For many philosophers, the search for an adequate or perspicuous representation of the contents of our thoughts, concepts and theories has been very important. The whole project of philosophical analysis, for example, involves just such a search. It emphasizes the possibility of a kind of semantic investigation which leads to a representation of a thought that renders all of its content fully explicit. Semantic complexity hidden within individual words or concepts is brought out into the open, and the misleading connotations of the familiar expressions of our thoughts can be overcome.
This vague description could fit a wide range of philosophical projects designed to meet a number of different philosophical needs. An early modern paradigm is provided by Descartes’s use of clear and distinct ideas. If I can replace my vague idea of matter or God by a clear and distinct alternative, then I attain an idea which contains nothing that is unclear. And, supposedly, when I obtain a clear and distinct perception of something, I can see that it is true: the confusion and obscurity which make doubt possible are absent. Replacing my confused perception with a clear one—or obtaining a clear perception of what my confused perception confusedly contained—is of value as a means of reaching the truth. In the same vein, we may judge that an axiomatization of a geometrical theory provides a perspicuous representation of a range of facts which renders their contents and the grounds of their truth wholly explicit. Other paradigms can be provided by Kant’s metaphysical deduction: if only we possess a perspicuous representation of the structure of all possible thoughts about the empirical world, then we can derive the categorial structure of that world from an examination of formal features of these representations. And a variety of empiricist and pragmatist philosophers have supposed that an attempt at unpacking the meanings of propositions, employing a general theory of the form that such unpacking should take, will enable us to see that some of them have no real content at all. The search for such canonical clear or perspicuous representations can be a tool for discovering truths about reality, formulating philosophical theories and guarding against metaphysical illusion.
Any project of this kind involves two distinct elements: a set of philosophical motivations which make a search for clarity desirable, and a view about the sorts of features which must be brought to the surface if such clarification is to be possible. One could be sceptical of the value of such projects either by denying the interest or importance of the purposes to which such clarification is intended to lead or by questioning whether such clarification is genuinely possible. Both Wittgenstein and Quine, it seems to me, stand in rather ambivalent relations to such projects. Quine, for example, introduces a canonical notation for the expression of scientific theories which is supposed to yield philosophical insights and, through questioning the analytic/synthetic distinction, rejects the idea that propositions and beliefs have a semantic content which can actually be clarified. Wittgenstein, through his use of the notion of grammar and through his explicit emphasis on the value of a perspicuous description of features of our practices, shows sympathy for some aspects of such projects. On the other hand, there are other features of his views which are alien to them. I want to explore these possible connections between their views by suggesting that both take seriously an idea which I shall call the shallowness of reflection. Since it is plausible that both reached their positions through critical reflection on views defended by Rudolf Carnap, I shall use their varying reactions to his positions as a focus for what I want to say. After a sketch of Carnap’s ideas about clarity and explication (second section), I shall use them to discuss Quine’s dispute with Carnap about analyticity and the bearing of his commitment to a naturalistic approach to epistemology (third section) before turning to some Wittgensteinian themes.
The writings of Carnap provide a good illustration of some of these themes, and there are good reasons for introducing them here.
Many of Quine’s philosophical views were constructed through sympathetic criticism of Carnap’s philosophy. And not only were Carnap’s views developed under the influence of his attempts to understand Wittgenstein’s Tractatus Logico-Philosophicus, but some of the developments in Wittgenstein’s thought after 1930 were a reaction to the work of Carnap and those in the Vienna Circle who were under his influence. I hope that we can begin to understand the complex relations between Quine and Wittgenstein by comparing their reactions to Carnap.
Carnap recommends that we provide rational reconstructions of areas of our knowledge by constructing axiomatic systems.1 These linguistic frameworks or constructional systems embody a system of logical rules, together with fundamental classifications. There are explicit rules which determine the criteria to be employed in applying the predicates that make up these classificatory systems. Among the true sentences which can be formulated by someone using such a system, a distinction is to be drawn between those which express the rules of the system and others which use those rules to make ‘substantial’ statements. This distinction between L-truths and P-truths corresponds to the more familiar distinction between analytic and synthetic truths. Carnap’s views about the character of these constructional systems and about the reasons for developing them evolved. However, there are several themes that can be noted in general terms without the need to become involved in scholarly details.
