifying our over-all reports. By bringing together scattered sense events and treating them as perceptions of one obj ect , we reduce the complexity of our stream of experience to a man ageable conceptual simplicity . The rule of simplicity is indeed our guiding maxim in assigning sense data to obj ects: we associate an earlier and a later round sensum with the same so-called penny , or with two different so-called pennies , in obedience to the demands of maximum simplicity in our total world-picture . Here we have two competing conceptual schemes , a phenomenalistic one and a physicalistic one . Which should prevail? Each has its advantages ; each has its special simplicity in its own way . Each , I suggest , deserves to be developed . Each

357

may be said, indeed, to be the more fundamental , though in different senses: the one is epistemologically , the other physically , fundamental . . . Physical obj ects are postulated entities which round out and simplify our account of the flux of experience , j ust as the introduction of irrational numbers simplifies laws of arithmetic . 6

These passages make it clear that for Quine phenomenal entities are 'epistemologically fundamental' while physical obj ects are 'postulated entities' . The difference seems to lie in the fact that phenomenal entities are objects of direct experience while physical objects are not . Thus phenomena are more fundamental in that they and they alone are directly give n . Physical obj ects , not being given, are introduced ( or postulated) only in order to round out and simplify our accounts of the given phenomena. They are , so to speak , introduced 'for the sake of the phenomena' . Phenomena, on the other hand, are not introduced for the sake of anything . In fact they are not introduced at all : we simply find them in experience . What Quine says seems to lie somewhere on a continuum between Berkeley and the neo-Kantians , yet it is hard to say j ust where . In terms of the explication of ontology expressed in ( o 2 ) , Quine 's characterization of physical obj ects as ' myths' could be explicated by his ( explicitly or implicitly) proposing an ideal language all of whose undefined descriptive signs referred to phenomenal entities and properties of such . The 'names' of physical obj ects , provided there were such , and of the characters they exemplify would then be defined signs . Thus we would have a perfectly clear cut explication of that picturesque phrase , "the myth of physical objects' . Such an explication would reflect an essentially Berkeleyan outlook . But this Berkeleyan alternative does not seem to reflect what Quine has in mind . This is implicit in his use of ·postulate ' as well as explicit in his doubts concerning the definability of physical obj ect terms by means of phenomenalistic ones . If this is the basis of the mythical character of physical obj ects , it would see m that phenomena, the data of immediate experience , are not myths . Not being myths, phenomena are surely existents . Thus in the first stage it seems that Quine is primarily concerned , positively , with securing some ontologically respectable status for physical obj ects � he is not concerned , negatively , with destroying the ontological status of phenomena. Such respectability for physical objects is needed and desired for three reasons . First , due to Quine 's commitment to the epistemological priority of phenomena , he holds that physical obj ects are not objects of direct experience . Hence they are ontologically suspect . Second , even though ontologically suspect , physical obj ects are needed in Quine's conceptual scheme to accomodate the statements of science . For, third , phenomenalistic reconstruction cannot be achieve d . In short , Quine's conceptual scheme 358

must contain physical object terms that make ontological commitments, but such commitments conflict with his phenomenalistic orientation. This shows that in addition to the two parts of (o 1 ) , which we considered above, Quine apparently also thinks that·to be is to be an object of direct experience'. His explication of ontology, at this stage of his thought , is thus really three-fold. Yet this " phenomenalism' puts Quine in a quandary. Since only objects of direct experience exist, and hence are not myths, he, not being a direct realist. must consider physical objects as myths. On the other hand , since a phenomenalistic conceptual scheme cannot literally reconstruct all statements of science, it cannot serve as an ideal language. Hence he must accept a conceptual scheme which · postulates' mythical entities - physical objects. But he does not, as yet, reject directly given existents. Thus, in the first stage Quine apparently adheres to a conceptual scheme committed ontologically to phenomena and to physical objects. This seems to be so in spite of the fact that he contrasts a phenomenalistic scheme with a physicalistic one, and not, explicitly, with a " mixed' one. At this stage, a 'physicalistic scheme' is for him apparently one which accepts physical objects in addition to phenomena. This interpretation is clearly supported , first, by his speaking of associating sense data with physical objects in a physicalistic conceptual scheme and , second , by his analogy between the physicalistic and phenomenalistic schemes on the one hand and the arithmetic of rational and irrational numbers on the other. From the point of view of the conceptual scheme of the elementary arithmetic of rational numbers alone , the broader arithmetic of rational and irrational numbers would have the status of a convenient myth , simpler than the literal truth (namely, the arithmetic of rationals) and yet containing that literal truth as a scattered part . Similarly, from a phenomenalistic point of view , the conceptual scheme of physical obj ects is a convenient myth , simpler than the literal truth and yet containing that literal truth as a scattered part . 7

Thus a physicalistic scheme, in Quine' s terms, contains the phenom­ enalistic one (i. e. , commits one to phenomenal entities) as a part. Yet, due to his phenomenalism he has misgivings; so much so that he is still willing to see to what extent phenomentalistic reconstruction can be achieved . Let us by all means see how much of the physicalistic conceptual scheme can be reduced to a phenomenalistic one ; stil l , physics also naturally demands pursuing, irreducible in toto though it be . 8

From philosophical-epistemological motives Quine is pulled towards phenomenalism; from 'scientific' motives he is pulled towards an

359

ontology acknowledging physical objects. Hence his views contain unrecon ciled tensions. This is inherent in ,v hat I called his threefold explication of ontology: (a) to be is to be the value of a variable, (b) t o be is to be an ontological commitment of the 'simplest' conceptual scheme adequate to the reconstruction of science, (c) to be is to be an object of direct experience. Given the criterion of ontological commitment expressed in (a) and the failure of phenomenalistic reconstruction , Quine's conceptual scheme cannot satisfy both (b) and (c). The adoption of a complex ontology provides neither a solution nor a reconciliation. Moreover, for one who philosophizes, as I believe Quine does, with Ockham' s dictum in mind, a complex ontology can only be a temporary expedient. That this is so is implicit in Quine' s discussion of simplicity. As Quine notes, 'simple' is an ambiguous term. In one sense one may hold that a phenomenalistic scheme is simpler than a physicalistic one in that phenomena are simpler and more fundamental entities. That is, sensa, for example,are neither posits nor composits. A second and third sense of 'simple' arise when Quine speaks of simplifying our account of physics by postulating physical objects. One thing he apparently has in mind is something similar to what some mean when they speak of a linguistic structure or scientific theory being simpler than another in some technical sense concerning the kinds of terms and the forms of statements. Another thing he seems to have in mind is a sense of 'simplify' which is synonymous with 'unify' . J ust as scientists achieve simplification when they unify different areas or theories by, for example, subsumption of one or more theories under a different theory, so we achieve conceptual simplification by associating different phenomena with one physical object. (A further sense of 'simplify' is involved in assigning phenomena to physical objects according to the 'demands of maximum simplicity' . But this is irrelevant to our discussion. ) In a fourth and crucial sense one conceptual scheme may be considered simpler than another if it commits one to fewer kinds of entities. In this sense neither a phenomenalistic nor a physicalistic (in the usual sense, i. e. , one committed only to physical objects) scheme would, as such, be simpler than the other. Both, however, would be simpler than a scheme which acknowledged both kinds of entities. Following Ockham's dictum, Quine would or could be led to seek a simple scheme in this fourth sense of 'simple' . This, I believe, provides a key to his ultimately reject ing phenomena, after once having considered them to be ·epistemologically fundamental'. This point may be seen as follows. As we n oted , the Quine of the first stage is confronted with two ' opposing' conceptual schemes: one phen omenalistic, the other committed to phenomena an d to physical objects. Suppose that both are adequate in that they can both serve for the description of all areas of our experience an d for the reconstruction of

360

all physical object statements of the scientist. The phenomenalistic scheme would be simpler in the first and fourth senses of that term, while the complex scheme would be technically simpler. To vacillate between the two because of the technical simplicity of the one would be to show the grossest neglect for both ontology and epistemology. Yet, as I read the above quoted passages of Quine, it seems that he might be swayed , or at least tempted, by such ·technical' considerations. Be that as it may, he does not have to face such a choice, since a phenomenalistic scheme is not likely to be adequate. It is not adequate because, we recall, a phenomenalistic reconstruction of the scientist' s statements, according to Quine, cannot be achieved. In fact we may even take the third sense of 'simple' that I distinguished - " simple' as synonymous with 'unify' - to reflect this inadequacy. For, if one could reconstruct physical-object terms in a phenomenalistic reconstruction, then one would achieve the · unification' of various sensa with a single physical 'object' without recourse to an ontology embracing physical objects. Thus Quine writes: The physical conceptual scheme simplifies our account of experience because of the way myriad scattered sense events come to be associated with single so-called obj ects ; stil1 there is no likelihood that each sentence about physical obj ects can actually be translated , however deviously and complexly , into the phenomenalistic language . Physical obj ects are postulated entities . . . 9

Physical objects are thus postulated not " merely' for technical simplicity but as a necessary condition for the adequate reconstruction of physics. Hence we would not be confronted by two adequate schemes one of which was technically simpler, the other ontologically and epistemologically simpler. Rather, we would have to choose between an adequate and an inadequate scheme. In this sense there is no real choice. Consequently, as we saw, physical objects must be postulated . Thus the only scheme which might be both adequate and ontologically simple would be one committed to an ontology of physical objects alone. In this manner a ' new' alternative arises and , for one who seeks ontological simplicity and is more concerned with the need to literally reconstruct all statements of science than he is with epistemological problems, the road to materialism is open. But if such a one was once impressed by the epistemological priority of phenomena he must first convince himself that the immediately given and 'epistemologically fundamental' is non­ existent. Quine seeks to convince himself by attacking phenomenalism. At first he develops his misgivings about phenomenalistic schemes by assaulting literal phenomenalistic reconstruction. Eventually he comes to consider phenomena to be posits like physical objects. (To help convince himself of this he will even adopt traditional holistic arguments against phenomena as self-contained givens). Then, alternative schemes 361

of postulated entities confront each other. Phenomena are thus no longer epistemologically fundamental. Convinced of this he can then embrace materialism without a bad epistemological conscience. To make doubly sure, as it were, he will even attack epistemology. We may now t urn to what I have called the second stage: Quine's well-known attack on the plausibility of phenomenalistic reconstruction.

III The second stage is explicitly characterized by a marked change of tone rather than by one of doctrine. 1 0 In the first stage we noted that while Quine did not hold that phenomenalistic reduction could be achieved, he was sympathetic to attempts to carry it out . Now the sympathy has completely disappeared, and belief in such a reconstruction is castigated as 'naive' and 'dogmatic'. With phenomenalism thus dismissed , there is no longer any thought of possible reduction of the mixed ontology of the second stage to a 'simpler' ontology of phenomena alone. Physical objects are now firmly incorporated in Quine's ontology: Physical obj ects are conceptually imported into the situation as convenient intermediaries-not by defi nition in terms of experience , but simply as irreducible posits comparable , epistemologically, to the gods of Homer . . . . But i n point of epistemological footing the physical obj ects and the gods differ only in degree and not in kin d . Both sorts of entities enter our conception only as cultural posits . The myth of physical obj ects is epistemologically superior to most in that it has proved more efficacious than other myths . . . 1 1

This passage also makes it clear that physical objects are still second-class ontological citizens. As 'myths' they are on a par with Homer' s gods precisely because, as we noted earlier, they are not objects of direct experience. Thus at this stage Quine apparently still retains his belief in the epistemological primacy of phenomena (experience) . Quine' s classifyin g physical objects with Homeric deities goes beyond his fondness for the catching phrase and startling formulation. We have seen how his phenomenalism, his belief that in some sense to be is to be an object of direct experience, makes physical object s postulational entities. On the other hand his requiring an ideal language to contain literal reconstructions of all statements of science leads him to reject a phenomenalistic conceptual scheme. Granting this requirement, Quine' s rejection of a phenomenalistic ideal language is not debatable. However, the requirement itself is. For one may hold that this requirement reflects the fact that Quine' s primary concern is with the act ual reconstruction of science and not with the solution of the

362

philosophical problems. We recall that , for Quine, if one's ideal language does not embody these problems he considers them to be solved . If, on the other hand ,one views an ideal language as a tool for the analysis and solution of philosophical problems by discourse about it, rather than by mere elimination within it , one need not and, reasonably, will not insist on a complete an d literal reconstruction of ordinary language (including that of the scientist). Instead , one may be content with reconstructing, from a phenomenalistic basis, only what is necessary for the analysis of the philosophical puzzles. That is, to put it figuratively, an ideal language is ideal not only in that it is an improved language but that it is, in a sense, a fiction. Thus, for example, not every physical obj ect statement, nor, indeed , all of what we · mean' by any such statement , need be reconstructed in order for us to deal with the philosophical problems surrounding physical identity and the notion of substance. In fact no contemporary phenomenalist who knows what he is about would claim that his ideal language can actually be employed by the scientist . We have, then, two opposed views regarding the nature and purpose of philosophizing by means of an ideal language. As in other things one may, perhaps, j udge the two ' enterprises' by their fruits. Earlier we noted that Quine 's views on ontology result in his mixing philosophical and ordinary uses of "exist' . Consequently,he is led to repeat, in his ideal language, philosophically problematic assertions. The same tendency may be noted in connection with his criteria for an adequate conceptual scheme (ideal language) . He is ultimately led to deny such common­ sense things as phenomenal entities. As for physical obj ects, a supposedly philosophical analysis leads to nothing more definite than the poetic conclusion that physical obj ects are myths, like the gods of Homer. Thus Quine clearly concludes his analysis with propositions j ust like those that gave rise to the linguistic turn in philosophy. This outcome, I submit , argues against Quine's conception of analysis. It shows clearly that Quine has either not accepted or not grasped the crucial distinction between ordinary non-philosophical usage and the extraordinary use of language that philosophers are prone to . In t he second stage Quine still faces the dilemma of the first . As we have seen he is led to include both mythical physical obj ects and directly given phenomenal entities in his ontology. B ut if one is still at heart in the classical tradition and concerned with ontological simplicity, he has to choose eventually between phenomenalism and realism. The dilemma is finally resolved - after a fashion - when he embraces Watsonian behaviorism as a philosophy of mind . This step is a natural one to take in view o f Quine 's criteria for an adequate conceptual scheme. For, to many in t he empiricist-analytic tradition, behaviorism is equivalent to scientific psychology. Hence one may be a behaviorist and, in principle, be able to

363

say all that a scientist may have to say about 'mind'. Since, for Quine, all we are concerned with when we construct a conceptual scheme are the statements of the scientist, we can see how his criteria for an adequate conceptual scheme may lead him to adopt behaviorism as a philosophy of mind. This is a further fruit of his conception of the philosophical enterprise. As a philosophical behaviorist, Quine may then explicitly reject phenomena. Yet, he will still call physical objects myths. For, as we shall see, while Quine the behaviorist may have disposed of phenomenal entities , he has not completely disposed of Quine the phenomenalist.

IV The third stage is proclaimed in Quine's paper 'On Mental Entities' when he writes I suggest that it is a mistake to seek an immediately evident reality , somehow more immediately evident than the realm of external obj ects . 12

With this statement he cancels the epistemological priority of phenomena. When we recall that 'epistemological priority' once attracted him to a phenomenalistic conceptual scheme, it is no surprise to find that he quickly proceeds to reject phenomena. Epistemologists have wanted to posit a realm of sense data , situated somehow j ust me-ward of the physical stimulus , for fear of circularity : to view the physical stimulation rather than the sense datum as the end point of scientific evidence would be to make physical science rest for its evidence on physical science . But if with Neurath we accept this circularity , simply recognizing that the science of science is a science ,then we dispose of the epistemological motive for assuming a realm of sense data . 1 3

Nor , in view of Quine's concern with 'science' throughout all this, need we be surprised when he 'replaces' phenomena in the scientifically adequate manner of behaviorist psychology. To repudiate mental entities is not to deny that we sense or even that we are conscious ; it is merely to report and try to describe these facts without assuming entities of a mental kind . . . . we construe consciousness as a faculty of responding to one's own responses . The responses here are , or can be construed as , physical behaviour . 14

The contrasting of epistemology, on the one hand, with a 'science of science', on the other, is rather strange. The science of science is 364

apparently behaviorist psychology (perhaps ·science of scientists' would be more apt). Yet behavior scientists. as scientists. are concerned with discovering lawful connections between items of overt behavior. They are not , as physicists are not , concerned with philosophical problems either epistemological or ontological. As such neither behavior science nor physics provide answe rs to philosophical questions. Thus, prope rly unde rstood , behavior science is not a philosophy of mind - just as phenomenalism is not an experimental science. Yet Quine speaks as if he advocated the replacement of epistemology by a science. Perception psychology construed along behavioristic lines becomes an alternative to phenomenalism. Thus Quine refuses to recognize that phenomenalists at least saw certain philosophical questions to which the psychology of pe rception has no answer. As we earlier saw that he does not distinguish between philosophical and ordinary usage , we now see that he mixes scientific with philosophical questions. By accepting the circularity he mentions, Quine therefore rejects epistemology, as it is ordinarily understood. He thus accepts uncritical ly the universe of the physicalistic behaviorist. Any question as to how we know all the things the scientist tel ls us is omitted. This rejection of philosophy , though strange , is not surprising. For Quine once , we recall , held phenomena to be epistemologically fundamental. Now he rejects phenomena, in favor of postulated physical objects. How could he do that in good conscience if he retained the epistemological questions? So , in rejecting phenomena, Quine simultaneously rejects epistemology. It suggests the ploy of one who . not being able to refute Berkeley , seeks to forget him. This rejection of philosophy in favor of science is what I find strange. 1 5 Yet my attitude reflects a certain conception of philosophy that Quine does not share. For. as we saw, in view of his criteria for an adequate conceptual scheme , behaviorist psychology may provide a philosophical answer to a philosophical question ( or set of such). This is the fruit of his conception of philosophy , which led him to reject phenomenalistic reconstruction in the second stage , that Quine enjoys in the third stage. All this is not, of course ,to be construed as an attack on either science or , specifically, behaviorist psychology. It is merely to suggest that be havior science is not to be mistaken for a philosophical position. One who makes such a 'mistake' either rejects philosophy or embraces materialism or , more often , does both (though , of course , not usually explicitly). I shall not here rehearse the classical objections that Quine thus inherits. What we should note , however. is the complications that ensue by his insisting on the mythical status of material objects. For with phenomena banished , we are faced with a world whose only entities are mythical. This brings us to another strange consequence of Quine 's rejection of phenomena.

365

In the third stage sensa become posits, posits less warranted than physical objects. This is a radical change from the first stage, though, again, not surprising in view of Quine's rejection of epistemology. However, it is extremely paradoxical. An ontological posit may be warranted if it is necessary for the simplification and ordering of our immediate data - sensa and other phenomenal entities. A posit surely is not warranted if it is not so needed. This, we know, is the strategy of ontological postulation. Now, if one holds that phenomena are given, as the Quine of the first stage did, then it is obviously nonsense to say that such givens are unnecessary posits (or even necessary posits). On the other hand, if one rejects phenomena as given, then what is it that we produce order among by making ontological posits? Since there are no longer phenomena, either as given or as posits, all that is left are posited physical things. This is the real circularity that Quine accepts. We posit physical objects to order positedphysical objects. This puzzling circularity is understandable, genetically, when we recall that the 'method' of ontological postulation was formulated in Quine's phenomenalistic first stage. For then he had directly given phenomena whose ordering seemed to require posited physical objects. Without such givens the method of ontological postulation is both pointless and paradoxical. Consequently, sois the ontology that he is ultimately led to adopt by use of this'method' . We have seen how Quine has been led to such a strange position. But we may reasonably wonder why he accepts it. A possible explanation lies in his use, at crucial points, of the term'experience'. The linguistic material is an interlocked system which is tied here and there to experience . . . The statement that there is the planet may be keyed with our sense experience by our seeing the planet , or by our merely noting perturbations in the orbits of other planets . And even the statement that there is a table right here may be keyed with our sense experience through touch or sight or he arsay . . . . how the overall system will continue to work in connection with experience . 1 6

What does 'experience' mean in such passages? If, in spite of Quine's possible protestations to the contrary, he is still thinking of experience in terms of phenomena, then the position of the third stage is blatantly inconsistent. Yet such lapses would make plausible both his retention of the terminology of ontological postulation and his failure to realize the paradoxical consummation of his views. What Quine should, and at places does, hold is that experience is to be construed in terms of items of overt behavior and physical stimulation. But to do this is to explicitly assert the paradox we noted above. For to speak of experience in terms of overt behavior and physical stimulation is to speak physicalistically. Hence we are back to postulating physical objects to order postulated 366

physical objects. Quine's use of the term· experience' thus enables him to castigate phenomenalists and to propose, unmindful of its absurd consequences, his mythological materialism. Actually, Quine's overlooking the consequences of his view may have been aided by an expository device . . . . I do not want to force the issue of recognizing experience as an entity or composite of entities . I have talked up to now as if there were such entities ; I had to talk some l angu age and I uncritically talked this one . But the history of the mind­ body problem bears witness to the awkwardness of the practice . 1 7

Perhaps his · uncritical' linguistic habits are in some measure responsible for what, I suggest, is an uncritical philosophy. For, in the first stage he accepted phenomena, in the third he speaks, at places, as if he did . It may be that he takes himself too literally when, in the third stage, he ties language to " experience'. Quine 's well-known contention that a language need only be tied to · experience' at certain points does nothing to alleviate the problem. In fact, if anything, such talk dramatically illust rates his dilemma. Consider a language , L , which, as we say, contains physical object terms, but no terms that refer to phenomenal objects or properties of such. Some, but not all, of such terms (or statements containing t hem) are coordinated to terms (statements) in a language, L ' , whose terms refer to 'experiences'. Thus L is "interpreted' by means of L ' , much like certain abstract calculi of the theoretical scientist are partially interpreted into empirical concepts (statements) . For one who accepted phenomenal entities referred to by the terms of L ', or perhaps characterized by the statements of L ' , this would seem, on first sight, to be a plausible suggestion. Instead of requiring that all physical object terms (statements) be defined (reconstructed) , in this case in L ' , by means of phenomenal terms (statements) , one now would only require a partial interpretation of L by means of L ' . But for Quine 'experience' is physicalistically construed . Hence the terms of L ' are also physical-object terms. This is, once again, the circularity he accepts. But it is a vicious circle. For the whole problem was to 'tie' physical object terms to something other than physical object terms. To drive home the point, one need only ask what would be gained toward reducing the abstract character of an uninterpreted symbolic system if we provided a 'coordinating dictionary' in terms of another uninterpreted system. {This analogy may suggest a connection between Quine's calling physical objects myths and his thinking of a language as partially tied to experience. Recall that certain scientists and philosophers of science sometimes speak of some of the terms of a partially interpreted calculus as 'referring' to 'fictional entities', and Quine also speaks, at places, of ' positing' microscopic entities. )

367

Aside from the above mentioned difficulties of Quine 's position, there is a further point to note concerning his behaviorism. First, recall that he rejected a phenomenalistic conceptual scheme because a complete and literal reconstruction of physical object statements can not be achieved in it. Reductionism, he contended, is a naive dogma. Yet, is it not strangely inconsistent for him to reject the phenomenalistic reconstruction of statements about 'experience' ? For the behavioristic reconstruction, like the phenomenalistic one, is a reconstruction only in part and in principle. It is just as 'naive and dogmatic' to hold that the one is literally feasible as it is to contend that the other is. Quine should then, for consistency's sake, reject behaviorism as well as phenomenalism. Yet the behaviorist, with his partial reconstruction, says everything about minds that can nowadays be said 'scientifically'. Perhaps this reflects Quine's concern with science rather than with epistemology. In any case, he, inconsistently, rejects phenomena twice - once because of an incomplete reconstruction in terms of them; once in favor of an incomplete reconstruction of them. In a later paper Quine seems to suggest an answer to such an objection when he writes: Contrary to popular belief, such a physical ontology has a place also for states of mind . An inspiration or a hal lucination can . . . be identified with its host for the duration . The feasibility of this artificial identification of any mental seizure , x , with the corresponding time-slice x ' o f its physical host , may be seen by reflecting on the following simple manoeuvre . Where P is any predicate which we might want to apply to x , let us explain P' as true of x ' if and only if P is true of x . Whatever may have been looked upon as evidence , cause , or consequence of P , as a pplied to x, counts now for P' as applied to x ' . This parallelis m , taken together with the extensionality of scientific language , enables us to drop the old P and x from our theory and get on with j ust P' and x' , rechristened as P and x. Such , in effect ,is the identification . It leaves our mentalistic idioms fairly intact , but reconciles them with a physical ontology . 1 8

If we take the closing sentences of this passage literally, it seems that the mind-body problem is solved by "rechristening' physical things with mentalistic terms. For, this procedure leads to a physicalistic ontology which preserves the 'mentalistic idiom'. The basis for this rechristening is that the physical state is present if and only if the mental one is. Moreover , if we know of no such physical state we me rely postulate one. Further , since w e need not know what this physical state is, we 'explain' it in terms of the mental one. Then we eliminate the mental one. This procedure makes no literal sense to me , since we require the mental state , and the assertion about it , to explain what it is we are saying when we assert the postulated physical state. What can we th en possibly mean by holding that afte r we have used the mentalistic asse rtion to explain the physical 368

one we · rechristen' the physical state by the mental istic idiom? For without such a mentalistic state and assertion we have no idea what we are saying when we assert the physicalistic proposition. It seems that Quine is suggesting we do not need mental entities in an ontology after we have made use of them to explain what we say when we speak physicalistically . I n short, without 'Px' we do not know what it is we are asserting when we use ' P' x' ' . But to say that we are asserting 'Px' surely will not do, since this means that we either have not purged our phenomenalistic term in terms from the language or have interpreted a physical-object term in terms of phenomenalistic ones, perhaps making use of a phenomenalistic conceptual scheme. It is one thing to say that ' P'x''and 'Px' say the 'same' thing when we offer the former as a reconstruction of the latter in• a physicalistic scheme. I t is quite another thing, and a patently illegitimate one for a physicalist , to then seek to provide meaning for 'P' x'' by asserting that it says the · same' thing as " Px'. Yet this is what Quine seems to be d oing by his extension of the · method of postulation' . For, according to him, we simply posit that there is a property P' such that 'Px' ' asserts what 'Px' asserts. This use of posits replaces the behaviorisf s attempt to at least formulate such a property. A natural candidate for such a property, as Quine recognizes, would be a characterization of a unique state of the nervous system associated with a certain mental state. But this returns us to what is programmatic in the behaviorist's thesis, as well as to what is problematic philosophically 1 9 . So, in effect, all Quine has said is that if we have no properties on hand to make our relevant bicondit ional statements true, we simply posit that there are such. The phenomenalist too might profit from this new use of 'posits' in philosophical analysis. Consider the sentence (S 1 ) 'This is a chair'. We may take a sentence (S) ) , asserting that there are chair percepts (visual and/or tactile) which fulfill a certain posited complex relational property R, such that (SJ is true if and only if (S 1 ) is. The evidence for (S2 ) is exactly what we would consider evidence for (S 1 ) . Thus we no longer need physical object statements like (S1 ) . Perhaps in this manner the phenomenalist may get rid of all physical object statements. Yet the phenomenalist , in his own terms, cannot specify the meaning of 'R', just as Quine cannot do so for 'P' '. For, if 'R' is undefined , what it refers to must be presented in experience; if 'R' is a defined term, its definition must be specified. Quine might overlook, or not recognize, this difficulty, because of his holistic conception of meaning - his view that language as a whole, rather than terms or sentences, is the ' unit of meaning'. One aspect of this question we have already considered in discussing Quine's talk of 'experience', to which his conceptual scheme is tied . Another aspect is the issue of holism it self. This brings us to another thread of his argument against sensa. 369

