II
These considerations about the ambiguity inherent in the theoretical introduction of syntactic rules did not figure explicitly in the young Quine’s first attempts to elaborate the procedures of logic, devoted as these were to a largely sympathetic exposition of Carnap’s syntax project. But they are nevertheless central to the minor inflectional differences that would already distinguish these first attempts from Carnap’s descriptions of the methods of syntax. The early article “Ontological Remarks on the Propositional Calculus,” published in 1934 (the same year asSyntax
), already bears witness to some slight, but significant, differences in conception between Quine and his teacher. The article poses the question of how best to construe the subject matter of the logician’s symbolic, propositional calculus. Should the formulas and sentences of logic be taken to stand for extra-logical items, perhaps facts or states of affairs, or (following Frege) the truth-values True and False? Each of these solutions, Quine suggests, invites problematic metaphysical speculations. We do better, if we can, to construe the functioning of the propositional calculus without countenancing such “inferred entities” that would take us “beyond the realm of everyday uses of words.”
Accordingly, Quine outlines two distinct strategies for construing the reference of the sentences and formulas of the propositional calculus without invoking propositions. First, we may take the special truth-functional symbols of the propositional calculus simply to beabbreviations
of ordinary English words and phrases. Thus, for instance, the special symbol ‘~’ can be construed as a definitional abbreviation for ‘not’ or ‘it is false that:’, and the other truth-functional signs conceived similarly. The approach has the desired effect of eliminating the suggestion of any special subject-matter for the logical calculus. But as Quine notes, it also means that the propositional calculus “ceases to be a system in the usual sense.”
For if the truth-functional connectives and variable signs are simply abbreviations for natural-language terms and sentences, the propositional calculus is itself no longer a system of actually existing elements subject to specific operations, but just a paradigm showing the use of these ordinary terms and sentences. The formation and derivation rules can help to show under what circumstances certain of these sentences are true - in particular, they show us more clearly which ordinary propositions can be consideredlogical
truths - but beyond this, they have no distinct denotational objects of their own.
As Quine suggests, a second way to construe the significance of the propositional calculus without countenancing propositions is simply to construe the variable symbols of the calculus as denotations ofsentences
, grammatically well-formed sequences of symbols. This is essentially Carnap’s solution inLogical
Syntax
, and with it the propositional calculus again becomes a system of rules constraining the legitimate manipulation of elements, the sentences of the ordinary language. The truth-functional connectives now become signs denoting sententialoperations
, for instance the operation of appending “not:” before a sentence or concatenating two sentences and interposing the word “or”. As Quine observes, on this second solution, the symbolic formulasof
the propositional calculus now become, themselves, symbolsabout
sentences, in particular variables which ambiguously stand for any ordinary-language sentence of a certain logical form. The theorems of the system then become, themselves, assertions to the effect that the sentences they denote are true, and the turnstile symbol ‘├’, previously used simply as an informal tag for theoremhood, must now be construed as apredicate
asserting the truthfulness of the sentences ambiguously denoted by the formula that follows it.
Both of these suggested Quinean solutions to the problem of the nature of the propositional calculus share the strongly anti-metaphysical attitude of Carnap’sSyntax
project in their staunch avoidance of propositional entities beyond actual sentences themselves. But it is significant that both Quinean solutions, in construing the propositional calculus as involving nothing more than actual sentences, construe the formational and inferentialrules
of the symbolic calculus as systematically dependent upon the actual patterns of sententialuse
evident in ordinary linguistic practice. For Quine, there is nothing beyond such patterns for the symbols of the propositional calculus to beabout
. Gone, already, is any suggestion of the logician’scomplete
freedom in creatingarbitrary
symbolic calculi. For Quine, even the possibility of interpreting the transformation rules as rules ofinference
requires some reference to the antecedently understood significance of inference in an already-understood language. Similarly, even identifying a sentence in the calculus as a postulate or a logical truth means asserting the truthfulness of a whole class of actual object-language sentences with a certain form. This intrinsic dependence on the antecedently more-or-less understood notions of inference, derivability, and truth cannot be eliminated completely, even if the syntactical procedure may be thought to sharpen and clarify these notions somewhat.
This Quinean appeal to antecedent use in the articulation of syntactic rules develops further in his subsequent reckonings with the legacy of Carnap’s project. In his 1934 “Lectures on Carnap” delivered at Harvard, Quine summarizedLogical Syntax
, presenting its main results to a non-specialist audience. But although the second and third lectures are wholly devoted to exegesis, in the first lecture Quine introduces Carnap’s notion of analyticity by describing an original semantic procedure that can be followed in order to arrive at clear definitions of terms, and in order to determine the range of sentences that are analytic in a given language. To carry out the procedure for any given term, we begin by considering the set of all the sentences involving that term that aretrue
in the language, or accepted on a commonsensical level by its speakers. Now, if we can lay down definitions that indeed make all such sentences true in each case, we will have arrived at an accurate definition of the term and, more generally, at a set of definitional conventions that expose the actual logical structure of the language:
Now suppose we are confronted with the job of defining K. If we can frame a definition which fulfills all the accepted K-sentences, then obviously we shall have done a perfectly satisfactory job. Nobody who was inclined to dispute the definition could point to a single respect in which the definition diverged from the accepted usage of the word K; for all accepted K-sentences would be verified.