There is a clear sense in which such a rational reconstruction is intended to provide an account of a theory or body of knowledge which renders its structure wholly open to view: nothing is hidden and all is perspicuous. This perspicuity is obtained through explicitly listing the primitive vocabulary of the framework and laying down the logical and other linguistic rules which are used in evaluating sentences expressed using this vocabulary. The advantages of constructing such systems are several. First, if the rules which constitute a system of knowledge are made explicit, then all will agree upon the bearing of (say) a piece of evidence upon a theory. Irresoluble differences of opinion will probably turn out to be due to differences in the linguistic rules employed by different inquirers. It promises an ideal of rule-governed rationality which may help to explain the progress of, and inter-subjective agreement secured within, the sciences. Secondly, the construction provides a system of categories, offering a set of fundamental classifications. Moreover, the sorts of categories employed by Kant will be described in a constructional system which we use for describing the structure of linguistic frameworks in general. The rules of our framework specify the form of an area of knowledge; this form provides a background against which genuine assertions can be made.
The linguistic framework/constructional system also promises philosophical insight. First, some apparently deep metaphysical issues are reformulated as apparently innocuous grammatical issues about the structure of linguistic frameworks. If somebody asks, in the ‘material mode’, ‘Are numbers objects?’, they can appear to pose a deep ontological puzzle. Within the theory of linguistic frameworks, this can be reformulated as: ‘Are words for numbers singular terms?’ or ‘Is quantification over numbers first-order quantification?’ A puzzling issue receives a clear sense. Metaphysical illusion results from turning these questions from ‘formal-mode’ questions about forms of language into ‘material-mode’ questions purporting to be about the structure of reality.2 We might suppose that, correctly formulated, our question about numbers should be ‘Will the correct linguistic framework contain singular terms for numbers?’, but Carnap rejects this notion of a ‘correct linguistic framework’. When we debate whether to make use of a linguistic framework, we are not concerned with its truth: acceptance of the framework ‘can only be judged as being more or less expedient, fruitful, conducive to the aim for which the language is intended’.3 In The Logical Syntax of Language Carnap enunciated his ‘principle of tolerance’: ‘In logic there are no morals. Everyone is at liberty to build up his own logic, i.e. his own form of language, if he wishes’4, and he later urged us ‘to be cautious in making assertions and critical in examining them, but let us be tolerant in permitting linguistic forms’. There are no ‘facts’ about which linguistic frameworks should be adopted: this is a pragmatic matter concerning how well they serve our current intellectual purposes. Talk of fact and objectivity makes sense only relative to a linguistic framework.
Thus in his first major book, Carnap sketched two possible constructional systems. One, which he called ‘autopsychological’, attempted to ‘reduce’ physical object language to sentences about subjective experiences.5 The other, which was ‘heteropsychological’, attempted the reverse ‘reduction’. There was no suggestion that one of these was correct, or that one correctly reflected the contents of our thoughts or the structure of the world. Which we should use would depend upon our purposes: the first, he suggested, would be useful for epistemology, while the second would be a better vehicle for psychological research.
One further point of importance. We clarify one area of our practice—for example our attempts to do epistemology—by developing a constructional system, a linguistic framework in which all of the rules employed are carefully and explicitly formulated. There is no suggestion that this framework provides a fully explicit description of our pre-theoretical practice: it does not provide clarification by making explicit what we know implicitly. Rather it offers clarification by providing a linguistic tool which is better than our confused everyday ways of speaking: it achieves all of the purposes that are worth pursuing with the aid of such a framework, but it is clearer and more perspicuous than the area of discourse which it reconstructs. For example, the framework we construct will embody an explicit distinction between L-truths and P-truths. These need not map onto truths which (pre-theoretically) we treat as analytic and synthetic respectively. Indeed, tracing an analytic/ synthetic distinction in our ordinary discourse may be simply impossible because of the vagueness and imperfection of natural languages: all the more reason, Carnap may argue, for reconstructing our knowledge in a form which introduces this invaluable distinction.
In one sense, as we have seen, Carnap’s theory is ‘pluralistic’: he advocates the development of a range of competing and complementary linguistic frameworks so that the strengths and weaknesses of each can be appreciated. But this pluralism belongs within a philosophical outlook which is strongly scientistic: Carnap is ambitious to render philosophy scientific and to provide a philosophical reconstruction of scientific knowledge and the methods we employ in seeking it. He has no interest in the reconstruction of non-scientific areas of discourse. The process of rational reconstruction, and the defence of what we have identified as a form of pluralism, are both internal to science: they are to be justified by the contribution they can make to understanding scientific reasoning and facilitating scientific progress.