V In addition to his Watsonian meditations, Quine provides two further arguments for rejecting sensa. The first attack concerns memory. Our present data of our own past experiences are , on this theory , some sort of faint present replicas of past sense impressions ; faint echoes of past sensation accompanying the blare of present sensation . Now it takes little soul-searching to persuade oneself that such double impressions, dim against bright , are rather the exception than the rule . Ordinarily we do not remember the trapezoidal sensory surface of a desk , as a color patch extending across the lower half of the visual field ; what we remember is that there was a desk meeting such-and-such approximate specifications of form and size in three-dimensional space . Memory is j ust as m uch a product of the past positing of extra-sensory obj ects as it is a datum for the positing of past sense data . 20

The arguments may be taken in two ways. First, Quine could be contending that a phenomenalistic conceptual scheme is inadequate for the reconstruction of memory statements in that physical object terms are required in some of them. The argument requires him to hol d, as he does, that statements involving such terms are not reconstructible. To answer this argument one would , of course, have to present a pattern for reconstructing memory statements within the framework of a phenomenalistic conceptual scheme. One would also have to argue, along the lines I mentioned earlier, that schematic reconstruction of memory statements, as of physical object statements , was sufficient. Here, I merely note that a contemporary phenomenalist need not limit himself to the somewhat Humean apparatus of vivid and faint data in dealing with the problem of memory. 21 Second, Quine could be suggesting that a phenomena listic reconstruction of memory statements won't do since we do not speak in sense-data terms in ordinary language. That is, Quine could be arguing along lines similar to those used by ordinary language analysts. If this is so , it is interesting to recall his earlier rejection of epistemology in connection with his arguing in the style of a movement that systematically rejects all philosophical questions. The second attack seeks to dismiss sensa as "tenuous abstractions". It would be increasingly apparent from the findings of the Gestalt psychologists , if it were not quite apparent from everyday experience , that our selective awareness of present sensory surfaces is a function of present purposes and past conceptualizations. The contribution of reason cannot be viewed as limited merely to conceptualizing a presented pageant of experie nce and positing obj ects behind it ; for this activity reacts , by selection and emphasis , on the qualitative make-up of the pageant itse lf in its succeeding portions . It is not an instructive

370

oversimplification , but a basic falsification , to represent cogmt1on as a discernment of regularities in an unadulterated stream of experience . 22

I n the above quoted passage Quine seems to assert three things. (1) Sensa have causes, in particular our 'reason ' and the psychological sets with which we confront the stream of experience, and these causes partly determine the qualitative character of the sensa. (2) Sensa are 'abstracted out' from a complex stream of experience . As such they are abstract entities. (3) Due to (1) and (2) there are not 'pure' sense data. Hence the reality of sensa is suspect. Concerning ( 1), aside from the Kantian flavor , we may note that the transition from the quite commonsensical observation that data have causes ·outside' of ·raw experience' to the banishing of such data as ontological entities is patently illegitimate. The results of perception psychology which show that we bring a 'set' to a perceptual situation and that this set may causally determine what we perceive does not in any way provide evidence against the ontology or epistemology of the phenomenalist. The point is not whether or not our data have causes , speaking commonsensically (scientifically, if you will), but whether or not we 'have' such data. If there is now a red sensum with which I am acquainted, then it is a fact that there is something red. Whether or not the scientist can discover certain correlations involving my reports of it, my perceptual set when I 'have' it, my brain states, etc. , has nothing at all to do with whether or not there is such a datum. There is also a suggestion of the holistic pattern in ( 1) , but this comes out much more explicitly in (2). For (2) is reminiscent indeed of the familiar Hegelian-pragmatic antipathy to 'vicious abstractions' from the holistic character of experience. According to this pattern, when we 'abstract' sensa from the flux of experience we falsify experience which simply doesn't present itself in an atomistic fashion. One may wonder why Quine adopts, at this stage, the Hegelian pattern . The explanation is , I believe, twofold. First, we saw that part of his strategy in rejecting phenomena in the third stage is to consider them as posits rather than as givens. Once they are considered to be posits, he can reject them as unnecessary. This procedure would hardly make sense, as we noted, for objects of direct acquaintance. This being so , Quine seeks arguments against sensa being obj ects of direct acquaintance. The holistic tradition provides such ready made arguments against the 'self-contained given'. Second, Quine had already incorporated the holistic pattern into his philosophy in the second stage. In the best known piece of that stage he had linked the dogma of reductionism with a second 'dogma of empiricism' -the synthetic-analytic dichotomy. Moreover, and more important, he implicitly saw a connection between this dichotomy and the thesis of logical atomism.

37 1

The dogma of reductionism survives in the supposition that each statement , taken in isolation from its fel lows, can admit of confirmation or infirmation at all . M y countersuggestion, issuing essentially from Carnap's doctrine o f the physical world in the Aujbau , is that our statements about the external world face the tribunal of sense experience not individually but only as a corporate body . The dogma of reductionism even in its attenuated form , is intimately connected with the other dogma - that there is a cleavage between the analytic and the syn thetic . We have found ourselves led, indeed , from the latter problem to the former through the verification theory of meaning. More directly, the one dogma clearly supports the other in this way : as long as it is taken to be significant in general to speak of the confirmation and infirmation of a statement , it seems significant to speak of a limiting kind of statement which is vacuously confirmed, ipso facto , come what m ay ; and such a statement is analytic . . . . But what I am now urging is that even in taking the statement as unit we have drawn our grid too finely. The unit of empirical significance is the whole of science . 23

In short, the meaning of any term or statement as well as the truth of any statement must be considered in the contextual setting of the total conceptual scheme. This is, of course, the holistic pattern and the rejection of logical atomism. Thus Quine's linking, in the second stage, of the two 'dogmas' of phenomenalistic reduction and the synthetic-analytic dichotomy lead him to adopt the holistic pattern . Following this pattern and rejecting logical atomism, he then, in the third stage, may consider phenomena to be 'abstractions' and not self-contained, directly given objects of experience. Then , as we saw, phenomena, being abstractions and not directly givens, become, for him, posits. But , under the impetus of his materialism, they are unnecessary posits. Hence they are banished from his ontology. We thus see how a combination of 'scientific materialism' and Hegelian holism combine to form a two­ pronged attack that, for Quine, obliterates phenomena. But then , the combination of materialism and holism is not new . What is i n this context new is Quine's concern with the mythological character of physical objects. This, we saw , is a result of his early commitment to a phenomenalism which lingers on to complicate his materialism. Perhaps by still labelling physical objects · myths' Quine manages to placate an earlier concern with what is ·epistemologically fundamental'. Notes 1

2

3

One such place is in 'The Scope and Language of Science', The British Journal for the Philosophy ofScience, 8, 1957, pp. 1- 17. W. V. Quine, ' Mr. Strawson on Logical Theory', Mind, 62 , 1 953 , p. 46. For a detailed analysis of Quine's explication of ontology see three of my papers , 'The

372

4

6 7 8 9 10 11

12 13 14

15 16 17 18 19

20 21 22 23

Ontological Operator', Philosophy of Science, 23, I 956, 25�259 ; 'Professor Quine, Pegasus and Dr. Cartwright' and ·on Pegasizing', in this volume, pp . 86 f. W. V. Quine , From a Logical Point of View, Cambridge : 1953), pp. 45 , 16-17. Hereafter, I will refer to this book as FLPV and footnote quoted passages according to where they occur in the book , rather than in the original papers. For detailed discussions of the role of ideal languages in philosophical analysis see G. Bergmann, The Methaphysics of Logical Positivism (Longmans, Green, and Co . : 1954) ; Meaning and Existence (Madison: 1959). FLPV, pp . 17-18. FLPV, p. 18. FLPV, p . 19. FLPV, p. 18. However, see the discussion below concerning the holism implicit in the second stage. FLPV, p . 44. W. V. Quine , 'On Mental Entities', Proceedings of the American Academy of A rts and Sciences, 80, 3, 202. I will refer to this paper as OM E. OME, 202 . OME, 203 . Recall that in a passage quoted above Quine seemed to contrast physics with phenomenalism . See note 8. OME, 198, 199, 200 . OME. 202. The Scope and Language of Science' , p. 1 5 . There is , of course , a n issue connected with the use some philosophers make of discovered or speculative (not postulated, in Quine's sense) correlations between mental and physical states , in their attempts to disavow mental entities . But this takes us beyond Quine's particular version of physicalism. For a critical examination of a less sophisticated attempt to use such correlations to establish 'identities' between the mental and the physical see my ' Physicalism , Behaviorism and Phenomena' , in this volume, pp . 374 f. OME, 20 1 . For example , see G. Bergmann , 'Some Reflections on Time' , in II Tempo (Archivio di Filosofia , 1958) , reprinted in Meaning and Existence. OME, 201 . FLPV, pp . 4 1 , 42.

Physicalism , Behaviorism , and Phenomena

The issue of materialism has recently been raised again. Mr. Putnam argues against philosophical behaviorism [Putnam 1957, p. 97]. Such a position holds, as he construes it, that statements like 'Jones is angry' can be analyzed in solely behavioral terms. When one argues against philosophical behaviorism, he might be expected to distinguish this metaphysical position from behavior science. Putnam, however, does not make the distinction.Consequently he argues against both. I shall first state the distinction between these two different things, namely, philosophical behaviorism and behavior science, as I see it.The behavior scientist adopts the thesis that in principle it is possible to predict future behavior on the basis of data concerning environmental, behavioral, and physiological variables. All three of these he considers in physical terms. The behavior scientist thus speaks about physical objects and properties of such. Talking in such terms, he believes that it is in principle possible to coordinate to statements asserting that person X has or is in state of mind Y another statement, employing only the above mentioned physical terms, such that either both are true or both are false. The reasons for the behavior scientist's program are the well known quandaries involved in the observation of other people's minds and the need for intersubjective verification in science. One can further distinguish between a narrower and a broader view of behavior science. The former restricts itself to environmental and behavioral variables at what some call the macro level; the latter includes, or even concentrates upon, physiological variables.As scientists neither the behaviorist nor the physiologist asks or answers philosophical questions, either epistemological or ontological, about minds, bodies, and mental contents. A philosophical behaviorist, on the other hand , is one who argues that there are neither minds nor mental contents - sometimes basing his argument on behavior science. This position has been taken by Watson, flirted with by Ryle , and recently adopted by Quine. Philosophical behaviorism is a form of materialism . Some philosophical behaviorists choose to identify what we would think of as mental contents with overt behavioral patterns. Others prefer to locate them within the brain. In either case they are physicalists (materialists).Thus any dispute between them is an argument within the materialist camp.To put it another way, commonsense distinguishes three kinds of 'things' : (1) mental 374

(phenomenal) objects and contents, (2) overt behavior, (3) physiological states. Some philosophical behaviorists (hereafter, briefly, behaviorists) 'remove' (1) by identifying it with (2) , others (hereafter, briefly, physicalists) achieve the " red uction' by identifying ( 1) with (3). Putnam argues against "behaviorism'. But he does not do so because he holds that there are mental contents. Rather, he chastises those who would iden tify (1) with (2) by contending that (1 ) should be identified with (3). Thus, for Putnam phenomenal entities are " really' brain states and not macro-be­ havioral states. This argu ment of his against behaviorism follows a pattern that is often used in defending the existence of mental contents a gainst the behaviorist. In this pattern one argues that it is logically possible for a subject to present all the behavioral characteristics of anger (including dispositions) and yet not be angry, in the sense of not being in the appropriate mental state. Putnam rej ects behaviorism as follows. Identifying anger with a physiological state he points out the logical possibility of a ·non-angry' physiological state accompanying the overt behavior which we presently take as an indication of anger. In so arguing Putnam accepts a position which is just as repugnant to commonsense as the one he argues against. I say repugnant because, first, it overlooks the commonsense distinction between ( 1) , (2) , and (3) , and , second, I do not see how one can deny the existence of what we are directly acquainted with. Moreover, no behaviorist or physicalist can present an adequate analysis of the philosophical puzzles about mind on the simple ground that his reconstruction of statements about mental contents will n ot reflect the fundamental differences between our talk about our own contents, our talk about the mental contents of others, and our talk about brain states (or, for that matter, behavioral states), be they our own or others'. This raises a question of intellectual motive. Why should anyone wish to maintain that mental states and brain states (rather than behavioral states) are one and the same thing. One possible motive is Occam's razor. Phenomenal entities would be dispensed with in favor of physical objects (tissues, etc. ), for one would then hold that statements about phenomenal entities may be construed as statements about physical objects. I just indicated why such constructions are not satisfactory. In any case, Putnam gives no hint, either implicit or explicit, that ontological economy is what he seeks. What seem to be the major motives behind Putnam' s program are, first, his concern with'meaning' , and, second, his implicit belief that good science is good philosophy. Putnam distinguishes between "meaning' and 'being identical with' by citing the familiar case of the morning star. This 'Physicalism' , expressed as a working hypothesis, amounts to this : a subj ective experience ( e . g. , a particular feeling of anger) is a particular kind of

375

physical state of the organism . This is of course a synthetic identity , if true ( as Feigl has very well pointed out) . Philosophers are quite right in saying that 'the sensation blue' cannot mean a physical state. But they are wrong when they m aintain that it cannot be a physical state. (Thus , 'the morning star' cannot mean 'the evening star' . But the morning star is the evening star . . . ) [p . 97]

He thus attempts to make his identification of, say, blue sensations with brain states plausible by introjecting the meaning-reference issue.But to bring in this issue at this point merely beclouds the matter in two ways. First, in cases where there is clearly one and only one thing, say, the planet Venus, we have learned that certain puzzles about'meaning' arise if we refer to that one object by two phrases. To dissolve these puzzles one must analyze certain questions about meaning. These questions about meaning are not what is involved in'identifying' brain states with mental states or objects.In other words Putnam is not dealing with a situation in which it is already clear that we are referring to one and the same thing in two different ways. For the philosophical issue here is precisely whether there are two kinds of things or only one.Hence, philosophers who insist that phenomena are not brain states cannot be put off by introducing the meaning-reference distinction into the dispute. Second, the scientific question as to whether or not the morning star is the evening star is of a kind radically different from that of the philosophical one as to whether or not phenomena are identical with either brain states or behavior. In this lumping together of these two questions we encounter the first indication of what I believe to be Putnam's tendency to assimilate philosophical to scientific questions.In the case at hand this tendency is aided by a verbal bridge that is provided by (1) the behavior scientist's problem about the meaning of mental terms and (2) by Putnam's threefold use of the term 'meaning'.To the structure of this bridge I now turn. Recall our distinction among three kinds of things. Corresponding to it we can distinguish phenomenal , behavioral and physiological terms.The behavior scientist seeks to introduce expressions like 'anger' in terms of his physical obj ect terminology. Ordinarily, I would have used 'define' instead of 'introduce' , but Putnam speaks of the 'specification of meaning' by 'symptoms' rather than by definition. Of course , we cannot give meaning to a term , say 'glub' , by saying, "By 'glub' we mean an at present not definable physical state" . But we can incomplete ly specify the meaning of such a term ( and thus make it usable in science) by providing symptoms. [p. 98)

In effect Putnam thus raises the further issue of terms given incomplete meaning through specification of symptoms (perhaps introduced by 376

reduction sentences?) versus terms whose meaning is completely specified (introduced by explicit definition) . Presently we shall come to understand the role that this now much debated issue plays in his rejection of phenomenal entities. Let it first be noted that one can and must distinguish the tenuous and program matic way in which scientists sometimes actually use terms whose meaning is incompletely specified, in what Reichenbach aptly calls the context of discovery, from the analysis or schematic reconstruction which the philosopher attempts in the context of justification. Failure to distinguish between the two contexts may lead to a failure to distinguish between scientific and philosophical questions. In view of this distinction nothing that follows · should be construed as an attempt to limit the freedom of the scientist in the con text of discovery . Suppose one introduces a term, say 'anger' , by specifying its meaning (not defining it) in term of a set of macro-behavioral characteristics (symptoms) . Let S be this set . What we then have is a conditional (or perhaps biconditional) statement mentioning anger as well as the members of this set . To introduce terms in this manner is held to secure their openness . This I take to mean, among other things , that one may, as our knowledge develops, alter the set by the addition of new or the omission of old members. This openness is thus a linguistic one concerning the introduction of a sign about whose meaning there is some question . Or, to put it another way, this openness is one of meaning . That explains why some speak of signs so introduced as having incomplete meaning or lacking complete specification of meaning. Consider next another situation. Behavior scientists and physiologists seek to correlate overt behavioral states to physiological ones. That is, they search for laws that will connect a physiological state 0 (or a class of such states) with a behavioral complex, say S. ('0' stands for a set of statements about cells, tissues , nerve fibres, etc. ) We do know some such laws . We also believe that we shall eventually know many more . Be that as it may, when we now speak of such laws yet to be discovered, our speech is necessarily vague or 'open'. For we can merely say that there is some brain state, which we can as yet not completely specify or perhaps not at all specify, that corresponds lawfully to a behavioral state. This 'vagueness' or 'openness' is clearly not the vagueness or openness we encountered before. Before we were concerned with the meaning of a term. Now we are not dealing with a question of meaning. Rather , we are searching for a thing , so to speak, and a law ; we are not seeking to give meaning to a term. Putnam in identifying brain states with mental states mixes these two kinds of vagueness or openness. Thus he writes: This , then , is the full contention : at least at the physical science level , ' anger' means

377

a particular micro-state of the cereb�al cortex which we are not able at present to define , but for which we are able to provide symptoms . [p. 98]

This leaves no doubt that he conceives of the future specification of the 0 (in the case of anger) as providing us with the answer to the question as to what 'anger' means.Apparently, he seeks for some kind of metaphysical closure.If my analysis is correct, he achieves it by collapsing the two kinds of openness I just distinguished, namely, the openness of the term 'anger', as he proposes to introduce it, and the (different) openness due to the lack of specification of the brain state 0 which is lawfully connected with anger.Or, if I may so put it, we have an 'open term' and an 'open thing'. Putnam achieves a kind of closure by having the former refer, or as he puts it, mean the latter. I have suggested that one reason why Putnam identifies brain states and phenomena is that he does not distinguish between two kinds of openness. Since, as we saw, this amounts to a mixing of 'meaningfulness' with 'lawfulness', one may, in turn, wonder why Putnam is led to do that. We shall now see that his fusing of 'meaningfulness' and 'lawfulness' is a consequence of his views on meaning. We may distinguish, among others, three uses of 'meaning'. First, there is the use we make of 'meaning' when we speak of giving meaning to a new term by introducing it into the language we actually speak.Let us call this 'meaning 1 ' . Second, there is a use of 'meaning' which is synonymous with 'significance'.In this sense we learn more and more of the 'meaning' of a term as we discover more and more laws in which it occurs. I will call this 'meaning2 ' . Third, there is a use of 'meaning' in which it is synonymous with 'referent'. I will refer to this use as 'meaning3 ' . If we 'introduce' a term, say 'anger' , in terms of certain behavioral characteristics, say S, we can say ·anger' has been given meaning (meaning 1 ) . But if we are thinking in terms of symptomatic specification, rather than of explicit definition , certain dangers arise.Specifically, one who thinks that in the context of justification terms may be introduced by symptomatic specification is in danger of mixing meaning 1 with meaning2 . This is readily seen when one considers that by adding (or omitting) symptoms we are guided by the discovery (or rejection) of lawful connections among the symptoms. Putnam , himself, in this context , speaks of discovering high inner correlations among symptoms. He also tells us As Abraham Kaplan has pointed out , we do not , in general , go directly from the list of symptoms to a definition ; what rather happens is a long process of adding to and revising our list of symptoms as we search for a more and more narrowly specified

378

term , before we reach a term which is , in the strict sense of the word , definable. [ p . 98)

Here he explicitly mixes the context of discovery with the context of justification and meaning 1 with meaningi - Such a fusion becomes even more apparent when, as an argument against the view that all meaningful terms are definable, Putnam holds that, in contrast to the "arbitrary' process of redefinition, . . . when a term is specified by means of a list of symptoms , the adding of new indicators as they are discovered, and the progressive sharpening of the term , are in no sense arbitrary . in the actual process of science . [ 4 , p . 99)

What is not arbitrary in the process of symptomatic specification is, of course , the use of discovered lawful connections to provide meaning for a term - as contrasted with meaning stipulated by definition. 1 The mixing of meaning 1 with meaning2 is the structural heart of the holistic-idealistic contention that in order to know what a term 'really' means we must know all the laws it figures in. Thus we find the avowed physicalist Putnam adopting a familiar idealistic pattern. It is more ironical still that his adoption of this pattern is one of the intellectual motives that causes him to reject phenomenal entities. Putnam's holism comes out explicitly in yet another way when he speaks of 'implicit definition'. Our assumption or hope is that more precise indicators will be found for 'anger' ( or for various kinds of anger) until eventually an actual definition is possible ; but the term ·anger' (like other partially specified terms, e. g. , 'species' , in zoology) is meaningful now, it does not merely become meaningful when it becomes possible to define it, strictly or 'implicitly' - i. e. , by incorporation into a theoretical system . [p . 99]

This passage may be interpreted in more than one way. If we take the phrase "strictly or 'implicitly "' to state that a term can be defined either strictly or implicitly then he could mean, first, to identify 'strict definition' with 'explicit definition'. In this case 'anger' would be defined (explicitly) in terms of a set of symptoms, say S, with high inner correlation. 2 Or, by 'strict definition' he could mean, second, the actual specification of a micro-state, say 0, correlated with S.This would flagrantly mix meaning 1 with meaning2 . The plausibility of this interpretation is supported by Putnam's assertion that '"anger' means a particular micro-state of the cerebral cortex which we are not able at present to define [italics added] . . . ". However, instead of contrasting strict and implicit definition, Putnam could mean to identify them. In any case, at least some terms, 379

possibly all, are 'defined' when they are 'implicitly defined' by incorporation into a (presumably comprehensive) theoretical system. This shows, in yet another way, Putnam's mixing of meaning 1 and meaning2 and his consequent holism. For, recall, those who speak of 'implicit definition' think of the statements (laws, axioms) of theoretical systems as providing meaning for the terms they contain. 3 To repeat. One who fails to distinguish between meaning 1 and meaning2 and thinks of symptoms as providing 'meaning' (in some undifferentiated sense) for a term will easily be led to believe that we discover more of the'meaning' of a term each time we discover a law in which it occurs. Thus such a one may be further led to assimilate questions of lawfulness and questions of meaningfulness.Hence, he may, in turn, be led to think that the discovery of the law specifying the 0 related to the set S provides'meaning' for'anger'. Putnam's position is further complicated by the third sense of 'meaning' (meaning3 ) we considered. However , before we see how meaning3 complicates the pattern we may profitably note some intellectual motives for the fusing of meaning 1 and meaning2 . Terms introduced by symptomatic specifiction are not eliminable abbreviations. In this respect they differ from terms explicitly defined. [This may lead one to hold that the meaning of terms explicitly defined is 'closed', as opposed to the 'openness' of those introduced by symptomatic specification. ] Furthermore , some hold, as I do, that we produce an adequate'explication' of terms in the context of justification only if such terms are eliminable. 4 Thus, introduction of terms by symptomatic specification of meaning does not enable us to provide for such an adequate analysis. Moreover, as we have seen , appealing to systematic specification may lead one into traditional philosophical puzzles.Thus Putnam was led into denying the existence of phenomenal entities. In addition, as we also saw, the adherence to systematic specification in the context of justification enmeshes philosophers, nominally in the empiricist tradition, in holism. In view of all this one may wonder why Putnam and those he cites are so anxious to reject explicit definitions in favor of symptomatic specification . One reason is furnished by the'paradoxes' due to the use of the logical connective 'if...then' in the definitions of dispositional properties.With this question I shall not deal. However, we may note that essentially the same problem is introduced by the use of the conditional sign in statements of lawfulness. Thus, we may wonder what is gained by acq uiring the additional troubles produced by speaking of symptomatic specification of meaning. Be that as it may, there is another appeal of symptomatic specification. Advocates of reduction sen tences have argued that the openness of meaning associated with such methods of introducing terms is more in keeping

380

with the actual progress of science. Thus they seek to turn what some consider to be a defect of analysis ( the noneliminability of such 'open' terms) into a virtue by suggesting that their approach is more ·scientific'. Putnam, as we saw, adopts this line of thought when he cites Kaplan. But, again, all this clearly mixes the context of justification with that of discovery, since logical analysis of knowledge is not concerned with the growth of knowledge or , for that matter, with what hinders or helps such growth. To put it another way, philosophical analysis and scientific discovery are two things and not one : and it is not the task or place of the former to imitate either the process or the progress of the latter. For what it is worth we may note how the merging of these two different activities fits with the holism that results from the merging of meaning 1 with meaning .., . To see the connection we need only recall that insistence on the ·opennes-s· of the knowing situation and its incorporation into a philosophical pattern is the core of the pragmatist variant of holism. Putnam states that terms like 'anger' do not mean brain states, even though anger is a brain state. Yet he also says, inconsistently, that 'anger' means a brain state. One source of this inconsistency is the fusing of meaning 1 and meaningr Another source may be a further fusion with meaning 1 . We recall that Putnam thinks that the meaning-reference distinction is relevant to our discussion. Thus he may hold that while ·anger' refers to a kind of brain state , the latter is not the meaning of the term. If he then, inconsistently , thinks of 'meaning' in the sense of 'referent' we can see how he is led to utter contradictory assertions. Moreover, we will not be surprised at Putnam's lapsing into such an inconsistency , if we recall three things. Recall, first, that Putnam spoke of theoretical systems providing implicit definitions and hence 'meaning' for terms. Such 'systems' may be thought of as artificial calculi or axiomatic systems. Recall, second, that we sometimes say that we give meaning to such a system by 'interpreting' it, i. e. , by tying some of its terms to ·referents'. Prior to interpretation such systems are often said to lack empirical 'meaning' or content. Finally, we recall that the behavior scientist, for his purposes, is not concerned with phenomenal entities. Hence, he will not consider interpretations that involve such entities. Rather, he will interpret terms like ·anger' into physical things or states. Putnam apparently chooses a particular kind of physical state for 'anger' - a brain state. All this could explain why he then comes to say, inconsistently, that 'anger' means a brain state. But it does not explain why he chooses to identify anger with a micro rather than a macro-state. A possible motive for this choice is provided by a current issue in the philosophy of psychology . For the macro-behavior scientist anger is a dispositional state. Thus the term 'anger' is, for such a scientist, an intervening variable. Some