Were there only a relatively small number of sentences, for any given term, that both involve that term and are accepted by the speakers of the language, the definition would be easily accomplished, simply by listing the true sentences and proposing that the term should be used in just those ways and no others. But because there are, in any actual language, an infinite number of sentences including any given term, it is in general impossible to define terms in this finitary way. Rather, explicit definitionalrules
must be introduced for each particular term to subsume, as much as possible, the infinite number of true sentences involving it. Since each sentence involves more than one term, framing the rules requires making determinations as to whether a particular term appears in a context more or lessmaterially
. For instance, the term “apple” appears materially in “Every apple weighs at least two grams,” but does not do so in the sentence, “Within any class of two apples there is at least one apple,” since it may be replaced, in the latter sentence but not the former, with any other substantial term.
In framing definitional rules for the language as a whole, we are likely to begin with rules for terms, such as mathematical ones, that tend to appear in many contexts non-materially or vacuously; but since no termalways
appears vacuously, our definitional procedure will always involve making decisions of relative priority. The result is a system of rules that determines certain sentences as analytic, or true by definition. But because of the inherent arbitrariness of the determination of priority, the extent of the set of sentences deemed analytic will itself be, to a certain extent, arbitrary. In the limiting case (as Carnap had indeed already suggested),all
of the currently accepted sentences of the language, in fact, could be rendered analytic, simply by framing the rules in such a way as to make them all come out true. But in actual practice, the decision of the best systematization for the language as a whole will presumably be guided by considerations of overall, systematic simplicity, while also aiming to respect our ordinary, intuitive notion of the distinction between formal or logical and empirical truth.
The “Lectures” therefore exhibit, as yet, no significant disagreement with Carnap over the extent and significance of the analytic/synthetic distinction for a given natural language. As for Carnap, on Quine’s procedure the determination of the set of sentences that are analytic depends on the conventional introduction of explicit, syntactical rules. And because there is some degree of arbitrariness in framing these rules, the question of whether any given sentence is analytic or synthetic does not have a completely determinate answer. But the suggested procedure of framing the definitional rules for a term by reference to the set of accepted sentences involving that term has no direct correlate in Carnap’s suggested procedure. For Carnap inSyntax
, after all, the introduction of syntactical rules is a wholly arbitrary stipulation, having no essential reference to or dependence on the set of sentences that are actually considered true or accepted in any antecedently existing language. Even when the introduction of rules is supposed to capture, in some intuitive sense, the actual logic of an existing natural language, Carnap makes no provision for this introduction to depend on reasoning about the range of sentences already accepted or considered true. For Quine, by contrast, the introduction of particular syntactic rules is already always legitimatedonly
by their ability to capture antecedent usage in the language. The rules can only purport to be syntactic rules at all, insofar as they can claim to capture the patterns of antecedent usage with reference to which they will, pragmatically, be introduced.
A year later, in 1935, Quine re-formulated the material of the 1934 lectures and added some further speculations about logical truth in the influential article “Truth by Convention.” The article, again, offers no outright challenge to what Quine here calls the “linguistic doctrine” of logical truths as true by convention. But it does argue that there is no motivated way, in schematizing a language, to demarcate truths that are intuitively logical or mathematical in character from those that are intuitively empirical, in such a way as to ensure that truths in the first class are analytic and those in the second, synthetic. Quine begins the article by rehearsing the procedure introduced in the lectures for formulating the definitional rules for a language by considering the range of true statements involving a particular term. On this procedure, the introduction of a new symbol into the calculus always amounts to a definitionalabbreviation
for some antecedently understood term or phrase, in conformity with its already-understood traditional usage:
To be satisfactory in this sense a definition of the sign not only must fulfill the formal requirement of unambiguous eliminability, but must also conform to the traditional usage in question. For such conformity it is necessary and sufficient that every context of the sign which was true and every context which was false under traditional usage be construed by the definition as an abbreviation of some other statement which is correspondingly true or false under the established meanings of its signs.
Here, Quine clearly holds, even more explicitly than he had in the earlier lectures, that definitional rules can do no more than to summarize antecedently existing traditional usage. In addition, he explicitlydenies
that the introduction of such rules can be considered to be the result of a purely arbitrary and free decision. Even if Quine’s method at this point does not demand any specific doctrinal break with the system of
Logical Syntax
, the methodological divergence from Carnap’s approach is therefore already substantial. Quine has no interest in, nor even any ability to make sense of, Carnap’s general constructional method, with its associated maxim of tolerance and arbitrariness in language-system creation. Instead, he insists that the inferred or derived rules, even for an artificially constructed language, can have significance only by reference to its already-understood practice.