So far we have noted some different motivations that Carnap’s constructional systems were designed to meet: epistemological, ontological and anti-metaphysical. It will be important for the ensuing discussion to notice some general characteristics of Carnap’s approach. First, constructional systems are constructions: they are formal systems which are made by human beings and used for a variety of purposes. Consequently these systems can always be held at arm’s length, described and evaluated: we can always sensibly ask questions about how well they serve our purposes. Secondly, they rest upon the assumption that maximal clarity can be achieved by laying down a body of rules. We need to be clear about the logical and grammatical form of the areas of discourse we are concerned with, and we should ideally impose a clear formal structure upon it. In general, Carnap appears to believe that it is both possible and desirable to render all of the norms which govern our processes of belief formation fully explicit. Although holistic pragmatic evaluations have a role in the growth of knowledge, such evaluations are most perspicuously presented as concerned with our choice of systems of rules. If we reconstruct our knowledge in that manner, we shall gain benefits. And thirdly, as is exemplified during the 1930s, when Carnap develops a framework described as ‘general syntax’, philosophical insight is to be gained by constructing a general theory of frameworks: describing the different status of internal and external questions, for example, and drawing general philosophical conclusions from this.
Quine’s attitude towards these Carnapian themes is complex. In Word and Object he describes a canonical notation for science— essentially extensional first-order logic with identity. In giving his motivation for doing so, he hints at a comparison with Kant’s search for a system of categories.6 Moreover, when we express a theory in this canonical notation, we are not revealing a structure that is claimed to be already implicit within it: we are not describing its semantical content. Instead we seek a reformulation which promotes clarity and serves a definite purpose: only by doing this can we make the theory’s ontological commitments explicit and, ultimately, reveal the groundlessness of questions of ontology.
However, Quine’s attitude towards the autopsychological construction of the Logiscbe Aufbau is altogether less sympathetic. His own story of how beliefs about an objective external world are related to sensory inputs is, in many respects, close to the one that Carnap defends. And when he discusses Carnap’s venture in ‘Epistemology Naturalized’, he acknowledges that to achieve such a reconstruction would be ‘a great achievement’. He then continues:
But why all this creative reconstruction, all this make believe? The stimulation of his sensory receptors is all the evidence anybody has had to go on, ultimately, in arriving at his picture of the world. Why not settle for how this construction really proceeds? Why not settle for psychology?
(EN 75)
Notice that Quine does not seem to be questioning the possibility of carrying out Carnap’s project. He is, rather, questioning its point: the suggestion seems to be that we do not need the kind of ultra-clear perspicuous representation of the matter which Carnap’s project offers. And he offers an alternative recipe for becoming clear about our ways of acquiring knowledge: a psychological theory of cognition gives us all we need. For epistemological purposes, clarity does not require an account of linguistic frameworks; it needs only a theory of how our knowledge grows. A naturalized epistemology provides as perspicuous an account of epistemological matters as we could require.
In the following paragraph, Quine notes one circumstance in which it would be useful: if we could show that we could ‘translate science into logic and observation terms and set theory’, we could use our rational reconstruction in order to show that theoretical terms were dispensable, thus solving some fundamental epistemological problems. But if this benefit is not to be achieved— as Quine and Carnap agree it is not—rational reconstruction has no advantages over psychology. Quine’s naturalized epistemology and Carnap’s autopsychological constructional system tell rather similar stories about how objectivity and inter-subjective agreement are based upon sensory stimuli, so the debate concerns the way in which these stories are to be told. Quine thinks that Carnap’s reconstructions are worth having (or are worth the effort that goes into their construction) only if they allow us to eliminate reference to theoretical entities.
Since Carnapian construction differs from Quinean regimentation primarily in its use of a distinction between L-truths and P-truths, it is plausible that Quine’s position is linked to his rejection of the analytic/ synthetic distinction. However, there is an interesting question about the direction of this linkage. Are Carnapian constructions to be dispensed with because they rely upon a flawed distinction between analytic and synthetic statements? Or is rejection of the analytic/ synthetic distinction grounded in the fact that we have no need for Carnapian reconstructions that make use of the distinction? Such questions rarely have clear-cut answers, and Quine’s criticisms of modal logic may suggest the first direction: he thinks that such constructions are internally incoherent or fail to measure up to standards of clarity desirable in a system of logic. However, I think that the best way to understand the argument of ‘Two Dogmas of Empiricism’ (for example) is to focus on the second direction of linkage: we simply have no need for reconstructions of our scientific knowledge which employ this distinction (FLPVa, ch. 2).