38 1

analysts feel that such terms are not satisfactory , since they do not refer to manifest states of the organism. For, not referring to manifest states, intervening variables are thought of, by these analysts, as 'reintroducing' mind into behaviorist psychology. Also, since some of these analysts hold that intervening variables are to be introduced by symptomatic specification, rather than by explicit definition, the problematic nature of such terms, and the apparent need for their having 'concrete' referents, is enhanced. Consequently , some have proposed that intervening variables should be thought of as referring to manifest brain states. In this way one supposedly avoids 'abstract' entities in favor of 'concrete' physical things. Feigl has been one that favors such 'concrete' interpretations of intervening variables. Putnam cites Feigl. He even adopts his phrase 'synthetic identity'. Thus it seems reasonable to conclude that this issue in the philosophy of psychology provides a motive for Putnam's identification of anger with a micro-state. In this connection we may note four points. First, if a term is explicitly defined by means of macrophysical object terms, then it is eminently reasonable to consider it also to be such a term , since it is simply an eliminable abbreviation for a set of such terms. However, if a term is introduced by symptomatic specification, it has, as we noted, a certain 'openness'. Because of this one may think that even though a term is introduced by means of macrophysical object terms it need not be one. Thus one may puzzle about what such terms refer to or 'mean' . If one is materialistically inclined then, quite naturally , he might pick manifest micro-states as such referents. In fact, one may even wonder if symptomatic specification appeals, unconsciously , to materialists by providing 'open' terms which then 'require' concrete micro-referents. Historically materialism has frequently appealed to micro-objects. Be that as it may , we have seen how the rejection of explicit definitions, ( 1 ), gives rise to a 'problem' about entities associated with intervening variables and , (2), reinforces (a) Putnam's physicalism, (b) his introduction of the meaning-reference distinction and (c) his fusion of various meanings of ' meaning'. Second , we may note a further connection between the fear of introducing 'abstract' entities via intervening variables and the rejection of phenomena. For, phenomena too have been rejected by some as abstract entities. This similarity is more than verbal: it indicates another connection between metaphysical physicalism and certain attitudes toward intervening variables , symptomatic specification and explicit definition. Third , the physiologist , as a scientist, is a commonsense epi­ phenomenalist. 5 As such he is not interested in, nor does he consider, phenomenal states (objects) as distinct kinds of things -just as the macro382

behavior scientist need not do so. Thus, in one sense, for the scientist there are no such things. To say this is one thing. To assert as a philosophical-ontological position that there are no phenomena is another thing.6 Thus to claim, as Putnam does, scientific support for his physicalism is to hold that philosophical questions are answered by the scientist. This tendency to answer philosophical questions in terms of scientific practice we noticed earlier. Fourth, Putnam's use of an assumed one-one correlation between phenomenal states (objects) and brain states to identify the two is puzzling in two ways.First, one wonders what the correlation is between if there is only one thing. If they are identical there is no correlation; if there is a correlation they are not identical. Second, since Putnam holds that one-one correlations establish ·synthetic' identities, one wonders what he would assert if we discover (or assume) that such correlations hold between macro and micro-states.Would he then hold that macro­ states are identical with micro-states? That is, would one's overt behavior be a state of one's brain? Perhaps a further motive for Putnam's identification of phenomenal states (objects) with brain states is furnished by the extensionality axiom. Thus where ·B' refers to a phenomenal property and'P' to a physiological one, he may believe that from the assumed law '(x) [Bx Px]' and the extensionality axiom we may infer 'B = P' . This would further explain why he mentions the meaning-reference distinction.For he might then hold that while B is identical with P, 'B' and'P' do not have the same meaning. This raises two questions. One concerns the legitimacy of so using the extensionality axiom to identify B and P; the other involves the difference in meaning of 'B' and 'P'. Concerning the first, it must be pointed out that the extensionality axiom does not enable us to identify B and P. For, in order to use this axiom, the individual's exemplifying B and P would have to be the same. Hence, the use of this axiom to identify phenomenal and physiological properties presupposes the identity of particular phenomenal and physiological states (objects). But this clearly begs the question. Moreover, the laws that Putnam assumes do not have the form '(x )[Bx = Px]'. Rather, they would be like '(3.x)Bx 0', where '0', as before, stands for a complex description of a physiological state. Thus '0' contains references to such physical things as tissues, cells, etc., and 'B' is predicated of a phenomenal particular.The extensionality axiom cannot then be used to identify phenomenal characters or particulars with physiological ones. One may, however, suggest a "weakened" use of this axiom. Consider two properties, designated by 'S 1 ' and 'S2 ' , such that one is a property of being in a certain physiological state and the other is a property of being in

=

=

383

(or, rather, having) a certain phenomenal state. Hence, both properties will be properties of organisms. Then, from '(x) [ S1 x Sz♦X] and the extensionality axiom, one may infer 'S1 = S2 ' . But we have not identified phenomenal and physiological states, objects, or properties. All we have done is to identity properties of being in, or having, certain states, since whenever an organism is in one it is in the other. But even this meager "result" must be qualified in several ways. First, to identify these properties in this fashion commits one to the identity of such properties as featherless biped and rational animal (assuming that they are coextensive). Second, both 'Si ' and 'S2 ' will be physical object terms. Consequently, one ends up by merely identifying one property of physical objects with another. Third, we should not expect to achieve any significant results by use of the extensionality axiom when we concern ourselves with the ontological status of mental contents and, hence, of "mind. " Such considerations notoriously involve "intensional" contexts. Hence, any use of the extensionality axiom to eliminate mental contents would undoubtedly involve either question begging or an inadequate analysis. I mentioned that a question also arises, in this context, concerning the difference in meaning between ' B' and ' P' . One way of accounting for this difference would be to hold that 'B' and 'P' refer to different universals (their intensions) , even though they have the same extension. This would involve identifying meaning and referent for predicate terms. However, one could still hold that these terms had different meanings (referred to different universals) while having the same "' referents" in the sense of identical extensions. But if the extensions are identical then the particular phenomenal states (objects) and brain states would be identical. Consequently, one would have to hold that the universals referred to by 'B' and 'P' are also identical. Otherwise what would the phenomenal property B be exemplified by-physical objects? In what sense could it then be called a phenomenal character? Thus we are forced to another alternative. The "two" properties are identical but the two predicates, 'B' and 'P' , referring to '"it" do not " mean" the same thing. Hence, we introduce, in ad dition to the universal referred to by both •B' and 'P', two additional entities - the meanings of 'B' and " P'. It is then interesti ng to note that in attempting to achieve ontological parsimony by rejecting certain commonsense objects (phenomena) . one may end up by extending one's ontology to include objects that are far from commonsensical (meani ngs) . These meanings will be universals, for they will be 'instanced' on more than one occasi on. In view of these entities one may plausibly hold that Putnam is not a complete physicalist. For some such entities might be considered as 'mental'. I say · some such entities' because there is a possible ambiguity in all this. Consider two

=

384

predicates, one a phenomenal character term, the other a physiological one, which refer to characters that are uniquely correlated. We would ordinarily so classify the terms precisely because one refers to a phenomenal character and the other to a physiological one. But for Putnam there are only the latter properties. Hence, for him, both terms refer to the same physiological property. Yet they have different meanings . Thus, the distinction between a phenomenal character term and a physiological one must be based on these different entities - the meanings . It then seems reasonable to classify the meanings which determine some terms to be phenomenal character terms as mental entities . Yet, in view of historical associations with the term ' meaning', one may well classify all such entities as 'mental' . In either case we may note two things. First, in attempting to eliminate mental particulars, an avowed physicalist may end up with mental universals. This is slightly paradoxical, to say the least. Second, in accepting such entities a "scientific' physicalist comes close to some familiar med ieval thought patterns - particularly if, for the moment, we introduce a universal 'mind' to harbor such 'meanings' ( concepts, ideas, intelligible species) . This too is paradoxical, or, at least, ironical. One could attempt to avoid these conseq uences by, first, holding in nominalistic fashion that predicates may be used without ontological commitment to properties. Hence, the predicates 'B' and 'P' are 'true of the same things though they, in turn, do not stand for a ( or more than one) property. This materialis t argument would then presuppose the viability of nominalism. If one did not embrace nominalism, he could still hold that while 'B' and 'P' st ood for the same property, we need not take the claim that they ' mean' different things to have ontological import in that ad dition al entities, meanings, are recognized. The difference in meaning is a matter of the different roles the predicates play in the conceptual scheme to which they belong. Whether such a use of 'meaning' is viable and ontologically neutral is disputable. But, we need not bother with probing into that matter here. For, it simply will not do to claim that 'B' and 'P' stand for the same property or that B is eliminable in favor of ( or reducible to) 0 . Where ' B' is a primitive predicate taken to refer to a presented property of a phenomenal obj ect, it cannot be identified with P or eliminated in favor of 0 . For to do so is, first, to ignore the fact that we deal with empirical correlations (between B and P or 0) and , second, that it is logically possible to have an instance of B without the correspond ing physical properties being instantiated. The materialist is guided by a misleading analogy: the case of the morning star and the evening star. In that case what we discover is that one thing has two properties, not that something is identical with something or that two expressions refer to the same thing. To take the case of the property B to be an alogous , we would 385

have to discover that that property, in turn, had two properties. But that is clearly absurd in this case. The most one can discover is a correlation of properties. To speak of postulating the elimination or identification of B is equally absurd. What can it mean to postulate that a presented property does not exist? As for postulating an identification of B with some physiological property, it is clear that the latter would have to be a complex property, for a complex physical characterization would be involved. What this means is that we deal with a number of physical properties and relations and several physical particulars that are the instances of the properties and the terms of the relations. One may then speak of a physiological state as being a complex or macro particular consisting of relatively micro particulars of certain kinds in certain relations. The phenomenal particular would then be 'identified' with such a macro state and the property B with the complex property that characterizes such a state. But, notice, how another favorite analogy of the materialist now dissolves. Materialists suggest that just as a macro physical object is really a collection of particles with certain properties and standing in certain relations, so a phenomenal particular is really a collection of physical particulars.The analogy is grossly misleading. In the case of a macro physical object the claim, which itself is problematic, is at least based on the fact that some physical particulars, the micro objects, are literally parts of others, the macro objects, in a straight-for­ ward use of the term 'part' - i. e., as a spatial term. What the materialist does is something quite different from what is done when one claims that a macro object is really a collection of micro objects. The materialist identifies a phenomenal object with a macro state or simply declares that no phenomenal objects or states exist. Taking the macro state as a collection of micro objects, with certain properties and standing in certain relations, he then identifies the phenomenal object with the micro state. But the objects he ends with are not objects that are spatial parts of the object he started out with, as we initially construe the phenomenal object or state.Thus, all the materialist does is claim that what is really there are the physical states (whether he takes macro physical states to be reduced to micro states or not).For this assertion, he has no argument, only a challenge that the defender of mental entities refute his 'postulate' . The refutation is the classic Cartesian argument. On the materialist's claim we either do not deal with an empirical correlation between mental and physical events or it is not possible to have mental events without the 'corresponding' physical events.In so far as we acknowledge that we deal with empirical correlations and recognize the logical possibility of such a correlation not obtaining, materialism fails.7 To avoid acknowledging such a possibility the materialist turns statements of empirical fact into analytic truths. Nor 386

must we be misled by a standard attempt to avoid this proble m . Holding that mentalistic terms and physicalistic terms do not mean the same thing , materialists sometimes argue that the statements of lawful connection between ' mental ' and physical events are not analytic . Hence , they do not turn e mpirical laws into analytic truths . Here we clearly see how the materialist makes use of themes from nominalism and holism . For, if ' B ' is taken t o b e a primitive predicate , then i f it is taken t o stand for a phenomenal property , there is no question that the correlational statements will not be analytic truths. But , there is also no way in which we can deny the possibility of there being an instance of B without the correlated physical state obtaining. Hence , there can be no use of physical properties with which or in terms of which we can identify or eliminate B . It is only if " B ' is permitted to be introduced into a schema · without either being correlated to a property or defined that the materi alist can suggest that it is not analytic to claim that an instance of B obtains if and only if certain physical conditions obtain , and yet ' B ' does not stand for a phenomenal property . Thus, the materialist can have both an empirical correlation and a claim of identity between mental and physical events . But, in so far as neither nominalism nor holism (regarding the me aning of terms) will do , this defense of materialism will not do either. Notes 1

3

4

5 6

7

The insistence on the arbitrari ness of the process of definition and redefinition also reveals a fai lure to separate the context of discovery from that of j ustification . The framing and revising of definitions in the context of discovery is far from 'arbitrary' . The so-called ·arbitrariness' simply reflects the fact that definitions are abbreviations and , hence , analytic 'truths' or, as some would misleadingly put it, 'l inguistic truths' . This latter feature we note in the con text of j ustification , i. e . , when we are concerned with the logic of definitions. It would be strange indeed for Putn am to identify strict and explicit definition . For what could he then mean by holding ( I ) that 'anger' is defined by a set , say S, of macrosymptoms , and also holding (2) that 'anger' means ( or anger is) a micro-state? But see the discussion of intervening variables and abstract entities below . For a discussion of the problems surrounding the phrase 'implicit definition' see my · Axiomatic Systems, Formalization , and Scientific Theories' , in Symposium on Sociological Theory, ed. L. Gross (Evanston : 1 959) , pp . 407-36. This raises certain questions about the 'meani ng' of some terms found in the partially interpreted calculi of the physicist , i . e . , the issues surrounding 'fictionalism ' . But one need not go into those , I believe , in connection with the explication of psychological terms. For an elaboration of this point see G . B ergmann , 'The Contribution of John B . Watson ' , Psychological Review, 63 , 4 ( 1 956) , pp. 265-76 . It should be clear that I am not claiming that there are special 'philosophical things ' , whatever t h a t might mean , b u t o n l y that t h e scientist , a s such , does not answer, n o r ask , philosophical questions . On this see my 'Belief and Intention ' , in this vol ume , pp . 388 f.

387

Belief and Intention

I believe that Jan loves Susanne . On one view held by philosophers such a sentence is expressed perspicuously or reveals its logical form i n a sentence containing a two-term relational predicate , 'Bel 2 ' for 'believes' , a singular term 'I' , and a second term which is itself a sentence , 'L(Jan , Susanne)' for 'Jan loves Susanne' . We thus get: ( 1 ) Bel2 [ 1 , L(Jan , Susanne)] reflecting the analysis of the opening sentence of this article . Taking the sentence 'L(J an , Susanne)' as a term in ( 1 ) is then held to reflect the use of the sentence to stand for a proposition or other 'intentional ' obj ect . This recognition of intentional entities is thought to pose a problem for such a view . A second problem is thought to ari se when we consider the true identity. (2) Susanne = Elisabeth . For, if we hold to the view that singular terms may replace other singular terms referring to the same entity , the n (3) Bel 2 [1 , L(Jan , Elisabeth)] results from ( 1 ) and (2) by application of such a rule . B ut, if l do not know that (2) is true , I may deny that (3) holds . Moreover, even if I know that (2) is true , I may deny that I believe what is expressed by ' L(Jan , Elisabeth)' in the sense in which I do believe what is expressed by "L(Jan , Susanne)' . Does the view which transcribes the belief sentence i nto ( 1 ) thus abandon the rule for the replacement of singular terms? A further problem concerns the question of existential generalization from ( 1 ) . Can we derive (4 ) Bel 2 [ 1 , (3x) L(Jan , x)] or (5 ) (3x)Bel 2 [ 1 , L( J an , x) ] 388

or both, or neither from (1 )? Moreover, just what is the proposition that I believe according to (5 )? A second way of treating our original sentence is to analyze it in terms of a four-term relation Bel4 holding among the three individuals and the two-term relation loves. Thus, we get (6) Bel4 (1 , L, Jan, Susanne). One supposed merit of this alternative is that we no longer need recognize propositions, or similar entities, correlated with the sentence "L(Jan, Susanne)' as we do on the first alternative.Moreover, some feel that from (6) and (2) we may unproblematically go to (7) Bel4 (1 , L, Jan, Elisabeth) and that (6) yields (8) (3x)Bel4 (1 , L , Jan, x) in a straightforward manner. However, on the first alternative we can distinguish between (9) I believe that there is someone whom Jan loves , taken as (4), and (10) There is someone whom I believe Jan loves taken as (5). On the second view we must reflect (9) by something like (1 1) Bel3 [I, (3x)L(y,x),Jan] where'(3x)L(y,x)' is taken as a predicate 'true of' Jan, i. e., it stands for a relational property of Jan. (8) can then be taken as the transcription of (10). Russell once advocated the second view.1 He gave it up for two reasons.Following Frege, he held that a predicate could not function as a subject term in a sentence and still play its'predicative' role.This he took to be especially pointed in the case of relational predicates. Thus, Russell believed that any sentence of the form 'F(g)', where 'g' and 'F' were constant predicates of appropriate types, was elliptical for a sentence in 389

which 'g' occurred only as a predicate and 'F' did not occur. 2 In this sense, there were no descriptive properties of the higher types for Russell.But ( 6) does not jibe with this view of properties and relations.This provides Russell with a reason for rejecting (6) as an analysis. One need not subscribe to the general Russell-Frege view, regarding predicates used as subject terms, to hold that (6) involves a problematic use of 'L'.In (6) we pack into the relation Bel4 not only the ordering of its terms but the ordering of the terms of L as well.For the ordering of the terms of the predicate 'Bel4 ' in (6) is so construed that it preserves the ordering of the terms 'Jan' and 'Susanne' in the sentence 'L(Jan, Susanne)'. To put it another way, the relation Bel4 is quite different from a relation like between. In the case of the latter and a sentence 'between (I, Jan, Susanne) ' there is an understood ordering in the sense of which terms in which order would fill in the gaps in '...is between ... and ...'.But in the case of Bel4 we get '...believes that ....... ..' and the filling in of the last three places gives us a sentence. One case shows that between is a three term relation ; the other shows that construing believes as a four term relation merely amounts to a disguised way of writing a two term relation, where one linguistic term is a sentence.This comes out when we realize that we will also have a three term relation as in (1 1), a five term relation for 'I believe that I am between Jan and Susanne', etc. What we get is an exercise in linguistic slight of hand, rather than a philosophical analysis.It is no accident that the view is currently offered3 by a philosopher who dispenses with names, 'Pegasus', for example, by defining a predicate 'Pegasizes', in terms of 'x = Pegasus', and then 'replaces' the name by the predicate. Just as we may delude ourselves that we have eliminated the name, we may delude ourselves into thinking that ( 6) is more than a mere rewriting of (1).In a way this is what Russell came to see on the basis of a criticism by Wittgenstein.As Wittgenstein put it we make use of the rule that in Bel4 (I , f, x, y) the sequence 'f(x,y)' is a well-formed sentence, i . e., not nonsense. In short, we explain ( 6) in terms of ( 1 ). Another problem arises for the second view due to the forced use of '(3x)L(y ,x) ' as a predicate. We must recognize properties corresponding to such predicates in addition to the relational property indicated by 'L', or 'Lxy' if one prefers.This is not something we must do in the case of nonintentional contexts. To say that Jan has the relational property of loving someone is to say that Jan stands in the relation of loving to someone.To put it another way , to say '(3x)L(y ,x)' is true of Jan is to say that '(3x)L(Jan, x)' is true, but to make such a replacement in 390

'Bel 3 [I , (3x)L(y ,x) , Jan]' is to arrive at 'Bel 2 [I , (3x) L(Jan, x) ]'. Thus one makes use of the notational difference between a one-term 'predicate' like ' (3x)L(y ,x) ' and a two-term relational predicate like 'L' to hold that 'Bel3 [1 , (3x)L(y,x),Jan]' is different from 'Bel 2 [I , (3x)L(Jan, x) ) '. This comes out quite explicitly if we consider the relational property of loving Susanne. We could then write: ( 1 2) Bel 3 [I , L(y. Susanne) , Jan] to express that I believe that Jan loves Susanne. But , it is clear that ( 1 2) is just another way of putting (6). Or, to put it succinctly, of stating (1) , if we just replace the variable by the name. Contrast 'Fa' with 'Ex (Fx ,a)' where 'Ex' is short for 'exemplifies'. The latter is hardly an insightful way of expressing the former. Hence , we are not forced to accept , by the use of ( 12 ) , a relational property indicated by 'L(y , Susanne)'. However, ( 1 1 ) cannot be considered as a paraphrase for a sentence involving only a belief relation, the singular terms 'I' , 'Jan', and 'Susanne' , and the relational predicate 'L'. Once again , to avoid using sentences as terms , one makes use of a notational device. This time, however , one is forced to recognize relational properties in addition to relations . The ontological price already paid was the recognition of an indefinite number of belief relations of logically different kinds. What such ontological commitments may be taken to reveal is that the attempt to use ( 6) and (1 1 ) amounts to a notationally , rather than a substantively, different view from that using ( 1 ). If this is correct the relational view will not do. But , how do we handle the problems mentioned in connection with what we may call the propositional view that employs ( 1 ) ? Following Moore and Bergmann I will speak of acts or states of belief and, using Bergmann's notation, consider the sentence 'L(Jan, Susanne)' put inside corner quotes to be a predicate standing for a property of such mental states or acts which , as we may say , have the content expressed by the sentence. Let 'm 1 ' , 'm2 ' , etc. , be individual constants standing for particular acts or states , and let 'Bel 1 ' be a predicate standing for the generic one term property that all such acts which are beliefs , rather than doubts for example, have. Thus ,

states that the act m 1 has a certain content and is a belief. Let 'H' stand for a relation between a person and one of his mental states.

39 1

then states that the act m 1 is an act of mine. To say that I believe that Jan loves Susanne may then be expressed by (15) (3x)[rL(Jan, Susanne)7(x) & Bel 1 (x) & H(I ,x)] . The introduction of predicates like rL(Jan, Susanne)7 involves the tacit acceptance of a rule to the effect that the propositional characters the predicates stand for mean or intend the state of affairs expressed by the sentence from which the predicates are formed. As Bergmann puts it, we have a further intentional predicate 'M' such that

holds for every sentence p. Consider the sentence (16) rL(Jan, Susanne)7 M L(Jan, Susanne). We may infer ( 17) (3x)[�(Jan, Susanne)7 M L (Jan, x)] from (16) as a straightforward case of existential generalization, but we could not go to rL(Jan, Susanne)7 M (3x)L(Jan, x), for the propositional character does not mean what is expressed by the sentence '(3x)L(Jan, x)' . Moreover , (3x) rL(Jan , x)7 M L(Jan, Susanne) would be a senseless pattern and, hence , not a consequence of (16). It is senseless since , rL(J an, x)7' is not a propositional predicate as it is not formed from a sentence and hence does not mean what a sentence expresses. ( 17), by contrast , is not senseless since it expresses thatthere is someone such that the propositional character �(Jan, Susanne)7 means a state of affairs which consists of Jan loving that someone. And that is true, on a view where all names name.This, of course, is not to say that the state of affairs exists, i. e., that "L(Jan, Susanne)' is true. Consider next the sentences (18) (3x)[rL(Jan , Elisabeth)7 (x) & Bel 1 (x) & H(I ,x)] 392

and (1 9) (3x) [ r (3y)L(Jan, y)7 (x) & Bel 1 (x) & H( I, x) ]. We can infer neither ( 1 8) nor ( 1 9) from ( 1 5). To do so would be to infer that because one mental act occurs in a person's mental history a mental act with a different content also occurs. We could of course introduce a further sense of belief, whereby if a mental act of belief which exemplifies a propositional character r p7 is had by an individual then if 'p' implies 'q' one may say in this new sense of belief that the individual believes that q. One may extend this to cover cases where 'p' and "q' differ only in the replacement of a name by another name of the same individual or in a replacement of definitional equivalence. In yet a third sense of belief one may say that an individual believes that q only if he has a state of belief expressed by "" ' p' implies · q"' or the identity statement, i. e. , if he believes in the first sense that the implication or identity holds. By distinguishing the three senses of belief and noting that in the basic sense we started with ( 1 8) and ( 1 9) do not follow from (1 5) , we resolve two of the puzzles we began with; those based on (3) and (4) . Note, moreover, that the question of the replacement of names, on the basis of true identities like (2), in belief contexts, does not rest on whether one knows or believes the identity to hold. Even if I believe or know that the identity holds, it does not follow from m y having a mental state exemplifying one propositional character that I have a mental state exemplifying another propositional character. Here, however, one may object to the claim that the propositional characters are different. Consider the sentence (20) �(Jan, Susanne) 7 M L(Jan, Elisabeth) . Is this not an allowable consequence from (2) and (1 6)? And , if so, since �(Jan, Elisabeth) 7 M L(Jan, Elisabeth) do not the propositional characters rL(Jan, Susanne) 7 and �(Jan, Elisabeth) 7 mean the same fact and , hence, are they not the same property? Bergmann holds that (20) is false. In doing so he does not appear to realize that he contradicts his theme that propositional characters mean possible states of affairs or facts, since statements of the form , r p7 Mp' introduce or give meaning to the propositional predicates. To save his view that (20) is not a consequence of (1 6) and (2) , he must either hold that 'L(Jan, Susanne) ' stands for a different fact than that indicated by 'L(Jan, Elisabeth)' or ban different names for the same thing from his 393

ideal language . The former would be paradoxical , while the latter smacks of the dogmatic and ad hoc. An alternative is to allow for the replacement and accept (20) as a consequence of (16) and (2) . This means that some replacements are permitted in intentional contexts , but not inside the corner quotes . This also means that different propositional characters may mean or intend the same fact . Such characters, or propositions , would differ since the sentences used to construct the predicates referring to them differ . I shall elaborate on this theme shortly . For the moment let us deal with the existential claim expressed by (5) on the view sketched out . One thing we may write is (21 ) ( 3x)( 3 f) [f(x) & Bel 1 (x) & H(I , x) & f M L(Jan , Susanne)] . Since we have rej ected Bergmann's refusal to allow only one propositional character to mean or intend a possible fact , (21 ) does capture one way of taking (5) . For , according to (2 1 ) , I have a mental state which intends that Jan loves Susanne , but my mental state is not characteri zed in terms of the names 'Susanne' or 'Jan ' . Thus , (2 1 ) suggests that there is someone whom I believe that Jan loves, b u t it does not suggest that I believe what is expressed by the sentence ' ( 3 x)L(Jan , x) ' . However, we may take (22) ( 3x)( 3 y)( 3 f)[f(x) & Bel 1 (x) & H(I , x) & f M L(Jan , y)] , to better reflect what is intended by (5 ) , since (21 ) may also be taken to reflect (3) . Note that since sentences like

are consequences of the rules for the corner quotes and the sign ·M' , we can consider (23) ( 3 x) [ri(Jan , Susanne) 7 (x) & Bel 1 (x) & H(I , x) & ri (Jan , Susanne ) 7 M L( Jan , Susanne)] equivalent , in such a system , to ( 15) . Since , (22) is a logical consequence of (23) , we have answered the question as to whether (5) is a logical consequence of ( 1 ) , since (5) and ( 1 ) are transcri bed as (22) and ( 15 ) , respectively . We have also answered the question regarding what proposition I believe according to (5) , since on this analysis it is any proposition which intends , or stands in M to , the fact that Jan loves Susanne , without specifying j ust which proposition. (5) , rendered as (22) , i nvolves an existential reference to a proposition , if I may so put it . 394

