سکوت

INTRODUCTION

Itis now well over half a century since the heyday of the philosophical move­ment known as logical positivism or logical empiricism. Depending on how one counts, it is now approaching half a century since the official demise of this movement.' Since that demise, it has been customary to view logical pos­itivism as a kind of philosophical bogeyman whose faults and failings need to be enumerated (or, less commonly, investigated) before one's favored new approach to philosophy can properly begin. Such an attitude toward logical positivism and its demise has been widely prevalent, not only in the narrower community of philosophers of science (who characteristically have proceeded against the background of Kuhn's well-known critique), but also in the broader philosophical community as well. With our increasing histor­ical distance from logical positivism, however, a more dispassionate attitude also has begun inevitably to emerge. No longer threatened or challenged by logical positivism as a live philosophical option, it is becoming increasingly possible to consider this movement as simply a part of the history of phi­losophy, which, as such, can be investigated impartially from an historical point of view. Indeed, we have seen in recent years a veritable flowering of historically oriented reappraisals of logical positivism.

In the course of these reappraisals, it has become clear-not at all sur­ prisingly, of course -that the above-mentioned postpositivist reaction gave birth to a large number  of seriously misleading ideas about the origins, motivations, and true philosophical aims of the positivist movement. (One can hardly expect philosophical critics, concerned largely with their own agendas rather than with historical fidelity, to generate anything other than stereotypes and misconceptions.) I will discuss what I take to be some of the most important of such misleading ideas in what follows, but I also hope to show-or at least to suggest-that  achieving  a better  understand­ ing of the background, development, and actual philosophical context of logical positivism is not merely of historical interest. For the fact remains that our present situation  evolves  directly-for  better  or for worse-from the rise and fall of positivism, and what I want to suggest is that we will never successfully move beyond our present philosophical situation until we at­ tain a properly self-conscious appreciation of our own immediate historical background.

Perhaps the most misleading of the stereotypical  characterizations views logical positivism as a version of philosophical "foundationalism." The posi­ tivists- so this story goes -were concerned above all to provide a philosophi­ caljustification of scientific knowledge from some privileged, Archimedean vantage point situated somehow outside of, above, or beyond  the actual (historical) sciences themselves. More specifically, they followed the lead of the logicist reduction of mathematics to logic, where the latter also is un­ derstood as fundamentally foundationalist in motivation  and import. Just as the logicists attempted to justify mathematical knowledge and place it on a secure foundation by means of a derivation from  (supposedly more certain) logical knowledge, so the positivists attempted to justify empirical science and place it on a secure foundation by logically constructing the concepts of empirical science on the basis of the (supposedly more certain) immediate data of sense.   Thus, formal logic furnished the foundational enterprise with  the required Archimedean  standpoint located  outside of the actual (historical)  sciences themselves,  and phenomenalist reduction­ ism, carried out rigorously using the tools of formal logic (as epitomized

in Carnap's [1928a] Der logische Aufbau der Welt) , then provided the desired epistemological justification of the sciences.3

This conception of the aims and posture of philosophy vis a vis the special

sciences represents an almost total perversion of the actual attitude of the logical positivists, who rather considered their intellectual starting point to be a rejection of all such philosophical pretensions. An eloquent example is found in the first paragraph of a 1915 paper by Schlick on the need for philosophy to adapt itself to the new findings of relativity theory:

We have known since the days of Kant that the only fruitful method of all theoretical philosophy consists in critical inquiry into the ultimate prin­ ciples of the special sciences. Every change in these ultimate axioms, every emergence of a new fundamental principle, must therefore set philosophical activity in motion, and has naturally done so even before Kant. The most bril­ liant example is doubtless the birth of modern philosophy from the scientific discoveries of the Renaissance. And the Kantian Critical Philosophy may itself be regarded as a product of the Newtonian doctrine of nature. It is primar­ ily, or even exclusively, the principles of the exact sciences that are of major philosophical importance, for the simple reason that in these disciplines alone do we find foundations so firm and sharply defined, that a change in them produces a notable upheaval, which can then also acquire an influence on our world-view. (1915/1978-9, p. 153)

Schlick goes on to argue that neither of the two prevailing philosophical systems -neither the "neo-Kantianism" of Cassirer and Natorp nor the "pos­ itivism" of Petzoldt and Mach -can dojustice to Einstein's new theory and therefore both systems must be abandoned. Entirely new philosophical princi­ ples, based on the work of Einstein himself and of Poincare, are necessarily required.