A central puzzle concerning the analyticity debate is that the epistemological picture (‘moderate holism’) which Quine obtains by rejecting the analytic/synthetic distinction is apparently endorsed by Carnap himself in The Logical Syntax of Language, a book in which he explicitly uses it. Carnap insists that the empirical testing of a theory ‘applies, at bottom, not to a single hypothesis, but to the whole system of physics as a system of hypotheses (Duhem, Poincaré)’.7 And he insists that: ‘No rule of the physical language is definitive; all rules are laid down with the reservation that they may be altered as soon as it seems expedient to do so.’ This applies not only to P-rules (fundamental physical hypotheses) but also to analytic truths, including principles of logic. Both thinkers agree that confirmation has a holistic character, and both think that analytic propositions are revisable—presumably under pressure of experience.8
It is possible to see how a retention of the analytic/synthetic distinction is compatible with this holistic perspective. Imagine two scientists who have agreed upon a rational reconstruction of their shared body of theoretical knowledge. In the course of doing so, they adopt a system of formal logic and a suitably formulated confirmation theory. They undertake to use this constructional system as a tool in their inquiry. It is easy to see how the decisions they have to make about how to revise their beliefs can be divided into different kinds. In some circumstances, when they face a new experience, their confirmation theory and logic may leave relatively little room for manoeuvre when they ask where adjustments should be made in order to accommodate the new information. In other cases, they may need to make use of criteria of overall coherence and simplicity in choosing between a number of revisions which would be licensed by the rules of their logic and confirmation theory. And in yet further cases, shaken by the growing recalcitrance of the world and the emerging incoherence of their system of ideas, they may appeal to ‘pragmatic’ considerations in questioning their system of L-truths, their confirmation theory or even their logic. The constructional system is of value, then, in enabling them to keep track of these different kinds of decisions: it makes perspicuous the kinds of criteria that should be employed on these different occasions. The constructional system serves a purpose.
This picture is more than a little grotesque: we cannot conceive that inquirers actually could or would employ constructional systems in this way. However, it might still be insisted that it provides a useful ideal type, revealing to us the kinds of error that we risk through falling short of this ideal. This response would have to rest upon believing that rationality requires us to be as reflective as possible, and holding that reflection is guided ultimately by rules. In practice we lack the time or ability to reflect as deeply as we might. And this means that we are not guided by norms or rules as ideal rationality would require; and this is a bad thing.
A rational reconstruction is judged by two standards: we must ask whether it preserves those features of our everyday practice which are worth preserving; and we must evaluate the improvements (in clarity and precision) which it offers. A defence of the analytic/ synthetic distinction must rest upon claiming either that the distinction is nascent or implicit in our ordinary practice, or that the lack of such a distinction is a flaw in this practice. The latter view would have to depend upon the view that rationality requires rules, reflection and explicit representations of our norms and opinions.
How might such a distinction be implicit or nascent in our practice? We might distinguish what we are sure of from those opinions that are more tentative, or we could link the analytic to what is obvious to us. Alternatively we might notice that some of our beliefs seem wholly indubitable: any doubt of them would be an empty paper doubt; we cannot imagine any experience that leads us to question them; they are taken for granted in the ways in which we formulate questions, design experiments or interpret their results. It is undeniable that there are such beliefs, even if there is no sharp distinction between them and others that form part of our corpus. Would there be a case for seeking a reconstruction of our body of opinions which explicitly identifies the opinions which occupy this position? Could this be our nascent analytic/synthetic distinction?
One reason to deny this is that talk of analyticity offers a distinctive kind of explanation of the role occupied by such opinions, and beliefs could occupy the position just described merely on account of their obviousness or familiarity, without being analytic in Carnap’s sense. With philosophers of the common sense school, it could be said of these beliefs simply that ‘everything counts for them and nothing counts against them’.9 Moreover, all such beliefs are unlikely to be general rules or principles of the kind normally thought of as analytic. Presumably for Quine the sentence ‘There is such a place as Boston’ serves this role, but it is not a candidate for analyticity. Moreover, Carnap’s analytic truths are not all as obvious as this: the L-rules of a tentatively adopted constructional system will be analytic in spite of the fact that they are used with no great confidence that they will not soon be abandoned.