In fact, since (21) may be taken as rendering (5), understood slightly differently, we may take our analysis to reveal an ambiguity about (5) in addition to the original ambiguity that led to the distinction between ( 4) and (5). Note, however , that (2 1) , like (22) , is a conseq uence of (23) . What we have just seen is how we can ·quantify in' , to use a now popular phrase. Yet , we can do so without the problematic consequences of the neo-Fregean views of Wilfrid Sellars and David Kaplan or the problems faced by the relational analysis of Russell and Quine. Consider a problematic case Quine has made notorious. On two occasions one man , Ralph, sees another, Ortcutt. On one occasion, Ralph believes the man he sees to be a spy. The second time , not recognizing the man he sees as the man he saw , he believes the man he sees not to be a spy. On Quine 's version of the relational view, the two belief situations are expressed as: (0 1 5 ) Ralph believes z (z is a spy) of Ortcutt. and

(0 2:) Ralph believes z (z is not a spy) of Ortcutt . And Quine concludes : Thus ( 1 5 ) and (22) both count as true . This is not , however , to charge Ralph with contradictory beliefs . Such a charge might reasonably be read into : (23) Ralph believes z (z is a spy . z is not a spy) of Ortcutt . but this merely goes to show that it is undesirable to look upon ( 1 5 ) and (22) as implying (23 ) . 4

What is puzzling about Quine 's conclusion is that there is a sense in which Ralph does have contradictory beliefs, but, by adopting the relational view for the rendition of some statements of belief, Quine handles belief in such a way that we cannot say what it is that is believed in the one case that contradicts what is believed in the second case. Yet, he is also correct in holding that Ralph does not have contradictory beliefs ; for in some sense , he does not. The point is to clarify that sense as well as the sense in which he does. The pattern we have developed provides a ready answer, in fact it enables us to make some further relevant distinctions. In our terms, we construe (0 1 5 ) to state that there is a belief state which Ralph has and which has a propositional character that intends the possibility that Ortcutt is a spy, while taking (0 22) to state that there is a belief state that intends that Ortcutt is not a spy. We may then recognize senses in which there are as well as senses in which there are not

395

contradictory beliefs. Ralph has contradictory beliefs in that the intended situations are contradictory . He does not have contradictory beliefs in the sense that the construals of (0 1 5 ) and (0 22 ) do not involve the use of propositional characters which are construed from contradictory sentences (we make use of a quantified predicate variable in each case). This reflects what we might ordinarily express by saying that he has contradictory beliefs without knowing that he does, since he does not believe the requisite identity. There is another point. Since we express his having a belief in terms of an existential claim about a particular belief that Ralph has,

( 0 1 5 ,) ,...., Ralph believes z (z is a spy) of Ortcutt, is naturally rendered as the claim that every belief state of Ralph's is not a state that has a propositional character intending that Ortcutt is a spy. This is not equivalentto ( 0 22 ) , nor doesit followfrom ( 0 22) . Hence, there is a further sense in which Ralph does not have contradictory beliefs. Finally, as Quine notes, it is not the case that there is a belief state of Ralph's that has a contradictory sentence as the basis for the propositional character it exemplifies. But, there is one awkward feature of Quine's account.Let us change the situation to a case where Ralph is aware that it is the same man on both occasions, but where Ralph never obliges us with a belief state that exemplifies a contradictory propositional character. For Quine, (0 1 5 ) and (0 22 ) would still describe the two situations and Ralph would still not have contradictory beliefs. This will not do. Kaplan's criticism of Quine will not do either. He holds that For in the same sense in which ( 0 22 ) and (0 1 5 ) do not express an inconsistency on Ralph's part , neither should (0 1 5 ) and (0 1 5 . ) express an inconsistency on ours . Indeed it seems natural to claim that ( 0 5 .) is a consequence of ( 0 22 ) . B ut the 1 temptation to look upon (0 1 5 ) and (0 1 y ) as contradictory is extremely difficult to resist . 5

But, it is not at all 'in the same sense' that we may hold that (0 1 5 ) and (0 15 ,) do not express an inconsistency.As we just construed them, they clearly do. What Kaplan apparently has in mind is that if one can truly say of Ralph that he has the belief that Ortcutt is not a spy , then (0 1 y ) is true.His next sentence bears that out.But , this appears to confuse the sense in which Ralph may be said not to have a belief, when no belief he has is a belief of that kind, with the sense in which he does not have the belief if he has a belief in the negation of the relevant claim. I say 'appears to confuse' because Kaplan proceeds to make a very similar distinction in his own terms. Before getting to that, we might note more precisely why his claim 396

that 'in the same sense' ( Q 1 5 ) and ( Q 1 ,J should not express an inconsistency is mistaken . ( Q 15 ) could be rendered as (E 1 ) (3f) (f = x is a spy & Bel (Ralph , f, Ortcutt)) and ( Q 1 y) would be (E 2 ) (3f) (f = x is a spy & -- Bel ( Ralph , f, Ortcutt) ) , assuming Quine would not shirk from the use o f the quantifier. He should not , since he is using predicate abstracts in subj ect place and has temporarily acknowledged "intensions' as ·creatures of darkness' that will be left behind when he eventually semantically 'ascends' . Recognizing such ·creatures of darkness' , however temporary the association is , one would recognize a suitable comprehension condition for · attributes' . Thus , given the predicate ·x is a spy' , we have ' (3f) (f = x is a spy ) ' as a "theorem' of the schema employed . Hence , ( E 1 ) and (E 2 ) are contradictories . Or. if one balks , the point may be put in terms of (3f) (f = x is a spy) & (E 1 ) & (E J being a contradiction . We may also render (0 22 ) as (E 3 ) (3f) (f = x is not a spy & Bel (Ralph , f, Ortcutt)) . We then clearly see that (0 1 5 . ) is not a consequence of (0 22 ) . One line of Kaplan's criticism thus misses the mark . Kaplan mixes the question of (0 1 5 ) and (0 15 .) being contradictories with another question . This latter deals with the supposed inability of Quine 's formulation to distinguish two cases that depend on the characterization of beliefs in terms of the signs used to 'express' their content . In a way , this criticism is beside the point , since , first , Quine has acknowledged that such distinctions can be made , but he finds them to be problematic, and , second , it is not clear that Quine's semantically ascended view cannot embody those distinctions , for he does retain beliefs in sentences . Where Kaplan is correct is that in accepting 'intensional' entities like propositions and 'attributes' , Quine has not acknowledged 'senses' of individuals and quantification over senses , concepts , and propositions . Moreover, Quine has not clearly specified the relation between belief sentences, taken 'intensionally' , and semantically 'ascended' versions ; nor has he clearly stated how he 397

construes the latter. We need not bother with Kaplan's notion of a 'vivid name' and its reflection of a mixing of questions about causal conditions, behavioral responses, introspective ruminations, Fregean entities, and philosophical problems about 'reference'. What he is claiming is that if we 'break-up' the Fregean propositions into typical Fregean constituents and quantify over these, we can distinguish cases that Quine does not distinguish on his analysis. A relevant case would be where we distinguish the claim: (I) There is a propos1t1on and a constituent sense that denotes Ortcutt and the proposition contains the concept is a spy connected to the sense, and it is not the case that Ralph believes that proposition. from the claim: (II) For any proposition that contains the concept is a spy connected to a sense that denotes Ortcutt , it is not the case that Ralph believes that proposition. This presents a distinction that involves a simpler case than Kaplan's, since I do not bother with talking about terms in Ralph' s vocabulary that express the sense in question. The point is that certain distinctions necessitate construing the proposition or sentence as a complex with constituents. But, now, if one takes the gambit to literally involve us with propositions, senses, and concepts , Kaplan has not presented us with an account that renders such entities unproblematic or, for that matter, with any detailed account at all. Were he to work one out, it would likely amount to a variant of Bergmann's. The heart of the latter analysis, recall, is the employment of specific sentence patterns in the account of belief, through the use of ' M' and sent ences to the right of that sign in patterns like ,r . . . 7M. . . ' . If Kaplan' s account involves believing sentences, rather than propositions, he has not made clear what this could mean, or even whether it would significantly differ in structure from the first alternative. Quine has also suggested such a view, and it is hard to imagine that one who speaks of believing sentences will not eventually treat sentences as composed of words and , consequently, acknowledge patterns like 'there is a word and there is a sentence such that the word is a part of the sentence . . . ' as well as distinguish between believing different 'sentences' with words for the ' same thing' . There is an obvious problem with Kaplan's analysis that is indicated by my use of (I) and ( 1 1 ) above. Suppose we take the corners to turn sentences into names of propositions as, on Bergmann's view, one turns 398

them into names of propositional characters. How do we then transcribe the mixture of symbols and English (.3a)Bel(Ralph,r a is a spy7 ) into an English sentence? We cannot say that there is a sense such that Ralph believes that .. . What we must say is something on the order of the following. There is a proposition and there is a sense that is a constituent of it, and the concept is a spy is a constituent of the proposition, and the form of the proposition is such that the concept is a spy is predicated of the sense and Ralph believes the proposition. I shall readily agree that my use of the phrase ·English sentence' above requires a bit of charity on the reader's part. But , what is clear is that we 'quantify over' a variety of entities as well as employ notions like constituent and form. We also make use of a predicative connection within the proposition and of a referential connection between senses and objects. Kaplan's understanding of his ·convention' covers up all this. Once we have unpacked it, as it were, it becomes clear that Kaplan has not quantified into corner contexts without radically complicating the notion of a corner context that we started with. Consider the case of forming a name of a sign, say ' Desdemona', by putting standard quotes around it, as I just did, and then considering , (.3x)( 'Desdemonx ) to be an existential quantification over letters and read as "There is a letter such that placed for the 'x' in 'Desdemonx' the latter becomes a name". Clearly, we are no longer merely using the quoting device according to the original convention, for we are not rejecting '"'Desdemonx'" as gibberish nor treating itas a name of the pattern within the quotes. One may raise a further problem concerning Bergmann's analysis. If we allow the substitution , based on (2), in (16) to obtain (20), must we not also allow one to derive (24{L(Jan, Elisabeth)7 M L(Jan, Susanne) from (16)? For both rL(Jan, Susanne)7 and�(Jan , Elisabeth)7 intend the same fact or possibility. On Bergmann 's view, as I see it, to allow the replacement to the right of M forces one to allow it to the left as well, since, on his view , the propositional character in question is the one intending a certain possibility. Thus, he must ban replacements to the right of M or acknowledge that the sentences 'L(Jan, Susanne)' and 399

'L(Jan , Elisabeth)' indicate different possible facts . The latter is unpalatable and the former, as we noted , would be an arbitrary ad hoc move . In so far as sentences of the form ,rp7 M p' state a relation between a propositional character and a possibility or possible fact , one may allow replacements on either side of the sign 'M' , which result in a truthful statement of such a connection . Hence , replacements are allowable to the right of 'M' which yield sentences to the right that indicate the same fact , and replacements are allowable to the left which yield , not a predicate referring to the same propositional character, but one referring to a propositional character that intends the same fact as the original predicate to the left of 'M' . (24) may then be derived from ( 1 6) via (2) . But this does not mean that (25) (3x) [�(Jan , Susanne) 7 (x) & Bel 1 (x)] and (2) yield (26) (3x) [�(Jan, Elisabeth) 7 (x) & Bel 1 (x)] , for the propositional characters indicated by the different predicates in (25) and (26) are different even though they intend the same fact . To see what is involved consider two particular mental states occurring at different times , but both consisting of the auditory sensations one gets if he pronounces aloud, but to himself, the sentences ·Jan loves Susanne · . Call these two states 'm 1 ' and 'm2 ' . Let 'm 3 ' and " m4 ' refer to two other particular mental states occurring at different times but both consisting of phenomena one experiences when he ·says to himself' , but does not pronounce using his vocal chords , the same sentence . We have four distinct states in that there are four different acts occurring at different times, but m 1 and m 2 have a property , call it 'F 1 ' , which neither m 3 nor m 4 has . They have this property in virtue of consisting of auditory sensations . Similarly, m 3 and m 4 have a property , call it 'F2 ' , which neither m 1 nor m 2 has. F 1 is a property had by all auditory patterns that are like m 1 and m 2 in a way analogous to the way in which different tokens of a linguistic type are like each other. F., is a similar kind of property applying to mental states composed of phenomena like those contained in m 3 and m 4 • Compare , for example , F 1 and F 2 to properties characterizing respectively all tokens of the sentence "L(Jan , Susanne)' written in black ink and all tokens written in red ink . Both F 1 and F 2 intend or mean the fact or possibility that Jan loves Susanne , j ust as we might consider the type ­ in-black as a sentence and the type-in-red as a sentence both of which would mean or stand for the same 'thing' . Suppose further that all four states are beliefs and hence characterized by Bel 1 • In one sense , both F 1

400

and F 2 may be taken as the same propositional character , just as the type­ in-black and the type-in-red may be taken as the same sentence type. In another sense they are different characters or properties, just as the sentence types are different if we bring in the color of type in addition to the properties of shape and arrangement . They are the same in a relational sense in that they intend or mean the same fact. Thus , we might consider a sense of ·same' as follows: S i (F, G)

= df . (3p) [F M p & G M p] .

They are different in that they are two properties and not one , i.e. , with ' D' as a primitive predicate read as 'different' the sentence 'D(F., F2 ) ' is true. We can then take a further sense of ·same' to be given by S/F, G)

= df . - D(F, G).

Since ' S 1 (F 1 , FJ ' is true , one may say that m 1 , m2 , m3 , and m4 are individual states of belief but involving the same belief. Since '- S2 (F 1 , F2 ) ' is also true, one may say that m 1 and m3 do not involve the same belief, though m 1 and m2 do. Let us say that they do involve the same 1 belief but not the same, belief. Consider next two states, ms and m6, such that m s is like m l except that it contains an auditory sensation of the word "Elisabeth' instead of the word ·Susanne· and m6 is like m3 except for containing an 'inner pronouncement' of the word ' Elisabeth' . In view of these features let m" and m6 be said to have , respectively, the properties F3 and F4 . Since 'F 3 M L (Jan , Susanne) ' and 'F4 M L (Jan, Susanne) ' are both true, 'S 1 (F3 , F4 ) ' , · s l (F 3 , F l )', etc. , are also true. , ,__, s 2 (F l , F3) ', , __ s 2 (F 3 , F4)', etc. , are also true, however. Thus , ms and m6 involve the same 1 belief as each other and as m 1 , m2 , etc. , but they do not involve the same2 belief. Yet there is a sense in which m 1 , m2 , m3 and m4 are instances of the same belief in a way in which none of them is an instance of the same belief as ms or m6 . None of the first four states contain, so to speak, tokens of the name 'Elisabeth' . In virtue of this let m 1 , m2 , mJ and m4 be instances of the same3 belief, while m5 and m6 are the same 3 as each other but neither is the same3 as any of the preceding states. The sense of 'same' involved in these latter assertions both raises some problems and provides an occasion for some clarifications. To get at these let us revert , again, to purely linguistic cases considered in terms of a model schema. Suppose we consider a model subject-predicate schema construed along the lines of Russell and Whitehead's Principia symbolism with a set of zero level constants , 'a 1 ' , 'a2 ' , etc. , and constant predicates of the various types, 'Ff, 'Gf, etc. Consider a second such schema with, as we 40 1

might ordinarily say, the ' same symbols' only in a different type font, so that capital 'A' s are used for the individual constants. Let each individual and predicate constant of the one schema be said to be correlated to its obvious 'partner' in the other schema. Consider the ' names' and predicates of the first schema to be coordinated to, or interpreted into, certain individuals and properties. Let it be understood that such an interpretation also interprets the correlates in the second schema. Suppose further that we allow different constants in the first schema to be interpreted onto the same objects. Let ' a/ and 'a2 ' be two such constants corresponding to the same object. Consider the two sentence types 'F� (a1 ) ' and 'F� (a2 ) ' of the first schema and 'F� (A1 ) ' of the second schema. In one sense we have three types, yet in another sense 'F� (a1 ) ' and 'F� (A1 ) ' are the same in a way in which neither is the same as ' F� (a2 ) ' . Even if we say all three 'say the same thing', this does not reflect what is involved . For, via the correlation of 'a/ with ' A/ and of 'Fr with itself the interpretation of one sentence type provides an interpretation of the other, whereas, 'F� (a2 ) ' only says the "same thing' in virtue of a further act of interpretation or interpretation rule. Thus, we can consider tokens of the types 'F� (a1 ) ' and 'F� (A1 ) ' to be tokens of the same type in a sense, without considering tokens of 'F� (a2 ) ' to be tokens of that type even though tokens of all three types stand for or state the same fact. This is not to deny that in another sense tokens of 'F� (a 1 ) ' and of 'F� (A1 ) ' are tokens of different geometrical patterns and , hence, of different types. In an analogous way m1 , m2 , m3 , and m4 are instances of the same belief over and above their exemplifying propositional characters which mean the same state of affairs. In that case the correlation between auditory sensations and inner pronouncements of the same verbal patterns replaces the correlation between the signs 'a 1 ' and 'A 1 ' and hence of the relevant sentence patterns. Thus one can hold F 1 and F 2 to be the same3 without denying that they are not the same., or collapsing the sense of ' same3 ' with that of ' same l '. This brings out something our earlier example of red and black type glossed over. This third sense of same, whereby F 1 and F2 are the same3 , enables us to resolve a puzzle. For one obj ection to distinguishing the belief that Jan loves Susanne from the belief that Jan loves Elisabeth as we did earlier would be that we would be forced to distinguish the belief that Jan loves Susanne from itself since F 1 is a different propositional character from F2 • m 1 and m3 are instances of the same belief not only in that the same fact is intended , as in the case of m4 and m5 , but in thatF 1 and F2 are the same3 . The auditory sensations and inner pronouncements involved in m 1 and m3 are correlated in the way that 'a 1 ' and ' A 1 ' are correlated and not connected as the result of two independent coordinations of signs and things, as are ' a/ and 'a2 ' and as

402

are the components of m 1 and m e; . Thus we have yet another sense of belief, whereby sentences related.as ·F� (a 1 ) ' and ·F: (A 1 ) ' may be said to expr�ss the same belief. That is, mental acts exemplifying the appropriate propositional characters formed from the sentences with 'a l and "A i ' would be instances of the same belief in a way that instances of characters formed from sentences with "a 1 ' and "a2 ' would not be.In short, when Samuel Clemens was baptized tokens of a written sign pattern (many such patterns, of course), of auditory patterns, etc., were involved or ·covered'. The tokens of the name 'Mark Twain' involve another ceremony or set of circumstances. With 'S1 ' for 'same1 ' , as I have been using the latter, and 'Cor' standing for the notion o( an interpretation correlate we may say that SiF, G) = df.Cor (F, G). A reader may wish for a further spelling out of the relational predicate ·cor'. One reason for this could be the feeling that the notion of an interpretation correlate has not been made precise enough. When are signs correlates? When are tokens instances of the same pattern? Which patterns were covered by the baptismal ceremony involving Samuel Clemens? What is involved in such a ceremony that is relevant for a name being attached to its designatum? How does such a ceremony differ from the one whereby the name ' Mark Twain' became another name for Samuel Clemens? etc., etc. There questions are not relevant for our concerns, interesting though they may be for those who are interested in the detailed workings of natural languages. One point of employing simplified schemas or model sign systems is to enable us to separate questions of a philosophical or logical kind from questions that are more properly the concern of linguists, sociologists, anthropologists, and so forth. Unfortunately the phrases 'semantics' and 'semantics of natural language' often signal a confused blending of questions of the one sort with questions of the other kind. Here I ignore questions about when two tokens are tokens of 'A' and about the complexities of what is involved in the correlation of written and spoken signs in natural languages by considering a simple schema to avoid the first and a stipulated pairing to avoid the second. This involves my claiming that for the issues being discussed nothing is lost by so doing.The justification of such a claim lies in the consequent clarifications we get of the various strands of the classical puzzles and the insights we obtain regarding alternative ways of handling such puzzles. I would claim, but will not argue here, that the logical or philosophical problems involved are precisely those which can be retained by considering such simplified schemata. For my purposes here I believe I have been as specific as one need be about the term 'Cor' . 403

Another point may lie behind a desire for further specification of what the predicate 'Cor' involves. This returns us to Bergmann's discussion of the intentional relation M. A long standing criticism of Bergmann's view is that propositional predicates are formed from sentences yet refer to simple properties. Or , to put the matter linguistically, the predicates referring to the propositional characters are formed from sentences by use of the corner quotes and yet such predicates are taken as undefined or primitive terms. One feature ofBergmann's view responsible for such an objection would be eliminated if we defined a predicate ,rL(Jan, Susanne)7' as �(Jan, Susanne)7 = df. ('f)f M L (Jan, Susanne) and, in general, used the pattern, rp7 = df. ('f)f M p for every such 'corner' predicate. But this would force one to hold that F 1 , F2 , F3 , and F4 were the 'same' since we hold that 'all four' intend the same possibility. As I have considered the issue here , we have four different predicates all linked to the same possibility but referring to different properties. What properties? Consider F 1 and F2 . Any mental state exemplifying F 1 consists of a series of auditory sensations - the sound of 'Jan', of 'loves' , of 'Susanne' occurring in that order in more or less immediate succession. Such a sequence constitutes a fact composed of three particulars in a temporal relation. We may continue to refer to such facts by terms like 'm 1 ' , etc. , and introduce a relational predicate for a notion like 'contains' , say 'Con', so that we may consider a sentence like 'F 1 (m 1 ) ' to be elliptical for or defined by (I) (3x)(3y)(3z)[Con(m 1 , x) & Con(m 1 , y) & Con (m 1 , z) & 0 1 (x) & Oi (y) & OJ(z) & R 1 (x , y, z)] where the 'O's are used to specify that something is an auditory token of 'Jan', etc. , and 'R 1 ' stands for a relational property ordering the three tokens. By contrast the sentence 'Fim 3 ) ' would be elliptical for (II ) (3x)(3y)(3z) [ Con(m3 , x) & Con(m3 , y) & Con(m3 , z) & U 1 (x) & U 2 (y) & U 3 (z) & R / x, y, z)] where the 'U 's are ,used to specify that something is an inner pronouncement of ' Jan etc. In short , something that is a O 1 is a correlate of something that is a U I ' just as tokens of 'a 1 ' and tokens of ' A 1 ' were

404

correlates. R 2 may be taken as correlate of R 1 or as the same temporal and ordering relation. Particular mental states like m 1 and m 3are thus facts, much as Wittgenstein spoke of propositions being facts in the Tractatus. They and properties like F 1 and F2 may be spoken of as correlates, since their constituent particulars, the properties they exemplify, and the patterns they form are correlated. Just as we simplified matters by considering only one correlate for "a 1 ' , i. e., · A 1 ' , in our earlier example let us assume we have to deal with only auditory patterns and inner pronouncements. We can then consider a propositional predicate like •rL(J an , Susanne) 7' to be defined by (III) rL(Jan , Susanne) 7 (x)

= df. F 1 (x) v F/x) .