For Schlick, then, philosophy as a discipline is in no way foundational with respect to the special sciences. On the contrary, it is the special sciences that are foundational for philosophy. The special sciences -more specifically, the "exact sciences" -simply are taken for granted as paradigmatic of knowledge and certainty. Far from being in a position somehow tojustify these sciences

from some higher vantage point, it is rather philosophy itself that is inevitably in question. Philosophy, that is, must follow the evolution of the special sciences so as to test itself and, if need be, to reorient itself with respect to the far more certain and secure results of these sciences. In particular, then, the central problem of philosophy is not to provide an epistemological foundation for the special sciences (they already have all the foundation they need) , but rather to redefine its own task in the light of the recent revolutionary scientific advances that have made all previous philosophies untenable.4

Moreover,  this conception  of the proper  stance of philosophy vis a vis

the special sciences is not peculiar to Schlick; it is in fact characteristic of the logical positivists generally. This is so even -and indeed especially-of Carnap's Aufbau (which work, of course, is supposed to be most representa­ tive of foundationalist epistemology) .s It is true that the Aufbau presents a phenomenalist reduction of all concepts of science to the immediately given data of experience. Yet the point of this construction has little if anything to do with traditional foundationalism. First, Carnap shows no interest what­ ever in the philosophical skepticism that motivates, for example, Russell (1914) in Our Knowl,edge of the External World, nor does the text of the Aufbau at any point engage the traditional vocabulary of "certainty," "doubt," · usti­ fication," and so on.6 Second, and more important, Carnap is perfectly ex­ plicit that the particular constructions he employs depend entirely on the actual "results of the empirical sciences" (§ 122) and, accordingly, that the particular constructional system that he presents is best viewed as a "ratio­ nal reconstruction" of the actual (empirical) process of cognition (§100) .7 Thus, for example,  Carnap's choice of holistic  "elementary experiences"

rather than atomistic sensations as the basis of his system is grounded in the empirical findings of Gestalt psychology (§67) , his definition of the vi­ sual sense modality depends on the (supposed) empirical fact that it is the unique sense modality having exactly five dimensions (§86), and so on. In particular, then, the findings of the special sciences, such as empirical psy­ chology, are in no way in question in the Aufbau: once again, it is philosophy that must adapt itself to them rather than the other way around.

The aim of the Aufbau, therefore, is not to use logic together with sense data to provide empirical knowledge with an otherwise missing epistemo­ logical foundation orjustification. Its aim, rather, is to use recent advances in the science of logic (in this case, the Russellian type-theory of Principia Mathematica ) together with advances in the empirical sciences (Gestalt psy­ chology in particular) to fashion a scientifically respectable replacement for traditional epistemology. Carnap's depiction of the construction of scien­ tific knowledge from elementary experiences via the logical techniques of Principia Mathematica enables us to avoid the metaphysical excesses of the tra­ ditional epistemological schools- "realism," "idealism," "phenomenalism," "transcendental idealism"(§177) -while simultaneously capturing what is correct in all of these schools: allowing us to represent, in Carnap's words, the "neutral basis [neutrale Fundament]" common to all (§178) .8

II

According to the standard picture of logical positivism briefly sketched ear­ lier, that movement is to be understood not only as a species of founda­ tionalist epistemology but also as a version of empiricist epistemology in the tradition of Locke, Berkeley, Hume, Mach, and Russell's external-world program.9 The immediately given data of sense are viewed as the primary examplars of knowledge and certainty, and all other putative claims to knowl­ edge are judged to be warranted -or, as the case may be, unwarranted -in light of their relations to the immediate data of sense. (When empiricist re­ ductionism is acknowledged to have failed, the positivists' naive empiricism then is thought to express itself in the doctrine of an epistemically privileged and theory-neutral observation language against which all scientific claims are to be tested.)   I dealt earlier with the conception of the positivists as

4. For details of this kind of approach to the Aujbau, see Part Two (this volume) and Richardson (1990, i992, i998).
5. This is, of course, the conception promulgated especially in Ayer's ( i936) Language, Truth

and Logic:  compare the first sentence of Ayer's Preface, where the views put forward are characterized as "the logical outcome of the empiricism of Berkeley and David Hume."

foundationalists, and this question is, I think, rather easily disposed of. The question of empiricism -and, in particular, of the relationship between the positivists and the traditional empiricism of Locke, Berkeley, Hume, and Mach -is, however, considerably more delicate.