In fact, Quine does not actually reject the analytic/synthetic distinction (see RR 78ff., and I 503–6). He admits that something like the traditional notion of analyticity can be captured by reference to truths which are learnt as part of learning the words that they contain. This explains the analyticity of ‘All bachelors are unmarried’ and of much of logic: one could doubt or reject such claims only by changing their meanings. By this criterion, the existence of Boston may also be an analytic matter. And, presumably, he could also allow that some truths about (say) protons or electrons will have to be acknowledged if one is to belong to a community that can carry out inquiries into theories using such notions. What he rejects is the claim that this is of any philosophical interest: we learn little about scientific rationality, about how theories are based upon experience, about the differences between mathematics and empirical science, by offering general explanations using this distinction. There is no philosophical benefit in trying to list or describe analytic truths, or in specifying ‘framework principles’. It is of no importance for science.
A second way to deny this importance is simply to question the point of labelling such beliefs, of giving them a special kind of status: we never need to appeal to this in order to explain anything. Consider Quine’s account of how we do (and presumably should) revise our opinions. It accords with his empiricism that we seek theories that accord with experience; when confronted with a surprising experience, we are to favour the least change that will accommodate the new experience and remove the appearance of contradiction; and we seek simplicity: scientific method is ‘a matter of being guided by sensory stimuli, a taste for simplicity in some sense and a taste for old things’ (WO 23). The standards of simplicity that guide our inductions are ‘implicit in unconscious steps [of inference] as well as half explicit in conscious ones’ (ibid., 20); but ‘this supposed quality of simplicity is more easily sensed than described’ (ibid., 19). Quine’s emphasis on ‘the looseness’ of the ideal of simplicity, and on the passiveness of our judgements of evidence and simplicity, suggests that he doubts whether this ideal could usefully be encapsulated in rules and principles. Evolution has equipped us to be reasonably good at induction, and we are unwise to push reflection and the search for principles too far. Epistemic reflection is generally shallow: where reflection has a role, the only relevant epistemic standards are provided by empirical psychology.10 If we do best to trust our epistemic endowment and not to demand explanations of why each application of our sense of simplicity is legitimate, then we have no need for the complex apparatus of rules and principles offered by Carnapian constructional systems11. We do better to trust our sense of simplicity and to weigh evidence in a passive manner than to seek the depths of reflection that Carnap purports to offer.
There is clear evidence that Quine believes that we are guided by implicit or tacit norms which we cannot, and need not try to, make explicit. As we have noted, he insists in the first chapter of Word and Object that we are guided by judgements of simplicity in deciding how to revise theories in the light of new experiences. He is doubtful that we can formulate principles of simplicity or a formal system of inductive logic which will describe the basis of our practice and enable us to carry it out under the guidance of explicit norms. Normative standards that are grounded in our genetic inheritance or, presumably, in our scientific training are operative in the growth of scientific knowledge, although they are not explicit. Moreover, the anti-conventionalist argument of ‘Truth by Convention’ shows that it would not be possible to make explicit all of the norms or rules that we rely upon (TCb 77ff.). Explicit rules are general formulations which have to be applied to particular cases. In applying them, we must use rules and norms specifying how they should be applied. If these too must be explicit, we embark on a regress of rules. Hence some norms must be manifested in the ways that we interpret and apply explicit rules and norms without themselves being explicit.
For Quine, we are always in the middle of things, trying to answer questions against a background partly constituted by our evolving body of theory. We cannot stand back and assess our beliefs as a whole, measuring them against some transcendental measure of truth or rationality. The understanding of theory and language that we rely upon in trying to make sense of our practice is itself part of science.Like Carnap’s, Quine’s conception of philosophy involves a double reference to science: philosophy is itself part of science; and its chief aim is to make sense of our scientific knowledge. One of his differences with Carnap lies in his doubt that we can best clarify our ideas about how theory relates to evidence by developing constructional systems that embody systems of precise rules. The assessment of theory and evidence is generally ‘passive’, and there is no reason to suppose that the attempt to make norms explicit in the form of rules and definitions is likely to be successful or useful. If there are ‘analytic truths’, this is not interesting. There is no reason to suppose that we can uncover an underlying ‘form’ which specifies the logic of a theory. Tools for clarification, systems of canonical clarification, are constructed for distinctive purposes as part of our total science.