This refl ects the point that it is specious to consider 'things' which exemplify properties like F 1 , F2 , and rL(Jan , Susanne) 7 to be simples; just as it is specious , in view of (I) and (II), to take the predicate •rL(Jan , Susanne ) 7' to be a primitive predicate referring to a simple property. One problem associated with Bergmann's view thus disappears. A second problem is also avoided. Consider , once again , ( 1 6) rL(Jan , Susanne) 7 M L (Jan, Susanne). Bergmann problematically holds this to be a logical truth. We can see what is justified in his claim. ( 1 6) is merely a reflection of our interpretation rules for patterns (and their tokens) of the kind that exemplify the property rL(Jan , Susanne) 7. Resulting from such a stipulation or interpretation , ( 1 6) may be held to be a 'logical' or "linguistic' truth. But there are two other aspects that must not be overlooked. We have held that propositional patterns stand for possibilities. This claim is hardly a linguistic or logical truth. It involves a problematic philosophical assertion. Bergmann packed both claims into the declaration that sentences like •rp7 M p' were logically true , since 'M' stood for an intentional relation. As treated here , the use of 'M' merely indicates that we have stipulated an interpretation for certain patterns. It thus involves no mystery. Nor is there anything mysterious about a propositional character like �(Jan, Susanne) 7 holding of different kinds of states , in view of the notion of a correlation. Propositions, as peculiar entities, are no longer required. A state like m 1 does not 'intend' what it does in virtue of a peculiar thing called a proposition. It intends what it does since it contains tokens whose types have been interpreted the way they have. We have had ,of course , to speak of possibilities or possible facts. But this we must do in any case , to deal with the questions of how sentences are about the world and true or false of it. 6 Once we speak of

405

facts and possibilities, propos1t1ons,as philosophers like Bergmann employ them, are not needed. Thus, the problem of the connection between propositions and facts disappears. It was this question which made Bergmann's claim that (16) is a logical truth suspect, since, as we noted, part of what is involved in his view is the claim that a propositional character (proposition) and a possibility stand in the relation M . The view presented here recovers, I believe, what is sound in Bergmann's claim that statements like (16) are logical truths and in Moore's earlier point that we use the same sentence to stand for both a fact and a belief. But the present view also recovers what is sound in the objections raised to Bergmann's view, since it is obviously not analytic to claim that possible facts correspond to certain sentences. All we need do, to do justice to both lines of thought, is explicitly separate the claim that atomic sentences, and mental states whose content is expressed by such sentences, correspond to possible facts from the point that all mental states and sentences with that content are correlates composed of correlated constituents. In so doing, I have treated mental states as composites, like sentences, which are composed of tokens that are correlates of the linguistic tokens in sentences like 'L(Jan, Susanne)'. It is obvious that we can introduce a further predicate , say 'LF' , parallel to the definitions of 'F i ' and 'F2 ' in (I) and (II), where such a predicate would stand for a property that would characterize, not individual mental states,but written sentence tokens. Such a defined predicate could then be added as a further disjunct in a modification of (III). What is involved is merely the recognition that mental states of belief are propositional not in that they are related to special entities called propositions or exemplify special propositional characters, but in that they are, so to speak , literally sentence tokens. 7 On such a view one will still recognize generic properties like Bel 1 . If one takes such characteristics to be properties of acts then we are faced with a fundamental extension of logic in that we allow for properties of facts. For, unlike 'Fi (m 3 ) ' which is merely elliptical for (II), something like 'Bel 1 (m3 ) ' would require a basic extension of our syntax. Alternatively, one could take such generic characteristics to be properties of particulars, rather than facts,which accompany the appropriate facts. Whether one could satisfactorily work out such a view without having to predicate primitive monadic or relational predicates of expressions for facts is something I do not wish to pursue here. Rather , assuming such difficulties can be resolved on another occasion, I wish to take up a different problem. Suppose someone asks about the mental state m 1 , granting that it is a token of the kind it happens to be, why it intends what it does. In short, he 406

suggests that the notion of •intending' is ambiguous . In one sense , a mental state that is a token of a certain kind may be said to intend a situation (possible fact) since the type of the token has been interpreted the way it h as. But , consider the case where such a token occurs in what we may consider an absentminded way . The pattern , so to speak , 'goes through one' s mind' without the person believing, doubting , etc . Thus , in a second sense , the mental state , which is the instance of the pattern , does not intend the situation which the type represents or stands for . What is the difference ? Having generic properties such that states, as tokens , are said to be of a certain kind in that they are instances of such properties ( or connected in some way to such instances) takes care of the problem . For a state to intend in one sense , intend 1 , is for it to be a token of a type . For a state to intend in a second sense , intend 2 , is not only to be such a token but to be an instance of a generic characteristic of acts such as believing , doubting, remembering , etc . But now, a standard problem arises to complicate matters . Suppose we have interpreted certain geometric patterns ( auditory patterns , etc . ) in two ways so that , for example , the name ·Jan· is the name of two people . How does one determine which situation is represented by an occurrence of a token of�(J an , Susanne ) 7 , even where the token , which is a mental state , exemplifies an appropriate generic property of acts? For Bergmann , the problem does not arise , since he considers philosophical problems in terms of an ideal language which does not involve such ambiguities . If we follow him in such a restriction , it obviously does not arise on the present view either. But , suppose we consider the question without taking such a way out . One can say that here we deal with tokens of different types in that two tokens of the geometric ( or auditory , etc . ) pattern rL( Jan , Susanne ) 7 are not tokens of the sentence type since the sentence type is a pattern with a specific interpretation . To speak somewhat paradoxically , the name 'Jan' would not be involved since we have two names , not one , in that we have one name being the sign pattern with one interpretation , while the other name is the same sign pattern with a different interpretation . Yet , this distinction does not solve our problem , since we must now specify why a token is a token of one sentence type rather than of another sentence type where both share a geometric (auditory , etc . ) pattern . To hold that there is an additional content in or accompanying the act , which solves the problem, should be seen to be j ust a complicated and obscure way of doing what Bergmann does simply and neatly when he disallows arqbiguities in an ideal language . Moreover, it invites an obvious infinite regress . To talk in terms of prior causal features , dispositions , or contextual circumstances is irrelevant . Such factors are relevant to someone else determining what my mental states are or my resolving some problems

407

about some states of my own , but they do not solve our problem in that it is clearly possible to satisfy any such conditions and yet not have the appropriate state , or not satisfy them and have it. 8 The case is analogous to attempting to 'analyze' someone's being in pain in terms of his overt behavior , dispositions to behave , physiology, environment, etc. Insofar as we admit the logical possibility of having the behavior , physiology, etc. , without the phenomenal state , and vice versa, such an analysis is inadequate. We are dealing with a third sense of 'intend' (inten d 3 ) , for we have an intention ! since we have a token , and we have an intention 2 , ' since we have a generic property characterizing the token , say B el 1 , but we are now asking , 'In virtue of what does the token intend the situation it does in fact intend?' In one sense this question has no answer , since it is not in virtue of anything that the occurrence of such a pattern is a token of the type representing one situation rather than another. There is no account to be given in the sense that one can point to a feature or constituent of the mental state which provides the answer. Rather, one has to point to a feature of thought. Sartre has held that all acts, while directed to their objects and in that sense 'aware of' such objects or contents, are also self-awarenesses, in a basically different sense of 'awareness'. He is, I think , mistaken phenomenologically and ad hoc logically, but his pattern is suggestive for our concern. Acts of thought not only intend situations in that they are intentions 1 and intentions 2 , but they are awarenesses of which situations they do intend, and hence, in having the act one is aware of which situation is intended. It is this feature that makes them intentions in the third sense of 'intend'. 9 Such a "solution' invites the contemptuous response that nothing is said since what is said is that acts intend what they do because they intend what they do. But that is not correct for a number of reasons. Acts intend what they do in the first two senses of 'intend' because of inte rpretations of types and generic properties they have. They do not inte nd what they do in the third sense because ofanything. This simply means that in specifying the structure of the situations corresponding to belief sentences of the kind we have considered there is no eleme nt of such situations that grounds the fact that a token refers to what it does. To speak of a 'conglomeration of images , names, and partial descriptions which Ralph employs to bring x before his mind' 1 0 is to resort to an antiquated 'thought as image ' psychology as well as a problematic use of 'name'. It is also to confuse two quite different questions. There is one question regarding the ca usal conditions for the occurrence of the belief states we have considered . Answe ring such a question may involve talk of images or even of 'bringing things before one's mind' , whatever that might mean. In a similar way, one may talk about conditions, including physiological ones, for a subject being in pain. But , just as such correlated

408

conditions do not provide a description or analysis of the situation which obtains when one is i n pai n , they do not provide an analysis of what takes place when we refer to something. The second question involved is a question about the structure of the situation which does obtain when I refer to Jan. What is the analysis of the situation which obtai ns when such reference takes place, or , to put it another way, what is the fact corresponding to the sentence 'I believe that Jan loves Susanne? ' To claim that the two questions are distinct is not to deny either that there may be causal accounts for the occcurrence of such mental events or that such causal explanations will involve physiological and environmental factors. It no more does so than the denial that phenomenal entities or states , say a pain or the state of bei ng i n pain , are 'analyzable' or , ·reducible to behavioral , physiological, and environmental things and states involves the denial that the occurrence of such phenomenal entities is causally explainable in terms of such physical factors. All that is claimed is that such considerations do not suffice to cogently analyze statements of the kind we have considered. Thus, the italicized uses of 'because' above are to be understood as indicating the connection between an analysans and an analysandum and not a causal connection. I was, then , merely claiming that one sense of "intend' (or 'refer') , whereby we say a thought token intends or refers to what it does is not analyzable. 1 1 There is an interesting feature of the relation between causal accounts and accounts or analyses of the 'nature ' of thought. On the one side we find those li ke Sartre, who, in one way or another, convince themselves that thought involves the feature I have discussed in terms of "intend3 ' and illicitly conclude that such a feature is irreconcilable with causal explanation. Denying causality in such matters, Sartre proclaims "freedom' . On the other side , we have those who seek to specify the intentional nature of beliefs by recourse to the physiological, behavioral (including dispositions to utter linguistic tokens) , and environmental factors. Determined to insist on the causal efficacy of the physiological processes and behavior conditioning, they end up by denying not only the existence of intentionality but of mental entities and processes. Neither side seems able to render unto the other its due. What this may show is that in spite of the homage it has received Hume's lesson about causality and necessity is too difficult to digest. For , in so far as one treats causality in terms of 'constant correlations' , rather than in terms of "necessary connections', and insofar as we distinguish causal accounts from analyses , why should we not have our i ntentional states along with causal explanations of them? 1 2 Distinguishing the causal account from the question of the structure of the relevant fact enables us to safely ignore the criticism that all that has been said is that acts intend what they intend. That criticism is based on a 409

natural, but mistaken, tendency to treat the queries as to why a token refers to what it does and what it 'means' to refer as requests for a causal account that explains, in one obvious and legitimate sense of 'explain' , the fact that a token referring to someone occurs. Consider someone asking 'Why is Jones in pain' ? A normal response would not be 'There is a pain, a phenomenal entity, which Jones has'. In ordinary circumstances to account for Jones being in pain in terms of such a claim would be pointless and irrelevant . What is wanted in such circumstances is either a causal explanation for Jones being in pain (having a pain) or a specification of the evidence for believing that Jones is in pain. Yet , to speak of phenomenal entities is at the heart of the antiphysicalist' s account in the context of a philosophical dispute about mental entities. The same distinction is relevant for the present issue. We can then understand the temptation to appeal to causal accounts or elaborations of the ordinary circumstances of cases of reference as resulting from the treatment of our questions as either a request for specifying what conditions are causally connected with the use of a token to refer to someone or as a request for specifying how one knows that someone uses a token in such a way. But to so take matters is to overlook the problems we have been concerned with and to fail to recognize a distinctive feature of thought and reference. One may easily convince himself of this. Think of someone or something. You know that you refer and to whom or what you refer. But now ask yourself what you mean by saying that you refer, why1 3 the t hought is about whoever or whatever it is in fact about , and how you know that you refer and to what you refer? The questions point to three distinct claims involved in the discussion. First, reference is not analyzable by appeal to causal circumstances or what we normally take to be evidence that I refer to someone on a given occasion, even though such considerations obviously suffice to resolve nonphilosophical problems regarding cases of reference. Second , the intentional feature of a thought whereby a token on that occasion does refer to Jan is neither analyzable nor represented by an element of the situation that obtains. If I may so put it , only the referring token occurs as an element of the act of thought and not the fact that the token stands for what it does. Third , I sometimes know what I refer to without there being any further account of that fact to be given and , hence, anything more to be said . If one sums all this up by saying that all that has been said is that thoughts intend what they do because they intend what they do, the slogan makes a point but not a critical one. Causal accoun ts of intentionality are linked with so-called dispositional accounts of thought . On the analogy with a property like solubility, being taken as dispositional in that one holds that objects are soluble if they dissolve when placed in water, one holds that thinking410

that-p can be treated as a disposition to behavior B under conditions C. Thus , to say that I am thinking-that-p at time t is to say that at t I have the disposition 8, where we understand '8' in terms of

with the use of the "==>' in place of · :) ' to recognize the traditional problem of counterfactuals. That problem wil l not be our concern here, since the notion is that it is no more problematic for '8' than for 'soluble'. But, there are serious problems with the dispositional account. Consider a case of pain and ·associated' behavior that reveals that a subject is in pain. We normally take the pain to be the, or a, cause of the behavior or take the behavior to be a ·symptom' of the pain or of someone being in pain. To say of the subject , s , that he is in pain is thus to account for or 'explain' his behavioc it is not to state, in other words, that he behaves the way he is in fact behaving. In a similar fashion, we take my exhibiting behavior of kind B under C to be explained by my having a thought. The statement that a subject who is thinking-that-p will do B under C is taken to be a significant factual claim. It is not taken to be a short-hand way of expressing the trivial claim that a subject who will do B under C will do B under C. To avoid the triviality, the proponent of the dispositional account asserts that an 'analysis' is being offered. But, that is not the point. The analysis is such that upon it a significant statement becomes trivial . This points up a difference between the case of thinking-that-p and 'soluble' . To speak of a piece of sugar being soluble is to hold that the object is of a kind, sugar, and 'sugar is soluble' . 1 4 In short, there is a law involved, however implicitly or explicitly we put it. One may wonder what the corresponding law is in the case of 8, and consequently, what the 'kind' is that correlates with the disposition. Clearly , the relevant characteristic is the property of thinking-that-p. This would be the ' law' that would account for s having 8, i .e. , any subject that thinks-that-p does B under C. But to recognize such a ' law' is to give up the dispositional analysis ! One may , then, look for another characteristic to lawfully connect with 8. Two obvious candidates are ( 1) antecedent conditions that would causally expJain s thinking-that-p and (2) supposed correlates at t , physiological perhaps, of s thinking-that-p at t .Call these C 1 and C2 , respectively , then (a) (x) (C1 x :) 8x) {b) (x) (C2x= 8x) wil l both be nontrivial 'laws'. But, (a) and (b) do not really help, for the proponent of the dispositional analysis must stil l take thinking-that-p to 41 1

be stipulated to be 0 and, hence, still trivialize the claim that if s thinks­ that-p, then s does B under C. He must also hold that it is impossible to think-that-p without doing B under C, and vice versa, just as the classical behaviorist held that it is ' absurd' to claim that s could exhibit all the behavior associated with pain without being in pain, or be in pain without exhibiting any of the relevant symptoms. This is not to say that a causal account would not be forthcoming in that I have denied the possibility of laws like (a) and (b) or 'laws' linking mental states with behavioral states such as

or (d) (x) [ (3y) (H(x, y) &rp7y) =0x] . All that was denied was that (c) or ( d) yields an analysis of s thinks­ that-p. This is no more surprising than the claim that even if we have laws correlating physiological with macro-behavioral states, we cannot take the statements about the physiological states to constitute an analysis of the macro-behavioral statements in that the former explicate the latter. There is also a problem with talk about ' laws' connecting phenomenal states with behavioral states or with physiological states . They are more in the realm of the philosopher's imagination than in the domain of the psychologist' s or physiologist's field of research. This is why one has heard of that fanciful invention 'the auto-cerebrescope', which will enable one to both feel a pain and observe the corresponding physiological state. The problem is simple. It is that the private data of the phenomenal realm are not the sorts of things one deals with in the " intersubjective' realm of science. What the psychologist observes is the behavior of the subject, not the pain. Even if one puts matters in terms of observing that the-subject-is-in-pain so that the relevant characteristic becomes being-in-pain, a characteristic of an observable subject, rather than the property of being-a-pain, a property of a · private' phenomenal entity, matters do not change. For, the property of being-in-pain will be, as the psychologist uses it, a behavioral property, specified in terms of behavior and physiology. Thus , to speak of 'laws' in such contexts is irrelevant to our concerns in a very basic way. When properly specified , the laws are not laws that correlate physiological and phenomenal states . We can, of course, speculate about there being correlations between phenomenal and physiological or behavioral states . But, such speculations are more for the purpose of proposing philosophical positions rather than projections of future scientific discoveries . Of 412

course, the behavioral scientist and the physiologist operate in a common-sense-context, where one assumes that pain-behavior indicates pain and verbal responses go along with mental states. Yet, these notions do not belong in either the theories or laws proposed . In the common­ sense context in which the scientist operates, they function as a set of unquestioned assumptions, whereby we take for granted that behavior of a certain sort expresses or ind icates pain or that certain processes produce pain in a subject. It is granted that such common-sense assumptions are not questioned in other than philosophical disputes. The simple point is that they cannot, then. be a basis for advancing a philosophical position regarding mental states. The situation is no different from that regarding our common-sense knowledge of other minds. That is one thing. It is quite another thing to base a philosophy of mind on such 'knowledge' . What is an acceptable basis for our common-sense approach to people and situations is one thing. What is an adequate basis for a philosphical resolution of the classical puzzles about mind and the mental is quite another thing. It is blatantly question begging to use the one to resolve the other. or. at least, it returns us to the sort of approach we find in the ordinary language philosopher' s dismissal of the phenomenalist on the basis of ·correct' usage and our ordinary conceptual scheme. Our analysis of intentional contexts enables us to understand a misleading feature that appears to support the dispositional-behavioral analysis. As we speak or write, we engage in mental activity, just as we do when thinking or speaking to ourselves. What, then, are the mental states in such cases of " public' speaking? Convinced that such activities reflect or express mental states, some philosophers declare that such inner states must accompany or precede the " outer' behavior. Others, convinced that there are no inner states, in ad dition to the outer states, in such cases, stipulate that the obvious cases where we have inner states are to be construed in terms of dispositions to manifest outer states. Once again, each extreme distorts the apparent facts for the sake of some sort of conceptu al coherence. How, then, are we to construe cases of public ' expression of thought' ? Note, first, that the very phrase 'expression of thought' lends itself to the view that the public speaking or writing expresses some other " state' or 'thought'. But, how literally need we take such an expression? The proponent of the view that there is an inner thought may leap to the conclusion that there is some sort of vague isomorphism between an inner episode and the written or spoken sentence, thus inviting an opponent to press for a description of such a ghostly occurrence. Ironically, he might then feel forced to find it in the physiological apparatus and, thus, turn a defense of the mental into a form of materialism. Clearly, there is something besides the sentence as a pattern of sounds or blobs of ink, but 'it' need not be a mysterious state 41 3

that is isomorphic to the overt behavior that produces the sounds or script. Of course, there is also an 'inner' physiological state, butthatis not a mental state, though it may correlate in similar ways to an 'inner' thought and a'public expression' . What, then, is there in such cases of speaking and writing? The written or spoken marks are instances of interpreted propositional characters; thus, they, like inner thoughts, may be intentions in the first sense of the term. But, even then, let us distinguish between such tokens as written or spoken, i. e . , as the kind of product that they are as opposed to the physical objects that they are . There is the old puzzle about the monkeys typing Shakespearean sonnets or the wind forming 'sentences' in the sand. Given that it is not a product of a certain process, being written by a person who comprehends the language, it is not, in another sense, an expressive token. There is nothing really worth arguing about. But, there is a more important theme . In producing such a product, a person intends, in the second and third senses, exactly as he does in the case of inner episodes of thought. There are acts of inner thought, of inner speech, of outer speech, of writing (gesturing, too, if you will) that are all intentional acts, though the constituents of some of the acts (which, recall, are facts) are mental and those of other acts are physical. The inner thought that Jan loves Susanne is, thus, contrasted with the spoken declaration. But, both may be intentions in the second and third senses. Does this mean that'physical actions' can intend objects and possible situations? Yes and no. There is a difference between my writing and the wind 'writing' in the sand. My writing of a token of the name 'Scott' differs in a way that enables one to say that the token of the name is not merely the blob of ink or the mark in the sand. It is a blob of ink asformed in a certain way and context that is the name, just as much as itis the blob of ink in acertain geometricalshape that is the typographical token. Alternatively, it is the inner image of that typographical shape that is the token in the case of an 'inner thought' . As formed in a certain way, the typographical (sentential) ink patterns are tokens of thoughts , just as inner patterns of images are such tokens. We may have sentences composed of images and sentences composed of blobs of ink or marks in the sand. So long as we recognize the filling out that the phrase'composed of' requires, we can say that in a similar way thoughts may be composed of images or of externally written patterns. We may, then, ignore the absurd temptations to throw out the inner thoughts in favor of dispositions to "external thoughts' or postulate a 'hidden' or 'unconscious' inner thought for every external one . I may truly think out loud or in writing just as I may think to myself. I may just as truly have mental states with no present or future macro-physical correlate or manifestation, dispositional or overt. Again, this does not deny the possibility of correlations involving mental states and macro- or 4 14

micro- 'behavior'. It does not reflect on the general belief that every mental episode has a physiological correlate , for example. In fact , establishing 'common' physiological correlates for "inner' thoughts and intentional physical states would neatly fit with what has been said. Bergmann 's propositional characters and acts ground the connection between a thought and its content and, hence , between a name and 'its' referent. This is done in Fregean fashion , since the act exemplifies a property that stands in M to a (possible or actual) fact. Thus, exemplification and M connect the thought to its intention. On Frege's pattern a sentence expresses a proposition that denotes a truth value. Replace Frege 's truth values by the facts of a correspondence theory and the formal similarity is even more striking. Following Russell , I have tried to suggest a way of retaining non-ghostlike acts and linking them to states of affairs. The link between a token and an intention is provided by a basic intentional relation, which requires no further 'ground'. But a problem remains in connection with the use of tokens of a name or primitive predicate as labels for objects and properties. How do I know what object or property I refer to when I use a token - How do I know what the intended object or property is? We must distinguish two cases. First . the re are cases where the object and property are presented to me when I refe r to them. In some (though perhaps not all) such cases I clearly know what object and what property I intend. Thus , presented with a color patch , I judge That is white' . There is no problem regarding what object I refer to nor that I use the predicate 'white' to indicate the color of the patch. That the re is an implicit judgment that I use the predicate for a color consistent with past uses of the predicate is besides the point, for we can distinguish two aspects to the use of the predicate: (1) that it is used for the presented color property; (2) that it has been used for that property in the past. (2) clearly involves memory and , hence , involves a judgment that goes beyond the present expe rience. But, the issue at hand involves ( 1) and not (2). And it is clear that in some cases of a presented object and property one knows that one refers and to what one refers. Second, there are cases where the object and the property are not presented when the reference is made. Suppose one holds that when a token of a name or primitive predicate occurs , in the absence of the object or property , there is no basis for saying that we intend the object or property. In such cases there is only the occurrence of the tokens as tokens of types that have a 'role' in the complex network of past conditioning , behavior , etc. That is, it is suggested that to claim objects and properties are thought of when they are not apprehended as baseless. For, even the occurrence of images will not help , as the question arises about the intentional role of images. And, if there is a token of a description (occurring in thought), with or without an image, the problem occurs in connection with the token of the 415

predicate in the description. One may insist that there is a difference in the case of properties and that of particulars: that one can intend properties not presently apprehended, but not particulars. The traditional notion that we conceive properties, but only conceive of a particular in terms of properties, thus enters the discussion, as does the question of whether or not a particular is analyzed in terms of a complex of properties. Let me assume that there is no basis for distinguishing between particulars and properties in such matters. Hence, one either acknowledges that particulars, not presently apprehended, are intended or denies that properties, not presently apprehended, are intended. The solution to the problem lies with memory. I can remember an object once presented, just as I can remember a property. I cannot remember an object or property that was never presented. To intend a nonpresented object or property is to remember it. This does not mean that one cannot be mistaken, since memory is notoriously fallible . But, all that means is that one cannot be certain that he is intending an object or property in cases of thinking about nonpresented objects or properties. But, when do I remember, hence , intend, an objector property, given the occurrence of a thought? There are no criteria in one sense. In another sense, one can make use, in principle, of the 'causal network' within which such kinds of occurrences take place. What one must not do is confuse such uses of empirical correlations with philosophical analyses . Nevertheless, this leaves us with a problem about thoughts and intentions that matches the classical problem regarding the connection between experience and the (physical) objects of experience. Though , there is also a fundamental difference, since in the case of the classical problem of perception, one assumes that only phenomenal objects are directly apprehended. In the case of thoughts and the objects intended by them , there are cases where both the thoughts and the intended objects are so apprehended: the cases where one thinks or judges about presented objects and properties. Given such cases , there is no basis for a general sceptical question concerning the connection between a thought and its intention. But, as the certitude inherent in the case of the presented object and property is lacking in thoughts of nonpresented objects and properties, one must rely on context , the ·causal network' , and memory in such cases.

Notes 1 2

3 4

5 6

7

8 9 10 11

12

13 14

B. Russell , 'Philosophy of Logical Atomism' . reprinted in Logic and Knowledge, ed. by R . C . Marsh (New York : 1968), pp. 224-7. Ibid. , pp. 205-6. W.V . Quine , 'Quantifiers and Propositional Attitudes' . reprinted in Reference and Modality, ed . by L. Linsky, (Oxford University Press: 197 1) , pp. 104 , ff. Ibid, p. 106. D. Kaplan , 'Quantifying In' reprinted in Reference and Modality, pp. 1 1 2- 1 44. On this question see my 'Facts and Truth· . in this volume , pp . 279 f. , and 'Intentions, Facts, and Propositions· in The On tological Turn , ed . by M. Gram and E.D . Klemke , University of Iowa Press: 1974, pp. 168- 1 94. This, I believe , is one of Wittgenstein's themes in the Tractatus. That is, one m·ay determine in such a way what his own states were . As far as the causal conditions are concerned there are no first-person privileges. In terms of intends 3 , thoughts are not merely sentences . Wittgenstein either ignored or chose to deny this in the Tractatus. D. Kaplan , 'Quantifying In' , p. 136. To claim that beliefs and thoughts are to be understood in terms of the three senses of 'intend' is not to deny that sometimes one may be puzzled regarding the 'object' of thought ; it is to simply note that sometimes we are not. In so far as this point about reference and intention is cogent we must recognize the inadequacy of any form of physicalism or behaviorism. Of course , this is not to say that one who denies the intentional feature must reject mental entities . After all that has been said 'why' is not to be taken in a causal sense . On this point see my 'Dispositional Properties' , Philosophy ofScience, 34 , 1967 , 10-17 .

. \

Bibliography

Abbreviations of Journals A AJP AJPP APF APO BJPS D E I

JP JPL JSL

M

Me Mo

MP N

N DJFL p PAS PASS

PPR

PQ

PR

PS PSc

R RM

s

T

TNS

Analysis Australasian Journal of Philosophy Australian Journal of Philosophy and Psychology Acta Philosophica Fennica American Philosophical Quarterly B ritish Journal for the Philosophy of Science Dialectica Erkenntnis Inquiry Journal of Philosophy J oumal of Philosophical Logic Journ al of Symbolic Logic Mind Methodos The Monist Metaphilosophy Nous Notre Dame Journal of Formal Logic Philosophy Proceedi ngs of the Aristotelian Society Proceedings of the Ari stotelian Society Supplement Philosophy and Phenomenological Research Philosophical Quarterly Philosophical Review Philosophical Studies Philosophy of Science Ratio Review of Metaphysics Syn these Theoria The New Scholasticism

Abbreviations of Anthologies CPPL EBR EO EW FAP

Contemporary Perspectives in the Philosophy of Language , ed. H . Wettstein , P . French , and T. Uehling (Minneapolis: 1 978) . Essays on Bertrand Russell , ed. E . D . Klemke (Urbana: 1 970) . Essays on Ontology , E. Allaire , et. al . (The Hague : 1 963 ) . Essays o n Wittgenstein , ed . E . D . Kl emke ( Urbana: 1 97 1 ) . The Fo undations o f Analytic Philosophy, ed . H . Wettstei n, P . French , and T . Uehling ( Minneapolis: 198 1 ) .

419

L&K L&O M&E OT PBR PRC R&M S&PL SPGM U&P W&O

Essays on Logic and Knowledge , ed . R . C . Marsh (New York : Allen & Unwin, 1 968) . Logic and Ontology , ed . M . Munitz (New York : 1973) . Metaphysics and Explanation , ed. D . Merill and W. Capitan (Pittsburgh : 1 966) . The Ontological Turn , ed . M . Gram and E . D . Klemke (Iowa City: 1 974) . The Philosophy of Bertrand Russe11 , ed. P. Schilpp (Evanston : 1 944) . The Philosophy of Rudolph Carnap, ed . P . Schilpp (La Salle , Ill . : 1963) . Reference and Modality, ed. L . Linsky (Oxford : 197 1 ) . Semantics and the Philosophy o f Language , ed . L. Linsky (Urbana: 1 952) . Studies in the Philosophy of G . E . Moore , ed . E . D . Klemke (Chicago : 1969) . Universals and Particulars , ed. M . Loux (New York : 1970) . Words and Obj ections, ed. D . Davidson and J . Hintikka (Dordrecht: 1969) .

Aaron , R. I . Achinstein, P . Ackermann , R. Acton , H. B . Addis, L.

Ajdukiewciz, K. Aldrich , V. C. Allaire , E. B .

Allston , W . P. Altham , J . E. J . Ambrose . A. Anderson , A. R. ed . Anderson , A. R . and Belnap , N . D.

Anderson , J . Angelelli , I .

Anscombe , G . E . M.

Aquila , R. Armstrong , D . M.