The first point to notice is that the positivists' main philosophical con­ cerns did not arise within the context of the empiricist philosophical tradi­ tion at all. Rather, the initial impetus for their philosophizing came from late nineteenth-century work on the foundations of geometry by Riemann, Helmholtz, Lie, Klein, and Hilbert-work that, for the early positivists, achieved its culmination in Einstein's theory of relativity. 10 The principal philosophical moral that the early positivists drew from these geometrical developments was that the Kantian conception of pure intuition and the synthetic a priori could no longer be consistently maintained -consistently, that is, with the situation now presented to us by the exact sciences. In particular, Hilbert's (1899) logically rigorous axiomatization of Euclidean geometry shows conclusively that spatial intuition has no role to play in the reasoning and inferences of pure geometry, and the development of non-Euclidean geometries together with their actual application to nature by Einstein show conclusively that our knowledge of geometry cannot be synthetic a priori in Kant's sense. All the early positivists were thus in agree­ ment that the strictly Kantian conception of the a priori must be rejected, and this rejection of the synthetic a priori constituted a centrally important element in what they came to call their "empiricism."

Yet it is equally important to notice, in the second place, that the positivists did not react to the demise of the Kantian synthetic a priori by adopting a straightforwardly empiricist conception of physical geometry of the kind tra­ ditionally imputed to Gauss (who is reported to have attempted to determine the curvature of physical space by measuring the angle sum of a terrestrial triangle determined by three mountaintops). On the contrary, all the early positivists also strongly rejected this kind of empiricist conception (which they attributed to Gauss, Riemann, and, at times, Helmholtz) and, rather, followed the example of Poincare in maintaining that there is no direct route from sense experience to physical geometry: essentially nonempirical factors, variously termed "conventions" or "coordinating definitions," must necessarily intervene between sensible experience and geometrical theory.

10 Most of the early writings of the positivists focused on these revolutionary mathematical­ physical developments. In addition to the 1915 paper of Schlick cited earlier, see Schlick (1917/1978), Reichenbach (1920/1965), and Carnap (1922). Interest in these themes also was stimulated by Weyl's 1919 edition of Riemann's 1854 dissertation and the 1921 edition of Helmholtz's epistemological writings by Schlick and the physicist Paul Hertz.

The upshot is that it is in no way a straightforward empirical matter of fact whether space is Euclidean or non-Euclidean. 11

This radically new conception of physical geometry-neither strictly Kan­ tian nor strictly empiricist-was formulated by Reichenbach ( 1920) in an especially striking fashion in his first book, The Theory of Relativity and A Priori Know/,edge. Reichenbach maintains a sharp distinction, within the con­ text of any given scientific theory, between two intrinsically different types of principles: "axioms of connection" are empirical laws in the traditional sense recording inductive regularities involving terms and concepts that are already sufficiently well defined; "axioms of coordination," on the other hand, are nonempirical statements that must be antecedently laid down before the relevant terms and concepts have a well-defined subject matter in the first place. (Thus, for example, Gauss's attempt empirically to deter­ mine the curvature of space via a terrestrial triangle inevitably fails because it tacitly presupposes that light rays travel in straight lines and, therefore, that the notion of a "straight line" is already well defined. But how, indepen­ dently of the geometrical and optical principles supposedly being tested, can this possibly be done?) These nonempirical axioms of coordination ­ which include, paradigmatically, the principles of physical geometry- are thus "constitutive of the object of knowledge," and, in this way, we can therefore vindicate part of Kant's conception of the a priori.

For, according to Reichenbach, the a priori had two independent aspects in Kant: the first involves necessary and unrevisable validity, but the second involves only the just-mentioned feature of "constitutivity." The lesson of modern geometry and relativity theory, then, is not that the Kantian a pri­ ori must be abandoned completely, but rather that the constitutive aspect must be separated from the aspect of necessary validity. Physical geometry is indeed nonempirical and constitutive -it is not itself subject to straightfor­ ward observational confirmation and disconfirmation but rather first makes possible the confirmation and disconfirmation of properly empirical laws (viz., the axioms of connection). Nevertheless, physical geometry can still evolve and change in the transition from one theoretical framework to an­ other: Euclidean geometry, for example, is a priori in this constitutive sense in the context of Newtonian physics, but only topology (sufficient to admit a Riemannian structure) is a priori in the context of general relativity.