There are puzzling similarities between views found in Wittgenstein’s later writings and the views of both Quine and Carnap. The discussion of rule following in the Philosophical Investigations has many points of contact with Quine’s claim that it is impossible that all norms or rules should be explicitly formulated: following a rule is, at root, a practice. They share, too, the view that the search for philosophical analyses (lists of analytic necessary and sufficient conditions for the applications of concepts) is misplaced. The Quinean suggestions about ‘analyticity’ mentioned at the end of the previous section may also find echoes in Wittgenstein’s work, and they could share the view that if there are ‘analyticities’, these will include singular propositions which appear to be ‘empirical’ (for example, ‘This is a hand’) as well as the general formulations employed by Carnap. But Wittgenstein’s attempt to describe our varied linguistic practices and to identify grammatical propositions reflects an approach to philosophical issues radically at variance with Quine’s. Given the resemblances I have just noted, why does Wittgenstein find it important or useful to identify sentences which have this distinctive kind of status?
The latter suggests similarities between Wittgenstein and Carnap; and we might suppose that the former’s emphasis on the variety of our linguistic practices bears similarities to Carnap’s pluralistic insistence that we should be liberal in allowing many linguistic frameworks to bloom. But this impression fades when we notice Wittgenstein’s determination to describe our practices rather than replace them with finely honed rational reconstructions. The goal of his descriptions appears to be the avoidance of philosophical error rather than the encouragement of scientific rationality. However, in other respects there are similarities. Wittgenstein’s account of necessity shares a conventionalist flavour with Carnap’s, although their versions of conventionalism are very different. And consider some passages from On Certainty. Wittgenstein emphasizes the special status of those propositions which form ‘the scaffolding’ (§211) of our inquiries. He notes that ‘the questions that we raise and our doubts depend on the fact that some propositions are exempt from doubt, are as it were like hinges on which those turn’, and he emphasizes that their possessing this status ‘belongs to the logic of our scientific investigations’ (§§341– 2). This all suggests that philosophers can try to understand the form of an area of our knowledge, identifying propositions with a distinctive ‘rule-like’ status. Even if his claim that these ‘propositions’ are ‘in deed not doubted’ accords with the idea that they express norms which are implicit in our practice, and which we do not need to formulate explicitly, it is hard to deny his un-Quinean adherence to a rough dualism of propositions which might be supposed to encourage a Carnapian direction.
One theme in On Certainty indicates that this would be wrong. The attempt to make these propositions explicit, to formulate them and inquire about their status, is seen as a temptation to philosophical error: if we ask how we know them, we shall never block scepticism; and what belongs to the logic of science is the very fact that they are ‘in deed not doubted’; ‘It is our acting which lies at the bottom of the language game’ (§204). If it is philosophically important to identify some propositions of this sort and to note their special status, this is not because we thereby approach some ideal of reflective rationality: it will not make us better scientists. Reflection is shallow: rationality does not even require us to notice these propositions. Here Wittgenstein is closer to Quine than to Carnap. But that only reinforces the question of why such propositions are philosophically important to the former but not to the latter.
For Wittgenstein, the systems of concepts we employ reflect our interests and concerns and general facts about the nature of our environment. Considering our colour vocabulary, he remarks that ‘one is tempted to justify rules of grammar by sentences like “But there really are four primary colours”‘. He continues:
We have a colour system as we have a number system. Do the systems reside in our nature or in the nature of things? How are we to put it? Not in the nature of numbers or colours.
Then there is something arbitrary about the system? Yes and No. It is akin both to what is arbitrary and to what is not arbitrary.
(Z §§357–58)
They are arbitrary in the sense that ‘the world’ does not require a unique colour system. They are non-arbitrary in that the system we use is wholly natural given our concerns and the general facts about the world in which we live.12 Compare Quine: ‘The lore of our fathers…is a pale gray lore, black with fact and white with convention. But I have found no substantial reasons for concluding that there are any quite black threads in it, or any white ones’ (CLTb 132).