1) The Theory of Universals (Oxford : 1967). 2) Two senses of the Word Universal, M, 68 , 1939. Rudolf Carnap , RM, 19, 1966. Perspicuous Languages, O T, 264-75. The Theory of Concrete Universals , I, M, 45 , 9. 1936 ; II, 46 , 1937 . 1) Moore and Ryle: Two Ontologists (The Hague: 1965) , with D. Lewis. 2) Particulars and Acquaintance , PSc, 34 , 1967. 3) The Philosophy of Gustav Bergmann , Algemeen Nederlands Tijdschrift voor Wijsbegeerte, 63 , 197 1 . 4 ) Presentness and the Ideal Language , M, 80 , 197 1 . 5 ) Time , Substance , and Analysis, O T, 148-67. On the Notion of Existence , Studia Philosophica, 1949-50. Colors as Universals, PR, 6 1 , 1952. 1) Existence , Independence , and Universals , PR, 69 , 1960. 2) The Tractatus: Nominalistic or Realistic? , £0, 148-65. 3) Bare Particulars , £0, 14-2 1 . 4) Another Look at Bare Particulars, U&P, 250-7 . 5) Truth , MP, 6 , 3-4 , 1975 . 6) Ontologically Speaking Things Are ... , T, 17 , 1976. 7) Relations: Recreational Remarks , PS, 34, 1978. 8) Bergrnann's Ontologies, O T, 38-64. 1) Particulars - Bare and Qualified, PPR, 15 , 1954 . 2) Ontological Commitment , PS, 9, 1958 . The Logic of Plurality (London: 1971). Wittgenstein on Universals, Ludwig Wittgenstein: The Man and his Philosophy, ed. K. T. Fann (New York : 1967) , 336--52 . 1 ) Some Open Problems Concerning the System E of Entailment , A PF, 1963 , 7-18. 2) Mind and Machines (Englewood Cliffs, N. J.: 1964) . 1 ) A Simple Treatment offruth Functions , JSL, 24 , 1959. 2) Tautological Entailments , PS, 13, 1962 . 3) The Pure Calculus of Entailment , JSL, 27 , 1962. 4) Entailment , v. 1 (Princeton: 1975). Studies in Empirical Philosophy (Sydney: 1962). 1) Studies on Gottlob Frege and Traditional Philosophy (Dordrecht : 1967). 2) Abstraction , Looking Around and Semantics, Die lntensionale Logik bei Leibniz und in der Gegen wart, Symposion der Leibniz-Gesellschaft Hanno ver (Wies­ baden : 1979) , 108-23 . 1 ) The Principle of Individuation , PASS, 27 , 1953 . 2) An Introduction to Wittgenstein 's Tractatus (London : 1959). 3) Substance , PASS, 38, 1964 . 4) Collected Papers, 3 v. (Minneapolis : 1981). Intentionality: A Study of Mental Acts (University Park, Penn . : 1977). 1) A Materialist Theory of the Mind (Londo n : 1969) . 2) Belief, Truth and Kno wledge (Cambridge : 1973) .

421

Aune , B Austin, J. L. Ayer, A. J.

Bacon, J . Baier, K. Baker, G. P . and Hacker , P. M. S . Bambrough, R . Bar-Hillel , Y. Barber , K.

Barker , S . F. Barnes , J . Barnes , W . H. F. Barnett, D. Batens , D. Bayliss , C. A.

Bell , D. Belnap , N. Belnap , N. and Dunn , M. Benacerraf , P. Benacerraf , P. and Putnam , H. eds. Bennett, J.

422

3) A Theory of Universals, 2 v. (Cambridge : 1978). 1 ) Knowledge, Mind, and Nature (New York : 1967) . 2) Quine on Translation and Reference , PS, 27 , 1975 . 1 ) Philosophical Papers (Oxford : 196 1 ) . 2) Sense and Sensibilia (Oxford : 1962) . 3 ) How to D o Things with Words (Cambridge : 1962) . 1) Universals and Particulars, PAS, 34 , 1933-34 . 2 ) On What There Is, PA SS, 25 , 195 1. 3) Philosophical Essays (New York : 1954). 4) The Problem of Knowledge (London: 1956) . 5) The Concept ofa Person (London: 1968). 6) Russell and Moore: The A nalytical Heritage (New York : 197 1 ) . 7) The Vienna Circle , PA P, 173-88. Ontological Commitment and Free Logic, Mo, 53 , 1970. Existence , PAS, 60 , 1960-61 . Wittgenstein - Understanding and Meaning: An Analytical Commentary on the Philosophical Investigations (Oxford : 1980) . Universals and Family Resemblances, PA S, 60, 1960--61. 1) Bolzano's Definition of Analytic Propositions , Me, 2, 1950. 2) Language and Information: Selected Essays on Their Theory and Application (Reading , Mass.: 1964). 1) Bare Particulars and Acquaintance , Dialogue, 5 , 1966. 2) A Note on a Paradox of Analysis, PS, 19, 1968. 3) On Representing Numerical Difference , Philosophical Topics, 10, 2. Philosophy of Mathematics (Englewood Cliffs , N. J . : 1969) . The Ontological A rgument (London: 1972) . 'Knowing' , PR, 72, 1963. A New Semantical Theory of Egocentric Particulars , S, 28, 1974. Outline of A Theory of Meaning : Semantical and Contextual , Philosophica, 19 , 1 , 1977. 1) Meanings and Their Exemplification , JP, 27 , 1930 . 2) Universals, Communicable Knowledge and Metaphysics , JP, 48 , 1953. 3) Logical Subj ects and Physical Obj ects, PPR, 17, 1957. Prege's Theory ofJudgment (Oxford : 1979) . See Anderson , A . J . and Belnap , N. The Substitution Interpretation of the Quantifiers, N, 2 , 1968. 1) What Numbers Could Not Be , PR, 74 , 1965 . 2) Frege : The Last Logicist , PA P, 1 7-36. Philosophy of Mathematics (Englewood Cliffs, N . J . : 1964) . 1) Substance, Reality and Primary Qualities , A PQ, 2, 1965 . 2) Real , M, 1966, 50 1- 1 5 . 3) Linguistic Behavior (Cambridge : 1976) .

Berg , J. Bergmann , G.

Bernays . P. Beth , E. W. Black , M.

Blanshard , B . Bochenski , I . M .

Boer, S. Bolzano , B . Borkowski , L . Bosanquet , B. Bradley, F. H.

Bolzano 's Logic (Uppsala: 1962). I ) A .t'ositivistic Metaphysics of Consciousness, M, 45, 1945. 2) Russell on Particulars , PR, 56 , 1947 . 3) Two Cornerstones of Empiricism, S, 8 , 1953. 4) Logical Positivism, Language , and the Reconstruction of Metaphysics , Rivista Critica di Storia della Filosofia, 8 , 1953. 5) Particularity and the New Nominalism, Me, 6 , 1954. 6) The Metaphysics of Logical Positivism (New York : 1954 ). 7) Intentionality , Semantica, Archivio de Filosofia ( Rome : 1955) . 8) Philosophy of Science (Madison: 1957). 9) Concepts, PS, 8, 1957 , with H. Hochberg. 10) The Revolt Against Logical Atomism , PQ, 7, 1957. 1 1) Meaning and Existence (Madison : 1959) . 12) Logic and Reality (Madison: 1964). 13) Realism: A Critique of Brentano and Meinong (Madison: 1967). 14) Diversity , Proceedings and Addresses of the American Philosophical Association, 5 1 , 1968 . 15) Sketch of an Ontological Inventory, Journal of the British Society for Phenomenology, 10, 1 , 1979. 16) Notes on Ontology , N . 15, 2 , 198 1. 17) Notes on the Ontology of Minds , FA P, 189-2 13. On Platonism in Mathematics , L 'Enseignement Mathematique, 34 , 1935. 1) The Foundations ofMathematics (Amsterdam: 1959). 2) Formal Methods (Dordrecht : 1962). 1) Russell's Philosophy of Language , PBR, 227-56. 2) The Nature ofMathematics (New York : 1933). 3) The Semantic Definition of Truth , A, 8, 1947-48. 4) Language and Philosophy (London: 1949). 5) The Identity of lndiscernibles. M, 6 1 , 1952 . 6) A Companion to Wittgenstein's Tractatus (Ithaca : 1964) . 7) Logic and Reality, Problems in the Philosophy of Language, ed . T. M. Olshewsky (New York : 1969) , 7 19-29. 8) The Elusiveness of Sets , RM, 24, 197 1. The Nature of Thought (London: 1939). 1) The Problem of Universals (Notre Dame: 1956), wi th A . Church and N . Goodman. 2) A History of Formal Logic, trans. I. Thomas (Notre Dame : 1961 ). 3) Logic and Ontology , Philosophy East and West, 24 , 3 , 1974. Reference and Identifying Descriptions , PR, 82, 1972. Wissenschaftslehre (new ed.) 4 v. (Leipzig: 1929) . Reduction of Arithmetic to Logic Based on the Theory of Types, Studia Logica, 8, 1958. The Essentials of Logic (London: 1948). 1) Appearance and Reality, 2nd. ed. (Oxford: 1897) . 2) The Principles ofLogic, 2 v . (Oxford : 1883) . 3) Essays on Truth and Reality (Oxford: 19 14).

423

Braithwaite, R. B. Brandt , R. B. Broad , C. D. Brodbeck , M.

Brody, B. A. Brownstein, D. Burge, T. Burkhardt , H. B utchvarov , P . K.

Butler , R. J. ed . Campbell, K. Carmichael , P . Carnap , R.

Cartwright, R.

Cassin , C. E.

424

4) Collected Essays, 2 v . (Oxford : 1935). Universals and the 'Method of Analysis' , PASS, 6, 1926 . The Languages of Realism and Nominalism, PPR, 1 7 , 1957 . 1 ) Examination of McTaggart's Philosophy (Cambridge: 1933) . 2) Leibniz: An Introduction (London: 1975) . 1 ) The New Rationalism: Dewey's Theory of Induction , JP, 46, 1 949. 2) Philosophy in America: 1900-1950 , American Non-Fiction, 1900-1950 , M. Brodbeck , et al. (Chicago : 1952) . 3) The Philosophy of John Dewey , EO, 1 88-2 1 5 . 4 ) Models, Meaning and Theories , Readings in the Philosophy of the Social Sciences, ed . M. Brodbeck (New York : 1968) , 579-600. Natural Kinds and Essences, JP, 64 , 1967. 1 ) Aspects ofthe Problem of Universals (Lawrence : 1 973) . 2) Negative Exemplification, A PQ, 1 0 , 1 , 1973 . 3) Denoting, Corresponding and Facts , T, 1 7 , 1976. Reference and Proper Names, JP, 70, 1973 . Logik und Semiotik in der Philosophie von Leibniz (Munich : 1980) . 1 ) Concrete Entities and Concrete Relations , RM, 1 0 , 1957 . 2) Resemblance and Identity (Bloomington: 1966). 3) The Relation of Resemblance, U&P, 1 84-99. 4) The Limits of Ontological Analysis , O T, 3-37. 5) Being Q ua Being (Bloomington: 1979). 1) The Logical Scaffolding of Russell's Theory of Descriptions, PR, 63 , 1954 . 2) Analytical Philosophy, v. 1 (Oxford : 1962) ; v. 2 (Oxford : 1965) . 1 ) Family Resemblance Predicates, APQ, 2, 1965 . 2) The Metaphysic of Abstract Particulars, FA P, 477-88 . 3) Body and Mind ( Garden City, N. Y . : 1970). Derivation of Universals , PPR, 8, 1948. 1) Der Logische Aufbau der Welt (Berlin: 1928) . 2) Logische Syntax der Sprache (Vienna : 1934) . 3) Foundations of Logic and Mathematics (Chicago : 1939) . 4) Meaning and Necessity (Chicago : 1947). 5) Empiricism, Semantics , and Ontology , Revue Internationale de Philosophie, 4, 1950 . 6) The Logicist Foundation of Mathematics, EBR, 34 1-54 . 1) Ontology and the Theory of Meaning, PSc, 4 , 1 954. 2) Comments on Dr . Hochberg's Paper : The Ontological Operator , PSc, 3 , 1956 . 3) Negative Existentials, JP, 56 , 1960. 4) Propositions , Analytical Philosophy, 1 , 8 1-103 . 1 ) Russell's Discussion of Meaning and Denotation, EBR, 256-72. 2) Russell's Distinction Between the Primary and the Secondary Occurrence of Definite Descriptions, EBR, 273-84.

Castaneda , H. N .

Castonguay , C . Caton , C . E. Causey , R. Cavell , S. Chandler, H. S. Chappell , V. C. Chihara , C .

Chisholm , R.

Chomsky , N. Chomsky, N . and Scheffler, I. Church , A.

Church, R. W. Churchman , C. W. and Ackoff, R . Chwistek, L .

1 ) Arithmetic and Reality, AlP , 37 , 2 , 1959. 2) On Math ematical Proofs and Meaning, M, 70 , 196 1 . 3 ) The Logic o f Knowledge and Belief , N . 1 , 1 , 1967. 4) Thinking and the Structure of the World, Philosophia, 4, 1974. 5) Individuation and Non-Identity : A New Look, APQ, 12, 2, 1975 . 6) Ontology and Grammar, T, 17 , 1976. 7) Intentionality, Minds and Perception (Detroit : 1967). Meaning and Existence in Mathematics (New York : 1972) . 1) An Apparent Difficulty in Frege's Ontology , PR, 7 1 , 1962 . 2) Strawson on Referring , EBR, 2 13--19. Unity of Science (Dordrecht: 1977) . Must We Mean What We Say? /, 1 , 1958. 1) Essence and Accident, A, 26, 1966 . 2) Defending Continuants, N, 4 , 1970. 1) Sameness and Change , PR, 69 , 1960. 2) Particulars Re-Clothed, PS, 15 , 1964. 1) Mathematical Discovery and Concept Formation , PR, 72, 1963. 2) Our Ontological Commitment to Universals , N, 2, 1968. 3) Russell's Theory of Types, Bertrand Russell, ed. D. Pears, (New York : 1972) . 4) Ontology and the Vicious Circle Principle (Ithaca: 1973) . 1 ) Perceiving: A Philosophical Study (Ithaca: 1957). 2) Realism and the Background of Phenomenology (New York : 1960). 3) Theory of Knowledge (Englewood Cliffs , N. J. : 1966) . 4) Person and Object (La Salle , Ill . : 1976). 1) Syntactic Structures (London : 1957). 2) Quine's Empirical Assumptions, W& O, 5 3-68. 3) Language and Mind ( New York : 1972) . What Is Said to Be , PAS, 59, 1958--59. 1) Propositions and Sentences , The Problem of Universals. 2) Review of Quine's 'Notes on Existence and Necessity' ,JSL, 8, 1943. 3) On Carnap's Analysis of Statements of Assertion and Belief , A , 1950 . 4) A Formulation of the Logic of Sense and Denotation , Structure, Method, and Meaning, ed. P . Henle , H. M. Kallen, and S. K. Langer (New York : 195 1). 5) Ontological Commitment , JP, 55, 1958. 6) The Need for Abstract Entities in Semantic Analysis, The Structure of Language, ed. J. Katz and J . Fodor (Engle­ wood Cliffs, N. J. : 1964). An Analysis of Resemblance ( London : 1952) . Methods of Inquiry (St . Louis: 1950) . Die Nominalistische Grundlegung der Mathematik , E, 3 , 1932-33 .

425

Clark, R. Clatterbaugh, K. Cocchiarella, N .

Cohen, L . J . Copi, I . Copi, I . and Gould, J . ed . Comman, J . W. Cresswell, M. J . Crittenden, C . Cummins, P . Curry, H. B.

Danto, A. Darrant, M. Dauer, F . Davidson, D .

Davidson, D . and Hintikka, J. ed. Dawes-Hicks, G. Demos, R. Dennet, D. Dewey, J.

Donagan, A.

Donnellan, K.

426

Concerning the Logic of Predicate Modifiers, N, 4, 1970 . General Ontology and the Principle of Acquaintance, PSc, 32, 1965 . 1) Whither Russell's Paradox of Predication, L & O, 133-58. 2) Properties as Individuals in Formal Ontology, N, 6, 1972. 3) Fregean Semantics for a Realist Ontology, NDJFL, 15, 1974. The Diversity of Meaning (London: 1962) . 1) Essence and Accident, JP, 5 1 , 1954. 2) The Theory of Logical Types (London: 197 1 ) . Contemporary Readings i n Logical Theory (New York: 1967). Language and Ontology, AJPP, 4 1, 1963 . Propositional Identity, Logique et A nalyse, 10, 1967. Thinking About Non-Being , I, 16, 1973 . Time for a Change, A , 26, 1965 . 1) Outlines of a Formalist Philosophy of Mathematics (Amsterdam: 195 1) . 2) Remarks on the Definition and Nature of Mathematics, D , 8, 1954. A nalytical Philosophy of Knowledge (Cambridge : 1968). Feature Universals and Sorta} Universals, A , 39, 1970. In Defense of the Coherence Theory of Truth , JP, 71, 1974. 1) Truth and Meaning, S, 17 , 1967 . 2) True to the Facts, JP, 66 , 2 1 , 1969. 3) Events as Particulars, N, 4, 1970. 4) The Individuation of Events, Essays in Honor of C. G. Hempel, ed . N. Rescher (Dordrecht : 1971), 2 16--32. 5) In Defense of Convention T, Truth , Syntax and Modality, ed . H. Leblanc (Amsterdam: 1973). 6) Belief and the Basis of Meaning , S, 27, 1974. 7) The Method of Truth in Metaphysics, CPPL, 294-304. Words and Objections: Essays on the Work of W. V. Quine (Dordrecht: 1969) . Are the Characteristics of Particular Things Universal or Particular?, PASS, 3, 1923. A Discussion of a Certain Type of Negative Proposition, M, 26, 1917. Brainstorms (Montgomery , Vt. : 1978) . 1) Essays in Experimental Logic (New York: 19 16) . 2) Experience and Nature (New York: 1929) . 3) Logic: The Theory oflnquiry (New York : 1938) . 4) Propositions, Warranted Assertibility, and Truth, JP, 38, 194 1 . 1 ) Recent Criticism s of Russell's Analysis o f Existenc e , A , 12, 195 1-52. 2) Universals and Metaphysical Realism , Mo, 47, 2, 1963. 3) Realism and Freethinking in Metaphysics, T, 42, 1976 . 1) Reference and Definite Descriptions, PR , 75, 1966. 2) Speaker Reference, Descriptions, and Anaphora, CPPL, 28----44.

Dretske , F. Dummett , M.

Ducasse , C . J . Duncan-Jones , A . E . Dunn , M . Eaton , R . M . Eberle . R . Echelbarger, C . Egidi . R . B .

Emmett, E . R . Essler, W . K . Evans , G . Ewing, A . C . Fan n , K . T . ed. Favrholdt , D . Feferman , S . Feibleman , J . Feigl , H .

Feigl , H . and Sellars , W . ed .

3) Speaking of Nothing, PR, 83 , 1974 . 1 ) Referring to Events , CPPL, 369-78. 2) Laws of Nature . PSc, 44 , 1977 , 248--68 . 1 ) Nominalism . PR , 65 . 1956, 49 1-505 . 2 ) Truth , PAS, 5 9 , 1 958---59. 3) Wittgenstein ·s Philosophy of Mathematics , PR, 68 , 1959. 4) The Phi losophical Significance of Godel's Theorem, R, 5, 1 963 . 5) Frege, Philosophy of Language (London : 1973) . 6) What Is a Theory of Meaning , Mind and Language, ed. S . Guttenplan , (Oxford : 1 975 ) . 7 ) What I s a Theory o f Meaning , I I . i n Truth and Meaning, ed. G . Evans and 1. McDowell (Oxford: 1 976) . 8) Truth and Other Enigmas (London : 1978) . Some Critical Comments on a Nominalistic Analysis of Resemblance , PR, 49 , 1 940 . Universals and Particulars, PAS, 34 , 1934. See Belnap , N. and Dunn , M . Symbolism and Truth (Cambridge : 1 925) . Semantic Analysis without Reference to Abstract Entities, Mo, 61 , 3. 1 977. Sellars on Thinking and the M yth of the Give n , PS, 25 , 1 974 . 1 ) La Consistenza Filosofica della Logica di Frege , Giornale Critico della Filosofia Italiana, 1 6 , 1 962. 2) Matematica , logica e filosofia nell ' opera di Gottlob Frege , Physics, 4 , 1 962 . Philosophy of Resemblances, P, 29 , 1 954. Analytische Philosophie I (Stuttgart : 1972) . 1 ) The Causal Theory of Names, Naming, Necessity, and Natural Kinds, ed. Schwartz, ( Ithaca: 1977) . 2) Reference and Contingency , Mo, 62 , 2 , 1 979. The Problem of Universals , PQ, 2 1 , 84 , 1 97 1 . 1 ) Wittgenstein 's Conception of Philosophy (Oxford : 1969) . 2) Wittgenstein: The Man and His Philosophy (Oxford : 1 969) .

An Interpretation and Critique of Wittgenstein 's ' Tractatus '

(Copenhagen : 1 964) . Systems of Predicative Analysis , JSL, 29 , 1 964 . Inside the Great Mirror (The Hague: 1958) . 1 ) Logical Empiricism , Twentieth Century Philosophy, ed. D . D . Runes , ( New York : 1943) . 2) Existential Hypotheses: Realistic Versus Phenomenalistic Interp retations, PSc, 17, 1950. 3) Logical Reconstructionism , Realism and Pure Semiotic , PSc, 1 7 , 1950. 4) The Mind-Body Problem in the Development of Logi cal Empiricism , Revue Internationale de Philosophie, 4, 1 950. 5) The Mental and the Physical (Min neapolis : 1967) . 6 ) The Wiener Kreis i n America , Perspectives in American History, I I (Cambridge : 1 968) , 630--73 .

Readings in Philosophical A nalysis (New York : 1949).

427

Field , H. Findlay, J . N . Fisk , M. Fitch, F . B.

Flew , A. G. N . ed . F0llesdal, D . Fraenkel , A. A . van Frassen, B.

Frege , G.

428

1 ) Tarski's Theory of Truth, JP, 69 , 1972 . 2) Theory Change and the Indeterminacy of Reference , JP, 70, 1973 . Meinong's Theory of Objects and Values (Oxford : 1963) . A Paradox in Frege's Semantics, PS, 1 4 , 1963 . 1) The Problem of the Morning Star and the Evening Star, PSc, 17, 1949. 2) Some Logical Aspects of Reference and Existence , JP, 57, 1960 . Essays in Conceptual Analysis (London : 1956) . 1) Quantification into Causal Contexts, R&M, 52--62 . 2) Quine on Modality , W& O, 175-85 . 3) Knowledge, Identity , and Existence , T, 33 , 1967. Sur la notion d'existence dans les mathematiques, L' Enseignement Mathematique, 1935-36. 1) Singular Terms, Truth-value Gaps, and Free Logic , JP, 63 , 1966 . 2) Presuppositions, Supervaluations and Free Logic, The Logical Way of Doing Things, ed. K . Lambert (New Haven : 1969) . 1) Begriffsschrift, eine der arithmetischen nachgebildete Formelsprache des reinen Denkens (Ha1le : 1879). 2) Die Grundlagen der A rithmetik (Breslau: 1 884 ) . 3 ) Funktion und Begriff(Jena : 189 1 ) . 4) Grundgesetze der A rithmetik, I (Jena: 1893). 5) Grundgesetze der A rithmetik, II (Jena : 1903). 6) Ober Sinn und Bedeutung , Zeitschrift fur Philosophie und philosophische Kritik, 100 , 1892 , trans. in ( 13). 7) Ober Begriff und Gegenstand, Vierteljahresschrift fur wissenschaftliche Philosophie, 16, 1892. 8) Was ist eine Funktion ? , Festschrift Ludwig Boltzmann (Leipzig : 1904) , 656--666 , trans . in ( 1 3). 9) Der Gedanke: Eine logische Untersuchung , Beitriige zur Philosophie des deutschen ldealismus, L 19 18. 10) Die Verneinung , Beitriige zur Philosophie des deutschen ldealismus, I , 19 19. 1 1 ) Logische Untersuchungen, ed . G. Patzig (Gottingen: 1966) . 12) The Foundations of A rithmetic, trans. J. L . Austin (Oxford: 1953). 13) Translations from the Philosophical Writings of Gottlob Frege, ed . P . Geach and M. Black (Oxford : 1960). 1 4) The Thought , trans. A. M. and M. Quinton , M, 65 , 1956 . 15) Compound Thoughts , trans . R. Stoothoff, M, 72 , 1963 . 16) The Basic Laws of Arithmetic, trans. in part , M. Furth (Berkeley: 1964) . 1 7 ) Conceptual Notation and Related Articles, trans. T . W . Bynum (Oxford : 1972) . 18) Logical Investigations, ed . P . Geach and R. H . Stoothoff (Oxford : 1 977). 19) Philosophical and Mathematical Correspondence, trans. H . Kaai (Oxford: 1979) .

Furlong. E . J . Furth , M. Gadamer , H . G. Gale , R. Garver, N.

Gasking, D. Geach , P. T.

Geach . P. T. and Black . M. ed. Goddard. L. Godel . K. Gombrich , E. Goodman , N.

Goodman , N. and Leonard , H . Goodman , N . and Quine , W. V . Gram , M. S.

Gram , M. S. and Klemke, E. D. ed . Grattan-Guiness , I .

20) Posthumous Writings, ed . H . Hermes , F . Kambartel, and F. Kaulbach. trans. P . Long and R. White (Oxford : 1979) . Abstract Ideas and Images , PASS, 27 , 1953. Editor's Introduction to Frege's Basic Laws of A rithmetic (Berkeley: 1964) , v-lvii. Wahrheit und Methode (Tiibingen: 1960). Negation and Non-Being (Oxford: 1976) . 1) The Range of Truth and Falsity, The Paradox of the Liar, ed. R. Martin , (New Haven : 1970) . 2) Wittgenstein ·s Pantheism : A New Light on the Ontology of the Tractatus, EW, 123-37. 1) Mathematics and the World, AIPP, 1 8 , 1940. 2) Clusters, AIPP, 38, 1960 . 1 ) Russell's Theory of Descriptions , A, 10, 1950. 2) Subj ect and Predicate, M, 59 , 1950. 3) On What There Is , PA SS, 195 1 . 4) Mental Acts (London: 1957) . 5) Russell on Meaning and Denoting , A , 19 , 1958-59. 6) Reference and Generality (Ithaca : 1962). 7) Intentional Identity , IP, 64 , 1967. Translations from the Philosophical Writings of Gottlob Frege (Oxford: 1960). Predicates, Relations , and Categories , AIPP, 44 , 1966 . Russell's Mathematical Logic, PBR , 123-54. 1) A rt and Jllusion(London : 1962) . 2) Art, Perception, and Reality (Baltimore : 1972) , with M. Black and J. Hochberg . 1) The Structure ofAppearance (New York : 1966). 2) Problems and Projects (New York : 1972). 3) Fact, Fiction, and Forecast, 3rd . ed. , (Indianapolis: 1973). 4) A World of lndividuals , The Problem of Universals , with I. Bochenski and A. Church . 5) Languages ofA rt (Indianapolis: 1968) . 6) Words , Works , Worlds, £, 9, 197 5 . 7 ) Predicates Without Properties, CPPL . 8) On Likeness of Meaning , S&PL , 67-74 . 9) The Significance of Der logische Aufbau der Welt , PRC, 545-58. The Calculus of Individuals and Its Uses , ISL , 1940. Steps Towards a Constructive Nominalism , JSL , 12, 1947 . 1) Ontology and the Theory of Descriptions, EBR , 1 18-43 . 2) The Reality of Relations , TNS, 44 , 1 97 1 . 3 ) Relations Again , TNS, 45 , 197 1 . The Ontological Turn (Iowa City: 1974). Bertrand Russell on His Paradox and the Mu ltiplicative Axiom , IPL , 1972 .