Reichenbach concludes that traditional empiricism is in error in not rec­ ognizing the a priori constitutive role of axioms of coordination, and it is clear, moreover, that, despite some terminological wrangling on this point, the other logical positivists are in substantial agreement. 12 One of the cen­ tral themes of Schlick's ( 1918) General Theory of Knowledge, for example, is a sharp dichotomy between raw sensible acquaintance and genuine ob­ jective knowledge. Immediate contact with the given is both fleeting and irredeemably subjective. Objective knowledge therefore requires concepts and judgments, which are to be carefully distinguished from intuitive sensory presentations. Concepts and judgments are, in fact, only possible in the context of a rigorous formal system, of which Hilbert's axiomatization of ge­ ometry is paradigmatic. More specifically, the Hilbertian notion of "implicit definition" of scientific concepts via their logical places in a formal system (which notion is now associated by Schlick with Poincare's conventionalist philosophy of geometry) can alone explain how rigorous, exact, and truly objective representation is possible. Here, we are obviously very far from traditional empiricism and very close indeed to the supposedly antipositivist doctrine of the theory ladenness of observation. 13

Once again, the same is true even -indeed, especially-of Carnap's Aujbau. Carnap outlines an elaborate construction of all scientific concepts from the immediately given data of sense using the logical machinery of Principia Mathematica. Yet he also holds that the objective meaning of scien­ tific concepts can in no way depend on merely ostensive contact with the given. On the contrary, intersubjective communication is possible only in virtue of the logi,cal structure of the concepts in question arising from their log­ ical places within the total system of scientific knowledge. More specifically, intersubjective meaning must derive entirely from what Carnap calls "purely structural definite descriptions" rather than from sensory ostension (§§12- 15). (Thus, for example, the visual sense modality is defined in terms of the purely formal properties of its dimensionality rather than in terms of its phenomenal  content.) In this sense, the elaborate logical structure erected

above the basic elements of the system (the elementary experiences) is actu­ ally more important than the basic elements themselves. Objective meaning flows from the top down, as it were, rather than from the bottom up. 14

Carnap illuminatingly  articulates the precise meaning of the resulting "empiricism" in the Preface to the second edition ( 1961) of the Aujbau:

For a long time, philosophers of various persuasions have  held  the view that all concepts and judgments  result  from  the  cooperation  of  experience and reason. Basically, empiricists and rationalists agree in  this view,  even though both sides give a different estimation of the two factors, and obscure the essential agreement by carrying their viewpoints to extremes. The thesis which they have in common is frequently stated in the following simplified version: The senses provide the material of cognition, reason works up [ver­ arbeitet] the material soas to produce an organized system of knowledge. There arises then the problem of finding a synthesis of traditional empiricism and traditional rationalism. Traditional empiricism rightly emphasized the con­ tribution of the senses, but did not realize the importance of logical and mathematical forms. . ..I had realized, on the one hand, the fundamental im­ portance of mathematics for the formation of a system of knowledge and, on the other hand, its purely  logical,  purely  formal character  to which  it owes its independence from the contingencies of the real world. These insights formed the basis of my book.... This orientation is sometimes called "logical empiricism" (or "logical positivism"), in order to indicate the two components. (1928a/1967,  pp.  v-vi)

In thus emphasizing the central importance of a priori formal elements in first providing objective meaning for the otherwise undigested immediate data of sense, logical positivism has, it seems to me, broken decisively with the traditional empiricism of Locke, Berkeley, Hume, and Mach -which tradition is understood by the positivists themselves as mistakenly giving epistemic centrality to precisely such undigested immediate sensory data. Perhaps the best way to put the point is that the logical positivists have staked out an entirely novel position that is, as it were, intermediate between tradi­ tional Kantianism and traditional empiricism: it gives explicit recognition to the constitutive role of a priori principles, yet, at the same time, it also rejects the Kantian characterization of these principles as synthetic a priori.

III

The novel philosophical position briefly sketched above, however, is faced with  several  fundamental  problems.  The  heart  of  the  positivists'  new

conception is the idea of constitutive but nonsynthetic a priori principles un­ derlying the possibility of genuinely objective scientific knowledge. There is a sharp distinction, in particular, between conventions, coordinating defini­ tions, or axioms of coordination, on the one hand, and empirical principles properly so-called, on the other. On the one side lie the principles of pure mathematics and, at least in the context of some physical theories, the prin­ ciples of physical geometry; on the other side lie standard empirical laws such as Maxwell's equations and the law of gravitation. But what is the basis for this distinction, and how, more generally, are we sharply to differentiate the two classes of principles? On a strictly Kantian view, such a distinction is, of course, grounded in the fixed constitution of our cognitive faculties, and the question of differentiation is correspondingly straightforward:  the a priori principles are precisely those possessing necessary and unrevisable validity. 15 Now, however, we have explicitly acknowledged that a priori con­ stitutive principles possess no such necessary validity and, in fact, that these principles may evolve and change -in response to empirical findings-with the progress of empirical science. So what exactly distinguishes our a priori principles from ordinary empirical laws properly so-called?