The explanation of these differences lies in Wittgenstein’s attitude towards science. As Stephen Hilmy has emphasized, Carnap’s search for a scientific philosophy stands in marked contrast to Wittgenstein’s disdainful repudiation of such an activity.13 Where Carnap insisted that all of the ‘emotional needs’ that fuel our philosophical endeavours can be met by ‘clarity of concepts, precision of methods, responsible theses, achievement through cooperation in which each individual plays his part’,14 Wittgenstein announced that ‘I am not aiming at the same target as the scientists and my way of thinking is different from theirs.’15 Wittgenstein saw philosophical and metaphysical puzzlement as stemming from intellectual needs that could not be addressed by a scientific philosophy. He saw a need to provide descriptions of our practices which would enable us to come to terms with those demands without allowing them to be transformed into a search for a philosophical or pseudo-scientific theory. Carnap’s ‘pluralism’ involved attaching value to the development of a range of alternative scientific frameworks; Wittgenstein, by contrast, was open to the variety of non-scientific outlooks forming part of our complex cultural response to our surroundings. In spite of his many differences with Carnap, Quine never doubts that philosophy should be scientific in its approach, and his philosophical interest has always been in making sense of science.
This difference is reflected in contrasting attitudes towards the philosophical tradition. The legacy of the philosophical tradition is complex, and it contains features of at least two distinguishable kinds. Consider, for example, Descartes’s Meditations. We may derive from our reading of it a sense of the need to reflect upon the sources of knowledge, the character of the inferences we employ and the possible sources of error and illusion in our opinions, if we are to make secure progress in our search for knowledge. If this is all that we derive from it, then it will remain an open question whether we should try to execute this project in the same way as Descartes himself. For example, we may not share his view that confronting the most radical sceptical arguments is the best way to do this. Reflecting on the ways in which errors are uncovered as scientific inquiry progresses, and confident in the ability of a community of fallible inquirers to make steady progress, we may dismiss Descartes’s strategy as neurotic and flawed, as placing obstacles in the way of scientific progress which should be simply swept aside. In this spirit, we may see an exponent of naturalized epistemology acknowledging the continuities of his work with the earlier epistemological tradition while impatient of anyone who feels an obligation to engage seriously with the details of Descartes’s sceptical arguments, his proof that the mind is better known than the body or the intricacies of the Cartesian Circle. These issues, it will be claimed, arise out of the details of Descartes’s own flawed response to more general issues that are better addressed in other ways. And the best defence of this may be simply to address those issues in a naturalistic spirit without engaging with these other possibilities: an ‘anti-Cartesian’ approach to epistemology can be vindicated by its success in providing for the possibility of reflective and responsible belief formation. The search for a distinctively philosophical foundation for scientific knowledge then seems an unessential part of this tradition.
A second response to the philosophical tradition finds in it a body of distinctively philosophical arguments which inexorably draw our thought in unappealing directions which resist easy refutation. The reader of Descartes may be drawn into philosophical reflection by the apparent impossibility of showing that one is not dreaming or by the problems presented by the philosophical arguments for the real distinction between mind and body. A sensitivity to (and a susceptibility to the force of) these arguments can then control what seems to be required of any serious epistemological work. Philosophy then becomes a permanent possibility for us because of the challenges presented by these distinctively philosophical problems and arguments.
Carnap’s work suggests both of these responses. He provides an account of rationality which is supposed to facilitate responsible and reflective belief formation. And the rational reconstructions he offers enable him to provide a diagnosis of the source of these distinctively philosophical problems and a way of dismissing them. Sceptical problems arise, for example, because we treat external questions (questions about the choice of framework) as questions of fact. And apparently fundamental ontological issues all emerge from our misidentifying questions about the formal structure of a linguistic framework as material questions about the nature of reality. Constructional systems provide us with representations of theories and reality which help us to avoid those problems: Carnap appears to be sensitive to the force of these distinctively philosophical problems and anxious to disarm them. Quine, by contrast, seems to be simply impatient of these traditional concerns. If he is aware that we are tempted down unappealing philosophical pathways, he believes that an adequate scientific understanding of language and science is all that is required to restore us to the straight and narrow. It is no part of his philosophical endeavour to engage with them, appreciating their force and undermining it. When they are referred to, they are diagnosed as overreactions to scientific information about perceptual illusion and error. This aspect of the legacy of the philosophical tradition does not appear to present him with any challenge. The representations of our beliefs provided by his canonical notation and offered by his search for a naturalized epistemology have nothing to contribute to this kind of wrestling with philosophical demons.