429

Gregory , J. C. Grice , H. P . Grice , H. P . and Strawson , P. F . Griffin, J. Grimm, R. H . Grossmann, R.

Gull , R. Gunderson, K. Haack , S .

Hacker, P . M. S . Hacking, I . Hahn , H. Hall , E. W. Hallden, S.

Hallet , G . Harman , G . Hartshorne, C. Hausmann , A.

Heidelberger, H. van Heij enhoort , J. ed . Heller, E. Hempel , C . G.

430

Leibniz, the Identity of lndiscernibles, and Probability , PPR , 14, 1954 . Meani ng , PR, 66 , 1957 . In Defense of a Dogma , PR, 65, 1956. Wittgenstein 's Logical A tomism (Oxford: 1964) . Names and Predicables, A , 26 , 1965-06 . 1) The Structure ofMind (Madison : 1959) . 2) Conceptualism, RM, 14, 1960 . 3) Frege's Ontology, PR , 70 , 196 1 . 4) Sensory Intuition and the Dogma of Localization , /, 5, 1962. 5) Reflections on Frege's Philosophy(Evanston : 1969) . 6) Russell's Paradox and Complex Properties, N, 1972 . 7) Ontological Reduction (Bloomington : 1973). 8) Meinong (London : 1974) . 9) Definite Descriptions , PS, 27 , 1975 . Bradley's Argument Against Relations, TNS , 45 , 197 1 . Mentality and Machines (New York : 197 1 ) . 1) Deviant Logic (Cambridge : 1974) . 2) Analyticity and Logical Truth , T, 43, 1977. 3) Platonism Versus Nominalism: Carnap and Goodman , Mo , 6 1 , 3 , 1978. Insight and Illusion (Oxford: 1972) . 1) A Language Without Particulars, M, 77 , 1968 . 2) What is Logic , JP, 76, 1979 . Logik, Mathematik und Naturerkennen (Vienna : 1933 ) . Logical Subj ects and Physical Obj ects, PPR , 17, 1957. 1) Certain Problems Connected with the Definition of Identity and of Definite Descriptions Given in Principia Mathe­ matica , A , 9, 1948-49 . 2) The Logic ofNonsense (Uppsala: 1949) . A Companion to Wittgenstein 's Philosoph ical Investigations (Ithaca : 1977) . 1) Quine on Meaning and Existence , II , RM. 2 1 , 1967. 2) Deep Structure as Logical Form, S, 2 1 , 1970 . 3) Thought (Princeton : 1973) . Are There Absolutely Specific Universals. JP, 68, 3 . 197 1 . 1 ) Goodman's Ontology , Carnap and Goodman : Two Formalists (The Hague: 1967 ) . with F. Wilson. 2) Hume's Theory of Relations, N, 1 , 3. 1967 3) Bergmann on the Analytic-Synthetic Distinction , O T, 230-44 . The Indispensability of Truth , APQ, 5, 1968. 1) Logic as Calculus and Logic as Language . S, 17, 1967. 2) From Frege to Godel (Cambridge : 1967) . Ludwig Wittgenstei n, Encounter, 1958. 1) On the Logical Positivist 's Theory of Truth , A . 2 , 1935. 2) On the Nature of Mathematical Truth, The American Mathematical Monthly, 52 , 1945. 3) Reflections on Goodman's 'The Structure of Appearance' , PR , 1953 .

Henkin , L. Heyting, A. Hintikka . J .

ed . Hiz , H . Hochberg, H .

4 ) O n Russell's Phenomenological Constructionism , JP, 58, 1 966. Are Logic and Mathematics Identical ? , Science, 138, 1962 . Intuitionism (Amsterdam : 1 966) . 1 ) Existential Presuppositions and Existential Commitme nts , JP, 56, 1 958. 2) Modality and Quantification , T, 27 , 1 961 . 3) Knowledge and Belief(lthaca : 1 962) . 4) Cogito Ergo Sum: Inference or Performance ? , PR, 7 1 , 1 962 . 5) Individuals, Possible Worlds , and Epistemic Logic, N, 1 , 1967 . 6) Words and Objections, see Davidson , D . 7) A re Mathematical Truths Synthetic A Priori ? , JP, 65 , 1 968. 1) On Some General Principles of Semantics of Natural Languages , Philosophica, 1 8 , 2, 1976. 2) On Assertions of Existence , L&O, 175-91 . 1 ) 'Possible' and Logical Absolutism , PS, 6, 1 955 . 2) The Ontological Operator, PSc, 23 , 1 956. 3) Pea110 , Russell and Logicism , A, 16, 1 956. 4) Descriptions, Scope , and ldentity , A , 18, 1 957 . 5) A Note on the Empty Universe , M, 1 957. 6) Concepts, PS, l, 1957, with G . Bergmann . 7) Phenomena , Value , and Obj ectivity , PQ, 8 , 32, 1 958. 8) St. Anselm's Ontological Argument and Russell's Theory of Descriptions , TNS, 1 959 . 9) Intervening Vari ables , Hypothetical Constructs , and Metaphysics , Current Issues in the Philosophy of Science, ed. H. Feigl and G. Maxwell (New York : 196 1 ) , 448-56, 459-60. 10) Moore's Ontology and Non-Natural Properties , RM, 1 5 , 1 962 , revised version i n SPGM. 1 1 ) On Being and Being Presented , PSc, 32, 1 965 . 12) Things and Qualities , Metaphysics and Explanation, ed . D . Merril and K. Capitan , (Pittsburgh : 1 966) , 82-97 . 1 3) Rej oinder to Severens and Shorter, M&E, 1 06--12. 14) Dispositional Properties, PSc, 1 967 , 1 84-204. 15) Moore and Russell on Particulars , Relations and Identity , SPGM, 1 55-94 . 16) Some Reflections on Mr. Nelson's Corrections, SPGM, 14 1-54 . 1 7) On Referri ng and Asserting, PS, 20 , 1 969 . 18) Metaphysical Explanation , MP, 1970. 19) Material Properties in the Tractatus, EW, 120--23 . 20) Possibilities and Essences i n Wittgenstein's Tractatus, EW, 485-533. 21) Intentions , Facts, and Propositions, O T, 1 68-92. 22) Thought, Fact and Reference: The Origins and Ontology of Logical Atomism ( Minneapolis: 1978) . 23) Intentionality , Logical Structure , and Bergmann's Ontol­ ogy , N, 1 5 , 2, 1 98 1 .

43 1

Hopkins, J . Hughes , G . E . and Cresswell , M. Hunter, G. Iseminger, G. ed . lshiguro, H. Jackson , F. C. Jackson , H. Jaeger, R. James , W. Johnson, W. E. Jones , E. E . C . Jones , J. R .

Joseph, H . W. Jubien , M. Kahn, C. H . Kanger , S. Kaplan , D. Katz, J . J . Kauf, D . Kearns , J. T . Kemp Smith, N . Kenny, A. Khatchadourian , H. Kielkopf, C . Klemke , E. D .

ed.

432

St. Anselm 's Ontological Argument (Minneapolis: 1976) . An Introduction to Modal Logic ( London: 1968) . Metalogic (New York: 197 1 ) . Logic and Philosophy (New York: 1968) . Leibniz and the Ideas of Sensible Qualities, Reason and Reality, Royal Institute of Philosophy Lectures, V (London : 1972) , 4963. Statements about Universals , M, 76 , 1977. 1 ) Frege's Ontology, PR, 69 , 1960. 2) Frege on Sense-Functions , A, 23 , 1962-63. 1) Russell's Denoting Complex , A, 20, 1959-60, 53-62 . 2) The Development of Bertrand Russell's Philosophy (London: 1972) . l) The Principles of Psychology (London : 1899) . 2) Pragmatism (London: 1907) . 3) The Meaning of Truth (London: 1909). l) Logic, Part I (Cambridge, 1921). 2) Logic, Part II (Cambridge , 1922). Mr. Russell's Objection to Frege's Analysis of Propositions , M, 19, 1910. 1) Are the Qualities of Particular Things Universal or Particular?, PR, 58, 1949. 2) What Do We Mean By an 'Instance' ? , A , 1 1 , 1950. 3) Characters and Resemblance , PR, 60 , 195 1 . Universals and the 'Method of Analysis' , PASS, 1926. Two Kinds of Reduction , JP, 66 , 1969. On the Theory of the Verb 'To Be' , L & O. 1-20 . 1) The Morning Star Paradox , T, 23 , 1957. 2) A Note on Quantification and Modalities . T, 23 , 1957 . 3) Provability in Logic (Stockholm: 1957) . 1) Quantifying In, W& O, 206---242. 2) DThat , CPPL, 383--400. 1) The Philosophy of Language (New York: 1966) . 2) The Neo-Classical Theory of Reference , CPPL, 103-24. A Comment on Hochberg's Reply to Storer , PS, 5, 4, 1954 . 1) Sameness or Similarity? . PPR, 29 , 1968. 2) Substance and Time , JP, 65. 1970 . The Nature of Universals, M, 36, 1927 . Oratio Obliqua , PASS, 37 , 1963 . 1 ) Common Names and 'Family Resemblances' , PPR, 18, 1958. 2) Natural Obj ects and Common Names , Me, 13 , 196 1. An Investigation of Wittgenstein 's 'Remarks on the Foundations of Mathematics ', (The Hague : 197 1 ) . 1 ) Universals and Particulars i n a Phenomenalist Ontology , PSc, 27 , 1960. 2) The Epistemology of G. E. Moore (Evanston : 1969) . 3) Logic and Ontology, EBR, 4 16--444. 4) The Ontology of Wittgenstein's Tractatus, EW, 1 04-19. 5) Essays on Frege (Urbana: 1968) .

ed . ed. ed. Kneale , W. C .

Kneale , W. C . and Kneale, M. Korner . S .

Kripke , S .

ed .

Kung , G. Lacey, H . Lackey, D. Laird , J .

ed .

Lambert, K. Lazerowitz, M. Leblanc, H . and . Hailperin, T. Leech , G . Lej ewsky, C . Leonard , H.

Lewis , C. I . Lewis, D. Lewy, C.

6) Studies in the Philosophy of G. E. Moore (Chicago : 1 969). 7) Essays on Bertrand Russell (Urbana : 1970) . 8 ) Essays o n Wittgenstein (Urbana : 197 1). 1) The Objects of Acquaintance , PA S, 34, 1933-34. 2) Is Existence a Predicate? , PA SS, 15, 1936. 3) The Notion of a Substance , PAS, 40 , 1939-40. 4) Truths of Logic, PA S, 46 , 1946. 5) Universality and Necessity, BJ PSc, 12, 1962. 6) Russell's Paradox and Some Others , BJPSc, 22, 197 1. The Development of Logic (Oxford : 1962) . 1) OnDeterminables and Resemblances , PASS, 33, 1959. 2) The Philosophy ofMathematics (London : 1960) . 3) Experience and Theory (London : 1966). 4) Philosophy ofLogic (Oxford: 1976). 1) Semantical Considerations on Modal Logic . A PF, 16, 1963 . 2) Is There a Problem about Substitutional Quantification? , Truth and Meaning, (Oxford : 1 976) , 325-419. 3) Naming and Necessity (Cambri dge : 1980). 4) Identity and Necessity , Identity and Individuation, ed. M. Munitz, (New York : 1973) , 135---64. 5) Speaker Reference and Semantic Reference, CPPL, fr-27. 1) Concrete and Abstract Properties , NDJFL, 5 , 1964 . 2) Ontology and the Logistic Analysis of Language (Dordrecht: 1967) . Quine o n the Logic and Ontology o f Time, A lP , 49, 197 1 . 1) Russell's 1913 Map of the Mind, FA P, 125-42 . 2) Essays in Analysis: Bertrand Russell (New York : 1973). On Certain of Russell 's Views Concerning the Human Mind. PBR, 293-3 16. l ) The Theory of Obj ects , /, 16, 1973 . 2) Impossible Objects, /, 17 , 1974. 1) The Existence of Universals , M, 55 , 1946. 2) Substratum, Philosophical Analysis, ed. M. Black (Englewood Cliffs , N . J . : 1963). Nondesignating Singular Terms , PR, 1959. Semantics (Harmondsworth : 1974). A Re-examination of Russell's Theory of Descriptions , P, 35 . 1960. 1 ) The Logic of Existence , PS, 1956, 49-64 . 2) Essences , Attributes and Predicates , Proceedings and A ddresses of the American Philosophical Association, 37 , 1964, 25-5 1 . 1 ) Mind and the World Order (New York : 1929). 2) The Modes of Meaning , PPR, 4, 1 944. 1) Counterpart Theory and Quantified Modal Logic, JP, 65 , 1968. 2) Counter/actuals (Oxford : 1973) . 1 ) How A r e t h e Calculuses o f Logic and Mathematics Applicable to Reality? , PASS, 20, 1946.

433

Lindstrom, P .

Linsky, L.

ed. Lloyd, A . C. Loar, B . Long, D. C . Lorenzen, P . Loux , M. J .

ed. Lowe , V. Lucas, J . R. Lukasiewicz, J.

Lycan, W. Lycan, W. and Pappas,G. MacColl, H .

MacDonald, M. ed. Mackie, J. L. Malcolm , N. Malmgren, H. Marcus, R.

Margolis , J .

434

2) Entailment, PASS, 32 , 1958 . 1) On Model-Completeness, T, 30, 1964. 2) On Relations B etween Structures, T, 32, 1966. 3) First Order Predicate Logic with Generalized Quantifiers, T, 32, 1966. 1) Substitutivity and Descriptions, JP, 58, 1966 . 2) Referring (New York : 1967). 3) Reference and Referents, Problems in the Philosophy of Language, ed. T. M. Olshewsky (New York : 1969), 338---50. 4) Names and Descriptions (Chicago: 1977). 5) Semantics and the Philosophy ofLanguage (Urbana : 1952) . On Arguments for Real Universals, A , 1 1, 195 1 . Two Theories o f Meaning , Truth and Meaning (Oxford: 1976) Particulars and Their Qualities, PQ, 18, 1968. Metamathematik (Mannheim : 1962) . 1) Recent Work in Ontology, A PQ, 9, 2, 1972 . 2) Kinds and the Dilemma of lndividuation, RM, 27, 1974. 3) The Concept of a Kind, PS, 29 , 1976. 4) The Problem of Universals, U&P, 3--15 . 5) Particulars and Their Individuation , U& P, 203 . 6) Substance and A ttribute (Dordrecht: 1978). 7) Universals and Particulars: Readings in Ontology (New York: 1970) . The Concept of the Individual, Me, 5 , 1953. Minds, Machines and Godel , P, 36, 196 1. 1) Die Logik und das Grundlagenproblem, Les entretiens de Zurich sur Les fondements et la metho de des sciences mathematiques, 1 938. 2) The Principle of Individuation, PASS, 27, 1953 . Occam's Razor, MP, 6 , 3-4, 1975. What Is Eliminative Materialism, AJP, 50, 1972. 1) Symbolic Reasoning, M, 5, 1880 . 2) Symbolic Reasoning, M, 12. 1903 . 3) Symbolic Reasoning, M, 15 , 1906 . 4) Symbolic Logic and Its Applications ( London : 1906) . 1) The Philosopher's Use of Analogy, PAS, 38, 1937-38. 2) Things and Processes , A, 6, 1938. 3) Philosophy and Analysis (Oxford : 1954). Truth, Probability and Paradox (Oxford: 1973) . 1) Defending Common Sense, PR, 58, 1949. 2) Anselm's Ontological Argument, PR, 69 , 1960 . 1) Intentionality and Knowledge: Studies in the Philosophy of Moore and Wittgenstein (Gothenburg : 1971) . 2) Immediate Knowledge of Other Minds, T, 42, 1976. 1) Extensionality, M, 69, 1960 . 2) Modalities and Intensional Languages, S, 27 , 1962. 3) Essentialism in Modal Logic, N, 1, 1967. 4) Quantification and Ontology , N, 6, 1972 . 1) Some Ontological Policies, Mo, 53, 1970 .

Martin, R . M.

Martinich , A. P . Marshall, W. Maslow. A . Mates .B. Matheson, G. McCall . S. McCloskey, H. J. McDowell, J . McMullin, E. McTaggart , J . M.E . Meiland, J . W. Meinong , A.

Mellor, D. H. Mendelson , R . L. Montague , R. Moore, G. E.

Moravscik , J .

2 ) The Problems of Similarity: Realism and Nominalism, Mo, 6 1, 3 , 1978. 3) Robust Relativism, Journal ofAesthetics and Art Criticism, 30, 1976. 1) On the Berkeley-Russell Theory of Proper Names , PPR, 13, 1952-53 . 2 ) Truth and Denotation (Chicago: 1958) . 3) On Tense, Aspect , Modality and Meaning , Philosophica, 19, 1, 1977 . Referring , PPR, 40 . 2 , 1979. Frege's Theory of Functions and Obj ects , PR, 62, 1953. A Study in Wittgenstein's ' Tractatus · (Berkeley: 196 1). 1 ) Synonymity , S&PL, 1 1-36. 2) Leibniz on Possible Worlds , Leibniz, ed . H. Frankfurt (New York : 1972) . The Semantics of Singular Terms , JSL, 27 . 1962. A Non-Classical Theory of Truth, with an Application to Intuitionism, A PQ, 7, 1970. The Philosophy of Linguistic Analysis and the Problem of Universals . PPR, 24 , 1964. On the Sense and Reference of a Proper Name , M, 86 , 1977 . 1) The Problem of Universals , PS, 8, 1958. 2) The Concept of Matter (Notre Dame: 1963). 1) The Unreality of Time, M, 17, 1908. 2) The Nature ofExistence, 2 v . (Cambridge : 1920). 3) Philosophical Studies (London: 1934) . 1 ) D o Relations Individuate? U& P, 258--63 . 2) Talking About Particulars (London: 1970). 1) Untersuchungen zur Gegenstandstheorie und Psycho/ogie (Leipzig : 1904) . 2) Ober die Ste/lung der Gegenstandstheorie im System der Wissenschaften (Leipzig: 1907 ) . 3) O n Emotional Presentation, trans. M.L. Schubert-Kalsi ( Evanston: 1972) In Defense of Dispositions , PR, 82 , 1974. Frege on Predication ,FA P, 69-82. Formal Philosophy (New Haven: 1974). 1 ) The Nature of Judgment , M, 30, 1899. 2) ldentity , PA S, 1 , 190�1 . 3) The Refutation of Idealism , M, 12 , 1903 . 4) Is There Knowledge by Acquaintance? , PASS, 2, 1919. 5) Are the Characteristics of Particular Things Universal or Particular?, PASS, 1923 . 6) Russell's Theory of Descriptions , PBR, 175-226. 7) Philosophical Studies (New York : 195 1). 8) Some Main Problems of Philosophy (London : 1953) . 9 ) Philosophical Papers (London: 1959) . IO) Commonplacebook, ed. C. Lewy (London: 1962) . 1 ) Strawson on Ontological Priority , Analytical Philosophy, 2, 106- 19. 2) Frege and Chomsky on Thought and Language, FA P, 105-24.

435

Moreno, J . Morris, C .

Munitz, M. K . ed. ed. Myhill,J .

Nagel , E. Nakhnikian,G . and Salmon , W. Nelson, E. J .

Nelson , J . M . Nerlich, G . C . O'Connor, D.J .

Odegard, D. Oldenquist , A. Olshewsky , T. M. , ed. Owens, J .

Pap , A.

Parsons, C . Parsons, K .

436

Eaton on the Problem of Negation , Transactions of the C. S. Peirce Society, 16, 1980 . 1) Foundations ofthe Theory ofSigns (Chicago : 1938) . 2) Pragmatism and Logical Empiricism , PRC, 87-98. 3) Signs, Language and Behavior (Englewood Cliffs, N .J . : 1946) . 1 ) Existence and Logic (New York : 1974) . 2) Identity and Individuation (New York : 197 1) . 3 ) Logic and Ontology (New York : 1973) . 1) Report of Some Investigations Concerning the Consistency of the Axiom of Reducibility , ]SL, 16, 195 1 . 2 ) On the Ontological Significance o f the Lowenheim-Skolem Theorem, Contemporary Readings in Logical Theory, 40-5 1 . 1 ) Mr. Russell o n Meaning and Truth , JP, 38, 194 1 . 2) Logic Without Metaphysics (Glencoe , Ill. : 1956) . 3) Logic Without Ontology in (2) , 55-92 . 'Exists' as a Predicate , PR, 1957. 1) Intensional Relations, M, 39 , 1930 . 2) On Three Logical Principles in Intension, Mo, 43 , 1933 . 3) Contradiction and the Presupposition of Existence , M, 55 , 1946. Mr. Hochberg on Moore: Some Corre�tions, SPGM, 28---40. 1) On Evidence for Identity, Al PP, 38 , 1959. 2) Presupposition and Entailment, A PQ, 2 , 1965 . 1) On Resemblance, PAS, 46, 1945-46. 2) Stout's Theory of Universals , AJPP, 27 . 1949 . 3) Names and Universals, PAS, 53 , 1952-53. 4) The Identity of lndiscernibles, A , 1 6 , 1954 . 5) The Correspondence Theory of Truth (London: 1975) . Indiscernibles, PQ, 1964 . Wittgenstein on Phenomenalism . Skepticism, and Criteria . EW, 394-422. Problems in the Philosophy of Language (New York : 1969). 1) Unity and Essence in St . Thomas Aquinas, Medieval Studies, 23 , 196 1 . 2 ) A n Elementary Christian Metaphysics (Milwaukee : 1963) . 3) The Content of Existence, L& O, 21-35 . 1 ) Logic , Existence and the Theory of Descriptions. A , 13, 1952-53 . 2) Analytische Erkenntnistheorie (Vienna: 1955) . 3) Nominalism , Empiricism and Universals: I, PQ, 9, 1959. 4) Nominalism, Empiricism and Universals: II , PQ, 10, 1960 . 5) Mathematics. Abstract Entities, and Modern Semantics , Problems in the Philosophy of Language, (New York : 1 969) , 699-719. 1) Frege's Theory of Numbers, Philosophy in A merica, ed . M . Black , (New York : 1965) . 2) Ontology and Mathematics, PR , 80 , 197 1 . A Criterion for Meaning Change, PS, 28 , 1975 .

Parsons, T. Passmore , J . Peacocke , C. Pears, D. F.

Peirce, C . S. Peters , F. Peterson . J. Phillips, E . D. Pitcher, G. ed. Place, U. T. Plantinga, A. Pompa, L. Popper , K. R. Price, H . H. Prior , A. N .

Putnam , H.

1 ) Essentialism and Quantified Modal Logic, PR, 77, 1969. 2) A Prolegomena to Meinongian Semantics , JP, 7 1 , 1974 . 3) Nonexistent Objects (New Haven : 1980 ) . 1) Philosophical Reasoning (New York : 1961). 2) A Hundred Years of Philosophy, 2nd. ed. (New York: 1967) . Truth Definitions and Actual Languages, Truth and Meaning. 1) U niversals, PQ, 1 , 1951. 2) Logical Atomism: Russell and Wittgenstein, The Revolution in Philosophy, A. J. Ayer , et. al. (London: 1956) , 4 1-55 . 3) The Nature of Metaphysics, (New York : 1957). 4) A Critical Study of P. F. Strawson's Individuals, I and I I , PQ, 1 1, 196 1 , 172-85 . 5) Bertrand Russell and the British Tradition in Philosophy, (London: 1967) . 6) Ludwig Wittgenstein (New York : 1969). 7) Questions in the Philosophy of Mind (New York : 1975). 8) The Relation between Wittgenstein's Picture Theory of Propositions and Russell's Theories of Judgment , PR, 1977 . Collected Papers, 8 vols. (Cambridge : 193 1-58) . Russell on Class Theory, S, 15 , 1963. Realism and Logical A tomism (Alabama: 1976) . On Instances, A , 1, 1934, 60--1 . 1 ) The Philosophy of Wittgenstein ( Englewood Cliffs, N .J . : 1964) . 2 ) Truth ( Englewood Cliffs, N. J. : 1964). I s Consciousness a Brain Process? , British Journal of Psychology, 47 , 1956. The Nature of Necessity (Oxford : 1974). Family Resemblances, PQ, 17 , 1967, 63-9. 1) The Principle of Individuation , PASS, 27 , 1953. 2) Objective Knowledge (Oxford : 1973) . l ) Thinking and Representation (London: 1946) . 2) Thinking and Experience ( London: 1953). 1) Determi nables, Determinates, and Determinants , I and I I , M , 53 , 1949. 2) Time and Modality (Oxford : 1957). 3) Nonentities, Analytical Philosophy , 1 , 120-- 132. 4) Oratio Obliqua , PA SS, 37 , 1963 . 5) Past, Present and Future (London : 1967) . 1) Mathematics and the Existence of Abstract Entities, PS, 1956 . 2) Psychological Concepts, Explication, and Ordinary Language , JP, 4 , 1957 . 3) The Thesis That Mathematics is Logic, Bertrand Russell: Philosopher of the Century, ed. R. Schoenman (Lon­ don : 1967), 273-303 . 4) Philosophy of Logic (New York: 197 1 ) . 5) O n Properties . Philosophical Papers, vol. l , (Cambridge : 1975).

437

Quine , W. V.

Quine ,W. V. and Ullian , J . Quinton, A . Radnitzky , G . Ramsey , F . P.

Randall, J. H. Raphael , D. D. Reeves, J. W. Reichenbach , H. Reid , T. Rescher , N.

Resnik , M. Richman, R. Robbins, B. Rogers, R. Rorty , R. ed. Rose , T. A. Rosenberg, J . F.

438

1) 2) 3) 4) 5) 6) 7) 8) 9) 10) 11) 12) 13)

Designation and Existence, JP, 36 , 1939. On Universals, JSL, 1 2 , 1947. Two Dogmas of Empiricism, PR, 60 , 195 1 . From a Logical Point of View (Cambridge: 1953) . Word and Object (Cambridge : 1960). Set Theory and Its Logic (Cambridge : 1963 ) . The Ways of Paradox (New York : 1966). Selected Logic Papers (New York : 1966). Ontological Relativity and Other Essays (New York : 1969). Philosophy of Logic (Englewood Cliffs , N . J. : 1970). Russell's Ontological Development, EBR, 3-14. Russell's Theory of Types, EBR, 372-387. Steps Towards a Constructive Nominalism , see Goodman , N . 14) The Roots of Reference (La Salle, Ill.: 1973). 15) Intensions Revisited, CPPL, 268-74.