A second problem is perhaps even more fundamental. The logical posi­ tivists, I have argued, strongly rejected a foundationalist conception of phi­

losophy vis a vis the special sciences.  There is no privileged vantage point

from which philosophy can pass epistemicjudgment on the special sciences: philosophy is conceived rather as following the special sciences so as to re­ orient itself in response to their established results. But what then is the peculiar task of philosophy, and how, in general, does it relate to the spe­ cial sciences? Is philosophy itself simply one special science among others, and, if not, from what perspective does it then respond to and rationally reconstruct the results of the special sciences? The positivists are nearly unanimous in explicitly rejecting a naturalistic conception of philosophy as simply one empirical science among others-a branch of psychology, per­ haps, or of the sociology of knowledge. 16 On the whole, they instead prefer

to think of philosophy as in some sense a branch of logi,c and to conceive the peculiarly philosophical task as that of "logical analysis" of the special sci­ ences. Yet the perspective or point of view from which such logical analysis is to proceed remains radically unclear.

No real answer to these questions was forthcoming until Carnap's ( 934c) Logi,cal Syntax of Language. Here, Carnap is once again responding to re­ cent developments in the exact sciences: to Heyting's formalization of in­ tuitionistic arithmetic, for example, and, above all, to Hilbert's program of "metamathematics." Moreover, he is once again attempting to neutralize the philosophical disputes arising in connection with these developments by showing how all parties involved are in possession of part of the truth; the remaining part that appears to be in dispute then is argued to be not subject to rational debate at all. More specifically, the dispute is declared to be a matter of convention in precisely Poincare's sense: there is simply no fact of the matter concerning which party is "correct," and thus the choice between them is merely pragmatic. 17

The dispute in question arises from increasing appreciation of how fun­ damentally the program of Principia Mathematica is threatened by the para­ doxes and involves the three traditional schools in the foundations of math­ ematics: logicism, formalism, and intuitionism. Carnap responds to this dispute by declaring that each side is simply putting forth a proposal to con­ struct a formal system or calculus of a certain kind: logicism proposes to construct an axiomatization of mathematics using the rules of classical logic, intuitionism proposes to construct an axiomatization using the more restric­ tive rules of intuitionistic logic, and so on. The essential point is that no such formal system or calculus is more correct than any other-indeed, the notion of correctness is entirely inappropriate here. Considered simply as proposals to construct formal systems of various kinds, all of the apparently opposing philosophies are then equally correct, and the choice between them can only be a purely pragmatic question of convenience. In the Fore­ word to Logi,cal Syntax, Carnap calls this standpoint the Principk of Tolerance and then remarks:

The first attempts to cast the ship oflogic off from the terrafirmao f the classi­ cal forms were certainly bold ones, considered from the historical point ofview. But they were hampered by the striving after "correctness." Now, however, that

impediment has been overcome, and before us lies the boundless ocean of unlimited possibilities.  ( i934c/ 1937, p. xv)

It is thus with  an  exuberant  sense of liberation  that  Carnap extends Poincare's conventionalism to logic itself.

But how is it possible for Carnap to maintain a stance of neutrality with respect to logic itself? Here is where the fundamental insights of Hilbert's metamathematics come into play. For we are to describe the logical rules governing the formal systems or calculi under consideration within the rnetadiscipline of logical syntax: each system is viewed simply as a set of strings of symbols together with rules for manipulating such strings (entirely inde­ pendently of any question concerning their "meanings"), and any and  all fomal systems then can be specified from the neutral standpoint of a purely syntactic meta-language. More precisely, the syntactic meta-language need employ only the limited resources of primitive recursive arithmetic, and we thus can describe the rules of classical systems, intuitionist systems, and so on from a standpoint that is neutral between them. Our aim is not tojustify one system over others as inherently more  correct, but simply to describe the consequences of choosing any such system.

Carnap is now in a position precisely to articulate the method and stance of logical analysis. This paradigmatically philosophical enterprise is simply a branch of logical syntax: specifically, the logical syntax of the language of science. 18 We thus are concerned with what Carnap calls "the physical language" (§82). The physical language, as opposed to purely mathematical languages, is characterized by two essentially distinct types of rules: logi.cal rules represent the purely formal, nonempirical part of our scientific theory, whereas physical rules represent its material or empirical content. More­ over, this purely syntactical distinction between logical and physical rules is Carnap's precise explication for the traditional distinction between ana­ lytic and synthetic judgments (§§51-2) . Finally, and what most concerns us here, it is also clear that these logical rules or analytic sentences of the language of science represent Carnap's precise explication for the consti­ tutive -but nonsynthetic -a priori discussed above: these logical rules, in other words, syntactically represent Poincare's (and Schlick's) conventions