If one is open to the beguiling nature of traditional philosophical arguments and problems, and if one thinks that these result from kinds of intellectual bewitchment produced by our misunderstandings of how our language functions, then we might have good reasons for describing how our language does in fact function: when we recognize that a ‘proposition’ is functioning as a ‘rule’ rather than like the empirical proposition it appears to be, this is a matter of great philosophical importance. If we think that the attempt to approach all subject matters in a scientific spirit is itself one cause of (or symptom of) such misunderstandings, then we have good reasons for paying close attention to the differing rules and logical patterns exhibited by our reasonings in the different areas. A clear description of the norms that guide us becomes essential to philosophical progress.
We have noticed two differences between Quine and Wittgenstein, and we shall conclude with a brief speculation about their relations. The primacy of scientific knowledge drives Quine’s philosophy, leaving him with few grounds for doubt that the understanding to be obtained from science will suffice to dispose of the sources of philosophical bewitchment. What I have called Wittgenstein’s pluralism, his resistance to this idea that serious philosophical problems are scientific ones, prevents his attempting to combat philosophical puzzlement in this way. Detailed description of our linguistic practices, identifying what functions as scaffolding and what is unsupported, becomes essential for distinguishing those practices which are of value from those which manifest the bewitchment and confusion which are characteristic of much traditional philosophy.
1 Carnap’s views developed from the 1920s until the end of his life, but these developments are not relevant to the points I wish to make. I shall refer to three texts: The Logical Structure of the World, tr. R.George (University of California Press, Berkeley and Los Angeles, 1967)—this is a translation of Der Logische Aufbau der Welt, first published in 1928; The Logical Syntax of Language (Routledge and Kegan Paul, London, 1937); and Meaning and Necessity, 2nd edn (University of Chicago Press, Chicago, 1956). The ideas about rational reconstruction and constructional systems discussed here are found in The Logical Structure of the World.
2 This theme becomes prominent in The Logical Syntax of Language.
3 Meaning and Necessity, p. 214.
4 The Logical Syntax of Language, p. 52.
5 The Logical Structure of the World, section 58.
6 ‘The quest of the simplest clearest overall pattern of canonical notation is not to be distinguished from a quest of ultimate categories, a limning of the most general traits of reality’ (WO 161).
7 The Logical Syntax of Language, p. 318.
8 In Quine: Language, Experience and Reality (Polity Press, Oxford, 1988), I suggested that it was a mystery that Carnap did not see that this admission undermined his distinction (p. 37). The present discussion allows that the position is more complex than I there admitted.
9 As we shall see later, this would not be an adequate account of Quine’s own position. It suggests that the analytic/synthetic distinction is to be abandoned because all propositions are synthetic: these ‘obvious’ truths are synthetic, but we can only gesture towards the mass of evidential support they receive. We cannot articulate it. It is difficult to guess what Quine would say of this, but we should recall Quine’s insistence that conventional decision does have a role in the formation of our system of opinions:
The lore of our fathers is a fabric of sentences. In our hands it develops and changes, through more or less arbitrary and deliberate revisions and additions of our own, more or less directly occasioned by the continuing stimulation of our sense organs. It is a pale gray lore, black with fact and white with convention. But I have found no substantial reasons for concluding that there are any quite black threads in it, or any white ones.
(CLTb 132)
Note that Wittgenstein was equally unsympathetic to these ‘commonsense’ formulations.
10 I have discussed Quine’s views about epistemic evaluation and the limits of epistemic reflection in ‘Naturalized Epistemology and Epistemic Evaluation’, Inquiry, 37 (1994), 465–85. See pp. 476–9. The parallels between the views of the relation of evidence to knowledge found in Quine’s naturalized epistemology and in Carnap’s auto psychological construction are explicitly signalled in the first two chapters of Quine’s From Stimulus to Science (Harvard University Press, Cambridge, Mass., 1995).
11 Although Quine is happy to appeal to evolutionary considerations to explain the reliability of (for example) our inductive habits. See, for examples, I 503.
12 I have borrowed a sentence or two here from ‘Wittgenstein and Knowledge: Beyond Form and Content’, Journal of Speculative Philosophy, 7 (1993), 77–91, p. 86. This paper contains a more extended discussion of related themes in Wittgenstein’s thought.
13 Stephen Hilmy, The Later Wittgenstein (Blackwell, Oxford, 1987).
14 Discussed by Hilmy, ibid., pp. 213f. The quotation is from Carnap’s The Logical Structure of the World, p. xv.
15 Cited by Hilmy, op. cit., p. 191.