The Web ofBelief (New York : 1970). 1 ) Properties and Classes, PASS, 58, 1957-8. 2) The Nature of Things (London: 1973). Contemporary Schools of Metascience (Gothenburg : 1968) . 1 ) The Foundations of Mathematics, Proceedings of the London Mathematical Society, 2, 25 , 1925. 2) Mathematical Logic, Mathematical Gazette, 13 , 1926. 3) Universals and the 'Method of Analysis' , PASS, 6, 1926. 4) Facts and Propositions , PA SS, 7, 1927. 5 ) The Foundations ofMathematics (New York : 193 1 ). 6) Universals, in (5) . Substances as Processes, RM, 10, 1957. 1 ) Universals, Resemblance , and Identity , PAS. 55 , 1954-5. 2) To Be and Not to Be , PAS, 60 , 1960--1 . The Origin and Consequences o f the Theory o f Descriptions . PAS, 34 , 1933-34. Bertrand Russell's Logic, PBR, 21-54. Essays on the Intellectual Powers of Man, ed. A. Woozley (London: 1941). 1) Definitions of Existence , PS, 1957. 2) The Logic of Existence . Topics in Philosophical Logic (Dordrecht: 1 968). 3) Evaluative Metaphysics , M& E, 62-72 . 4) The Coherence Theory of Truth (Oxford: 1973). 5) A Theory of Possibility (Oxford : 1975) . 6 ) Methodological Pragmatism (Oxford : 1977). Frege ·s Theory of Incomplete Entities, PSc, 32 , 1965. Somt;thing Common . JP, 1962. Ontology and the Hierarchy of Languages , PR, 67 , 1958. A Survey of Formal Semantics, S, 15, 1963. I ) Philosophy and the Mirror of Nature (Princeton: 1979) . 2) The Linguistic Turn (Chicago : 1967) . The Nominalist Error , AJPP, 27 . 1949 . 1) The 'Given· and How to Take It: Some Reflections on

Rosenthal , D . ed . Routley , R. Rudner . R. Rundle. B. Russell , B.

Ryle . G.

Savery, B. Savigny , E. Saw, R. L. Schaffner , K. Scheffler , I .

Phenomenal Ontology , MP, 6 , 3--4, 1975. 2) Linguistic Representations (Dordrecht: 1977) . Materialism and the Mind-Body Problem (Englewood Cliffs, N.J.: 1971) . 1 ) Some Things That Do Not Exist , NDJFL, 7 , 1963. 2) Another ' Fatal' Objection to Meinongian Obj ects , PS, 24, 1973 . On Sinn as a Combination of Physical Properties, M, 6 1 , 1952. Modality and Quantification, Analytical Philosophy, 2, 27-39. 1) A Critical Exposition of the Philosophy of Leibniz (Cambridge: 1900). 2) The Principles ofMathematics (Cambridge: 1903) . 3) On Denoting, M, 14, 1905 ; L& K, 41-56. 4) Review of Meinong's Untersuchungen zur Gegen­ standstheorie und Psychologie, M, 1 4 , 1905. 5) On the Nature of Truth, PAS, 7, 1 906--7. 6) On the Relations of Universals and Particulars, PAS, 12 , 191 1 -12. 7) The Problems of Philosophy (London: 1912). 8) Principia Mathematica, 3 vols. (Cambridge: 1910-1913), with A. N . Whitehead. 9) The Philosophical Importance of Mathematical Logic, Mo, 22 , 4, 1913. 1 0) Theory of Knowledge , manuscript ( 1913) in the Russell Archives. 1 1 ) The Philosophy of Logical Atomism, Mo, 28, 29 , 1918, 1919; in ( 1 8) , 177-281 . 12) On Propositions: What They Are and How They Mean , PASS, 2 , 1919; in ( 1 8) , 285-320 . 1 3) Introduction to Mathematical Philosophy (London: 1919) . 1 4) The A nalysis ofMind (London: 192 1 ) . 1 5) The A nalysis of Matter (London: 1927). 1 6) An Inquiry into Meaning and Truth (London: 1940) . 17) Human Knowledge: Its Scope and Limits (London: 1948). 18) Logic and Kno wledge, ed . R. C. Marsh (London: 1956). 19) Mr . Strawson on Referring, My Philosophical De­ velopment (London: 1959) , 23�245. 1 ) Systematically Misleading Expressions, PA S, 32, 193 1-32, 139-70 . 2) Plato's Parmenides, M, 48, 1939 . 3) The Concept ofMind (London : 1949) . 4) Heterologicality , A , 1 1 , 1952. 5) Dilemmas (Cambridge : 1954). 6) Collected Papers (London: 197 1 ). Identity and Difference , PR, 5 1 , 1942 , 205-12. A nalytische Philosophie ( Miinchen: 1970). Our Knowledge of lndividuals, PAS, 52 , 195 1-52 , 167-88. Approaches to Reduction, PSc, 34 , 1967, 137--47. 1) An Inscriptionalist Approach to Indirect Quotation, A, 1 4 , 1967. 2) Science and Subjectivity (New York: 1967) .

439

Scheffler, I . and Chomsky , N . Schiffer, S . Schiller, F. C. S . Schlick , M .

Schock , R . Scott,D. Searle, J . R.

Segelberg, I. Segersberg , K. Sellars, W.

Severens , R. Shimony, A. Shoemaker, S. Shorter. J. M. Shwayder, D. Singer . E. A. Sinisi , V. F. Skyrms. B.

440

What Is Said to Be, see Chomsky, N . Meaning (Oxford : 1972) . Logicfor Use (New York : 1930) . 1) Positivismus und Realismus, E, 3 , 1933 . 2) De la relation des notions psychologiques et les notions phy�ques, S, 10, 1935 . 3) Meaning and Verification , PR, 44 , 1936. 4) Philosophy of Nature (New York : 1949) . 5) General Theory of Knowledge, trans. A . Blumberg , (New York : 1974) . 6 ) Philosophical Papers, 1909-22, trans. P . Heath (Dordrecht: 1978) . Logics Without Existence Assumptions (Stockholm : 1968) . 1) Existence and Description in Formal Logic, Bertrand Russell: Philosopherofthe Century (Lon don : 1966) 181-200. 2) Advice on Modal Logic , Philosophical Problems in Logic. 1) Russell's Obj ections to Frege's Theory of Sense and Reference, A, 18, 1957-8 . 2) Proper Names, M, 57 , 1958. 3) On Determinables and Resemblance, PASS, 33 , 1959. 4) Speech Acts (Cambridge : 1969). 5) What Is an Intentional State? , M, 88, 1979. 6) Referential and Attributive , Mo, 62 . 2, 1979 . 1) Begrippet Egenskap (Stockholm: 1947) . 2) Studier Over Medvetandet Och lagiden (Stockholm: 1953 ) . Two Dimensional Modal Logic, JPL, 2, 1973 . 1 ) On the Logic of Complex Particulars. M, 58 . 1949. 2) Particulars, PPR, 13, 1952 . 3) Logical Subj ects and Physical Obj ects. PPR, 17, 1957. 4) Grammar and Existence: A Preface to Ontology , M, 69. 1960. 5) Abstract Entities, RM, 16. 1963. 6) Presupposing, EBR, 173-89 . 7) Science, Perception and Reality (London: 1963). 8) Science and Metaphysics (London : 1968) . 9) Hochberg on Mapping . Meaning , and Metaphysics . CPPL, 349-359. 10) Ontology and the Philosophy of Mind in Russell . Bertrand Russell's Philosophy, ed . G . Nakhnikian (London : 1974) , 57-100 . 1 1) Naturalism and Ontology (Reseda: 1979) . Comments on 'Things and Qualities' , M& E. 98-100. The Nature and Status of Essences. RM, l . 3. Self- Knowledge an d Self-Iden tity (Ithaca : 1963) . Comments on Things and Qualities·, M& E, 10 1-5 . The Modes of Referring and the Problem of Universals (Berkeley : 196 1 ). Mind as Behavior (Columbus, 0 . : 1924 ) . Nominalism and Common Names . PR , 7 1 . 1962 . Notes on Quanitification and Self-reference . The Paradox ofthe

Slater, B. A . Smart , J . J . C.

Smiley , T. J. Smullyan. A . F . Smullyan, R . Sommers, F . Specht, E . K. Sprigge . T. Stace, W. T. Stampe . D. Staniland, H. Stebbing , L. S.

Steenburgh, E . W. Van Stegmuller, W.

Stenius , E.

Sternfeld , R. Stoothoff, R. H. Storer, T. Stout , G. F.

Stove , D. C. Strawson, P. F.

Liar, ed . R. L. Martin, (New Haven: 1970). Talking About Something , A, 23 , 1962-63 . 1 ) Whitehead and Russell's Theory of Types, A , 10, 1949-50. 2) The Theory of Types Again , A, 1 1, 195 0-5 1 . 3) Sensations and Brain Processes, PR, 68, I 959. 4) Materialism , JP, 60, 1963. 5 ) Philosophy and Scientific Realism (London : 1963) . Sense Without Denotation, A , 20, 1960 . Modality and Description, JSL, 13 , 1948. Languages in which Self-Reference is Possible , JSL, 22 , 1957. 1) The Ordinary Language Tree , M , 1959. 2) Types and Ontology , PR , 72 , 1963 . 3) Are There Atomic Propositions? , FAP, 59-68. The Foundations of Wittgenstein 's LaterPhilosophy (New York : 1969) . 1) Consciousness, O T, 1 14-47. 2) Facts, Words and Belief(London : 1970) . 1) The Refutation of Realism , M, 43 , 1934 . 2) Russell's Neutral Monism , PBR, 35 1-84 . Toward a Causal Theory of Linguistic Representation , CPPL , 8 1-102 , 1956. Universals (Garden City , N . Y . : 1972). 1) Concerning Substance, PAS, 30, 1929-30. 2) Logical Positivism and Analysis, Proceedings of the British Academy , 19, 1933. 3) Thinking to Some Purpose (London: 1939) . The Problem of Simple Resemblance , PS , 25 , 1974 . 1 ) Ontologie und Analytizitat, Studia Philosophica , 1956. 2) Das Universalienproblem einst undjetzt (Darmstadt : 1965). 3) Metaphysik, Skepsis, Wissenschaft (Berlin: 1969) . 4) Main Currents in Contemporary German, British and American Philosophy trans. A. Blumberg , (Bloomington : 1970) . 1) Wittgenstein 's ' Tractatus ' (Oxford : 1960) . 2) The Problem of Color Incompatibility, O T, 245-63 . 3) Beginning with Ordinary Things, W& O , 27-52. 4) Sets, S, 27, 1974 . 1) A Restriction on Frege's Use of the Term 'True' , PS, 6, 1955. 2) Frege's Logical Theory (Carbondale , I ll. : 1966). 1) Note on a Doctrine of Frege, M, 72 , 1963. 2) What Actually Exists, PASS, 42 , 1968. An Analysis of Logical Positivism , Me, 3, 12 . 195 1 . 1 ) The Nature of Universals and Propositions (London : 1921) . 2) Are the Characteristics of Particular Thi ngs Universal or Particular?, PASS, 3, 1923 . 3) Studies in Philosophy and Psychology (London : 1930) . 4) Universals Again, PASS, 1 5 , 1936. 5) Things, Predicates , and Relations , AJPP, 18, 1940. Two Problems About Individuality , Al PP, 33 , 1955. 1) On Referring, M, 59, 1950.

44 1

Introduction to Logical Theory (London : 1952) . Particular and General, PAS, 54 , 1953-54. A Logician 's Landscape, P, 30 , 1955 . Singular Terms, Ontology, and Identity , M, 65 , 1956. Logical Subj ects and Physical Objects, PPR, 1 7 , 1957. Individuals (London : 1959) . Singular Terms and Predication , JP, 58, 1 96 1 . Identifying Reference and Truth-values, T, 30 , 2, 1964. The Bounds ofSense (London : 1966). Subject and Predicate in Logic and Grammar (London : 1974) . 12) Belief, Reference and Quantification , Mo, 62 , 2 , 1979. Meaning , Referring, and the Problem of Universals, I, 4, 1961. 1 ) The Concept of Truth in Formalized Languages, Studia Philosophica, i. 1935 ; and in Logic, Semantics, and Metamathematics (Oxford: 1956) . 2) The Semantic Conception of Truth, PPR, 4, 1944. Universals and Common Properties, A, 29. Ontological Reduction , JP, 68, 197 1. Two Theories of Truth : The Relation Between the Theories of John Dewey and Bertrand Russell, JP, 44 , 1947. l) Sense and Reference in Frege's Logic (Dordrecht: 1968). 2) Gottlob Frege : Die Abstraktion , Grundprobleme der groBen Philosophen , hrsg. v. Josef Speck , Band Philosophie der Gegenwart I, (Gottingen : 1972) , 8-44. 3) Ontologie im Mathematikunterricht? , Zentralblatt fiir Didaktik der Mathematik, 9, 3, 1977. Perception and Individuation , L&O, 261-85. 2) 3) 4) 5) 6) 7) 8) 9) 10) 1 1)

Stroll, A. Tarski, A .

Teichmann , J. Tharp , R. H. Thayer, H. S. Thiel, C .

Thomason , R . Thomason R. and Stalnaker , R. Thompson , M. H.

Thomsom , J. J. Toms, E. Trapp , R. W. Turnbull, R. G. Urmson , J. 0. Vaught , C. G. Veatch , H.

442

Modality and Reference , N, 2, 1968. 1 ) On the Distinction between Thing and Property , Th e Return to Reason, ed. J . Wild, (Chicago: 1953). 2) Abstract Entities, PR, 69 , 1960 . 3) Abstract Entities and Universals , M, 74 , 1965. Space , Time and Objects , M, 74 , 1965. Non-Existence and Universals, PQ, VI, 1956 . A nalytische Ontologie: Der Begriff der Existen z in Sprache und Logik (Frankfurt am Main : 1976) . Ockham's Nominalistic Logic : Some Twentieth Century Reflections, EO, 79-9 1. 1) Recognition , PAS, 56 , 1955-56. 2) Philosophical Analysis (London : 1956) . The Identity of Indiscernibles and the Concept of Substance , Southern Journal of Philosophy, 6, 1968. 1) Two Logics: The Conflict Between Classical and Neo­ Analytic Philosophy (Evanston : 1969). 2) Realism and Nominalism Revisited (Milwaukee : 1 954). 3) The Philosophy of Logical Atomism: A Realism Manque, EBR, 102-1 1 7. 4) To Gustav Bergmann : A Humble Petition and Advice , O T, 65-85 .

Waismann, F.

Walker, J . D. B. Wang, Hao

Warnock, G. J. Webb, C. W. Weinberg, J.

Weinstein, S. Weitz, M.

ed.

Welken. D. D. Wells, R. S. Wettstein, H . White. A, R. White , M. G. Whitehead, A. N. Wienpahl, P. D. Wiggins . D. Wilkin, J. Williams, B . A. 0 . Williams, C . J . F . Williams, D. C. Wilson, F.

1) Verifiability, Logic and Language, ed . A. G. N. Flew ( Oxford: 195 1 ), 1 17-44. 2) Introduction to Mathematical Thinking (New York : 1 95 1 ). 3) The Principles of Linguistic Philosophy, ed . R. Harre ( London : 1 965) . 4) Philosophical Papers, ed. B . McG uinness (Dordrecht: 1 977). A Study ofFrege (Ithaca : 1965). 1 ) The Logic of Many-Sorted Theories, JSL, 1 7, 1952 . 2) What Is an Individual ?, PR, 1953. 3) The Formalization of Mathematics, JSL, 1 9, 1954. 4) Russell and His Logic, Ratio, 7, 1 965. Metaphysics in Logic, Essays in Conceptual Analysis, ed . A. G. N. Flew (New York : 1966), 75-93 . The Antinomy of lndividuals, JP, 55, 1 958. 1 ) An Examination of Logical Positivism (London: 1936). 2) Contrary-to-Fact Conditionals, JP, 48, 1951 . 3 ) Concerning Undefined Descriptive Predicates of Higher Level , M, 53, 1954 . 4) Abstraction, Relation, and Induction (Madison: 1965). Truth and Demonstratives, N, 8, 1974. 1) Ryle's Theory of Concepts, FA P, 32 1-34. 2) Twentieth- Century Philosophy: The A nalytic Tradition (New York : 1966). Linguistic Nominalism, M, 79, 1970. 1 ) Frege 's Ontology, RM, 4, 195 1. 2) Is Frege 's Concept of Function Valid? , JP, 60, 1963 . 1 ) Can What Is Asserted Be A Sentence?, PR, 85, 1976. 2) Indexical Reference and Propositional Content, PS, 36, 1979. 1 ) G. E. Moore: A Critical Exposition (Oxford: 1958) . 2) Truth (New York : 1970) . I ) Toward Reunion in Philosophy (Cambridge : 1956) . 2) The Analytic and the Synthetic ; An Untenable Dualism, S&PL, 272-86. 1 ) Principia Mathematica, see B. Russell. 2) The Concept of Nature (Cambridge : 1 9 1 9). Frege's Sinn und Bedeutung, M, 59, 1950. 1 ) Identity Statements, Analytical Philosophy. 2) Iden tity and Spatio- Temporal Continuity (Oxford: 1 967). 3) The Individuation of Things and Places, U& P, 307-35. Sellars on Bradley's Paradox, PS, 36, 1 979. Mr. Strawson on Individuals, P, 36 , 1 96 1. What Is Truth ? (Cambridge : 1976). 1 ) The Elements of Being, RM, 6, 1 953, 3-1 8, 171-93. 2) Matter and Fonn, I and ll, PR, 67, 1958, 29 1-3 1 2, 499-52 1 . 3) Necessary Facts , RM, 1 6, 1 963 . 1 ) Definition and Discovery, BJPSc, 1 8, 1967. 2) The Role of a Principle of Acquaintance in Ontology, The Modern Schoolman, 47 , 1 , 1969 , 37-56. 3) Dispositions: Defined or Reduced? Al P, 47, 2 , 1 969 . 4) Why I Am Not Aware of Your Pain , O T, 276--300 .

443

Wilson , N. L. Wisdom, J . Wittgenstein, L .

Wollheim , R. Wolterstorff, N .

Woodger, J . H. Woods, J . Woods, M. J . Woozley, A . D. Wright, C. von Wright , G . H. Zabeeh , F. Ziff, P. Zin k, S.

5) Weinberg's Refutation of Nominalism, D, 8, 3, 1969 . 1) Description and Designation, JP, 50 , 1953. 2) Space , Time , and Individuals, JP, 52 , 1955. 3) Substances Without Substrata, RM, 12, 1959. Logical Constructions (New York: 1969) . 1) Notebooks 19 14-16, ed. G . H. von Wright and G . E. M. Anscombe (Oxford: 1961). 2) Proto Tractatus - A n Early Version of Tractatus Logico Philosophicus, ed. B. F. McGuinness, T. Nyberg and G. H. von Wright , trans. D. F. Pears and B. F. McGuinness (London : 197 1). 3) Tractatus Logico-Philosophicus, trans. D. F. Pears and B. F. McGuinness (London : 1961) . 4) Some Remarks on Logical Form, PASS, 9, 1929. 5) The Blue and Brown Books (Oxford: 1958). 6) Remarks on the Foundations ofMathematics, ed . G . H . von Wright, R. Rhees and G. E . M. Anscombe , trans. G. E. M. Anscombe (Oxford: 1964) . 7) Philosophical Investigations, ed. G. E. M. Anscombe and R. Rhees, trans. G. E. M. Anscombe (Oxford: 1953). 8) Remarks on Color, ed. G. E. M. Anscombe. trans. L. McAlister and M. Schattle (Berkeley: 1977). F. H. Bradley (Harmondsworth: 1959) . 1) Are Properties Meanings? . JP, 57 , 1960. 2) Qualities, PR, 69 , 1960. 3) On the Nature of Universals, U&P, 159-85. 4) On Universals (Chicago: 1970). Science Without Properties , BJPSc, 2, 1952. Fictionality and the Logic of Relations , The Southern Journal of Philosophy, 7, 1969. 1) Identity and Individuation, Analytical Philosophy, 2 . 20-30. 2) The Individuation of Things and Places. PASS, 38, 1963 . 1) Universals, Encyclopedia of Philosophy, 8 , 194-206 . 2) Universals, Theory of Knowledge (London: 1949) . Wittgenstein on the Foundation ofMa thematics (London: 1 980) . Logical Studies (London: 1957) . Universals ( The Hague: 1966) . Semantic Analysis (Ithaca: 1960). The Meaning of Proper Names. M, 71, 1963.

Index of Names

Allaire , E . B . 2 1 , 183 , 202 Anscombe , G . E. M . 156 Aquinas, T. 243 Aristotle 1 8 1 , 236 Armstrong, D . 31 1 Barber, K . 256f. Benacerraf, P. 336 Bergmann , G. 9, 1 1 , 12, 2 1 , 33, 34 , 4 1 44 , 1 04, 1 57, 22 1 f. , 23 1 , 243 , 275-277 , 337 , 373 , 39 1 f. Berkeley , G. 20, 240 . 242 , 243 , 358, 365 Bradley , F. H. 1 1-18, 20 , 22 , 33 , 42, 206, 210, 212, 214, 2 1 9 , 220 , 242, 243 , 253, 256 . 278, 279 , 286 , 308 . Brodbeck . M . 33

Gasking, D . 337 Geach , P. 60, 61 Gombrich , E . 30 Goodman , N . 1 1 , 28-30 , 45 , 161-163 , 166, 173 , 174, 180, 182, 185, 1 86 Grosseteste , R. 243 Grossmann , R . 2 1 Hahn, H . 335 , 338 Hegel , G. W. F. 102, 372 Hume , D . 40, 370 , 409 Jackson, H . 59 Johnson, W. E . 42

Kant , I. 1 03 , 37 1 Kaplan , A . 378 Kaplan , D . 395f. Carnap , R . 1 1 , 2 1 , 32, 34 , 35 , 45 , 60, Kripke , S. 37-39, 40, 97 339f. Kuratowski , C. 223 , 225 , 234, 258, 259, Cartwright, R. 88f. 276 Cassin , C. 60, 80 , 81 Castaneda , H. N . 337 Leibniz, G . W. 20 , 252 Church , A. 60, 1 85 Lindstrom , P. 1 83 Churchman , C. W. 30 Loux , M . J . 222 , 229 , 230 Cocchiarella, N . 149 Davidson , D . 1 3 , 19, 37 , 46 , 167, 173- McTaggart , J . M . E. 1 4 Meinong , A. 286 , 328 , 333 1 76, 180, 1 83 , 187 , 279f. Moore , G . E . 9 , 1 1 , 1 5 - 2 1 , 29-3 1 , 34 , Descartes, R. 9, 243 4 1 , 44, 68, 98 , 103 , 240, 243 , 278 , 39 1 Dewey , J . 32, 33 Donagan , A. 1 96, 278 von Neumann, J . 276 Donellan , K . 1 29, 1 30 Dretske , F. 3 1 1 Dummett, M . 19, 56, 57 Occam , W. 375 Feigl , H . 33 , 382 Frege , G. 1 1-3 1 , 35 , 36, 44-85 , 1 42-144, 1 49, 207, 209, 212, 221 , 226, 228, 235 , 279 , 336, 389 , 390, 415

Peano , G . 322 , 335 Pears , D. 228 Pia to 1 9 Putnam , H . 374f.

445

Quine , W. V . 1 1 , 28 - 37, 46, 90-1 03 , 130, 1 33 -135 , 140-149, 155 , 165 , 166, 170-174 , 180, 182, 192 , 194 , 204, 246 , 249 , 279 , 282 , 283 , 337, 353f. , 374, 395f.

Spence , K. W . 337 Spinoza , B. 14 Storer , T. 237 Strawson, P. F. 12, 38, 40, 42, 105 -132, 141 , 149

Russell , B . A . W . 9 , 1 1-22, 27 -45 , 58 , 60- 85 , 90 -93 , 98 , 99 , 105 -132, 180, 196f. , 205 , 234 , 244f. , 263 , 282 , 283 , 286 , 294 , 303 , 3 1 3 , 321f. , 34 1f. , 389f. , 415

Tarski , A. 23 , 140, 174, 175, 181 , 1 83 Tooley , M . 3 1 1

Watson , J . B . 353 , 374 Weinberg , J. 237 Wiener, N. 223 , 225 , 234, 258, 259, 276 Sartre , J . P. 294, 408, 409 Wilkin , J . 230 Schlick , M . 1 1 , 32 Wilson , F. 2 1 Searle , J . 60, 61 , 74 , 130 Whitehead , A . N . 322, 352 Sellars , W. 1 1 , 19, 28 -37, 45 , 46 , 165 - Wisdom , J . 42 167, 173, 174 , 1 80, 183, 186f. , 210f. , Wittgenstein , L. 1 1-14 , 19, 24- 3 1 , 4 1 , 224 , 230, 395 45 , 145 , 1 80-184 , 226, 306, 3 1 3f. , 325 , Singer, E. A . 30, 32 334, 336 , 390, 405 , 417

Author's Note

The articles appeared originally as follows : 1 . Frege on Concepts as Functions : A Fundamental Ambiguity , Theoria, 37 , 197 1 . 2 . Russell's Attack on Frege ' s Theory of Meaning , Philosophica, 1 8 , 2 , 1 976 . 3. Professor Quine , Pegasus , and Dr. Cartwright , Philosophy of Science, 24 , 2 , 1957 . 4. On Pegasizing , Philosophy and Phenomenological Research, 1 7 , 1957 . 5 . Strawson and Russell on Reference and Description , Philosophy of Science, 37 , 3 , 1970. 6. Nominalism , General Terms , and Predication , The Monist, 7 1 , 1978 . 7 . Nominalism , Platonism and Being True of, Nous, l , 3 , 1 967 . 8 . Mapping, Meaning , and Metaphysics , Midwest Studies m Philosophy , II , 1 977 . 9 . Sell ars and Goodman on Predicates , Properties and Truth , Midwest Studies in Philosophy, III , 1978 . 1 0 . Russell's Proof of Realism Reproved , Philosophical Studies, 37 , 1980 . 1 1 . Logical Form , Existence , and Relational Predication , Foundations of Analytic Philosophy, ed . P . French et . al . (Minneapolis: 198 1 ) . 1 2 . Elementarism , Independence , and Ontology , Philosophical Studies, 1 2 , 1 96 1 . 13 . Ontology and Acquaintance , Philosophical Studies, 1 7 , 1966. 1 4 . Things and Descriptions , American Philosophical Quarterly, 3 , 1 , 1 966 . 1 5 . Universals, Particulars , and Predication , Review ofMetaphysics, 19, 1965 . 1 6 . Facts and Truth , this paper is based on 'Explaining Facts' , Metaphilosophy, 6 , 1975 . 1 7 . Negation and Generality, Nous, III , 3 , 1969. 1 8 . Arithmetic and Propositional Form in Wittgenstein's Tractatus, Essays on Wittgenstein, ed . E . D . Klemke (Urbana: 197 1 ) . 1 9 . Russell's Reduction of Arithmetic to Logic. Essays on Bertrand Russell, ed . E . D . Klemke (Urbana: 1 970) . 447

20. Properties, Abstracts, and the Axiom of Infinity, Journal of Philosophical Logic, 6, 1 977. 2 1 . Of Mind and Myth, Methodos, II, 1 959. 22. Physicalism, Behaviorism, and Phenomena, Philosophy of Science, 26, 2, 1959. 23. Belief and Intention, Philosophy and Phenomenological Research, XL, I, 1979. The necessary permissions have been obtained from the editors (and for 8, 9, and 1 1 from the University of Minnesota Press, for 1 0 and 20 from D. Reidel publishing company, and for 1 8 and 1 9 from the University of Illinois Press) and are gratefully acknowledged. Additions and alterations have been made to some of the essays.