At the end of the article, Quine poses another, even deeper problem for the “linguistic doctrine” according to which logical and mathematical truths are rendered true by convention. The problem, one of infinite regress, derives originally from Lewis Carroll, who had introduced it in the form of a dialogue between Achilles and the tortoise.
On the conventionalist doctrine, in any actual language, Quine argues, there will be aninfinite
number of statements that we may take to be logically or analytically or conventionally true. It follows that any conventional introduction of them must rely on the introduction of afinite
set of rules or paradigms that are considered to govern an infinite number of instances. Quine in fact considers, in some detail, how the tautological formulas of the propositional calculus might actually be introduced as logically true through one such set of conventions. Each of these paradigms is taken to assert the logical truth of the infinite number of particular sentences of a certain form; their adoption corresponds directly to the fixation of basic, syntactical rules for the language, as described by Carnap. The difficulty, though, is that the application of these paradigms, constitutive of logic, to generate any of the infinite number of particular sentencesitself
depends on the very conventions of logic that they are supposed to formulate. The doctrine of the conventionality of logic is then rendered circular; or, if the introduction of the basic conventions is construed as giving meaning to the primitive logical signs, this meaning is rendered incommunicable:
In a word, the difficulty is that if logic is to proceed mediately from conventions, logic is needed for inferring logic from the conventions. Alternatively, the difficulty which appears thus as a self-presupposition of doctrine can be framed as turning upon a self-presupposition of primitives. It is supposed that the if-idiom, the not-idiom, the every-idiom, and so on, mean nothing to us initially, and that we adopt the conventions … by way of circumscribing their meaning; and the difficulty is that communication of [these conventions] themselves depends upon free use of those very idioms which we are attempting to circumscribe, and can succeed only if we are already conversant with the idioms.
The problem becomes evident as soon as the rules or paradigms of logic are taken to provide information about the derivation or inference of true statements from other true statements. For instance, one of the rules that we may take to be definitive of the material conditional states that,if
we substitute any true sentence for “p” and for “pÉq”,then
the sentence substituted for “q” is true. But the application of this rule to any particular triad of sentences, say “a”, “aÉb”, and “b”, then itself depends on the use of the material conditional. In a similar manner, the application of any of the general rules of logic to particular cases itself depends on the rules themselves. As Quine concludes, there is no hope of taking the rules simply to be conventionally introduced, without relying on any prior understanding or basis, all at once.
In its implications for a general understanding of the basis of meaningful language, the Carroll infinite-regress problem cuts deeper than any objection Quine had hitherto formulated to Carnap’sSyntax
project. The earlier objections, both in the “Lectures” and in the first sections of the “Truth by Convention” article, had established the arbitrariness of any particular circumscription of the rules underlying the practice of a language to include, as analytic, only “logical” and “mathematical” truths. So far as this goes, however, it would still be reasonable to suppose that thereare
such rules, implicit in practice even if not non-arbitrarily capable of explicitation, and actually operative in governing the practice of inference and reasoning for both “logico-mathematical” and “empirical” propositions. The Carroll infinite-regress objection, though, challenges the coherence even of this, more cautious, supposition. If the logical rules governing the practice of a language cannot even be made explicit without circularity, the significance of supposing them to have beenimplicit
all along, in the practice of the language itself, begins to lapse. For any other set of rules, themselves introduced circularly, might enjoy an equal claim to represent the actual logic of the language, provided that they, too, are consistent with the facts of antecedent usage. Quine draws the conclusion near the end of the article:
It may be held that we can adopt conventions through behavior, without first announcing them in words; and that we can return and formulate our conventions verbally afterward, if we choose, when a full language is at our disposal. It may be held that the verbal formulation of conventions is no more a prerequisite of the adoption of the conventions than the writing of a grammar is a prerequisite of speech; that explicit exposition of conventions is merely one of many important uses of a completed language. … It must be conceded that this account accords well with what we actually do. We discourse without first phrasing the conventions; afterwards, in writings such as this, we formulate them to fit our behavior. On the other hand it is not clear wherein an adoption of the conventions, antecedently to their formulation, consists; such behavior is difficult to distinguish from that in which conventions are disregarded. When we first agree to understand ‘Cambridge’ as referring to Cambridge in England, failing a suffix to the contrary, and then discourse accordingly, the role of linguistic convention is intelligible; but when a convention is incapable of being communicated until after its adoption, its role is not so clear. In dropping the attributes of deliberateness and explicitness from the notion of linguistic convention we risk depriving the latter of any explanatory force and reducing it to an idle label.
The point, though cautiously formulated here, is a general and decisive one. The character of a language as a rule-based calculus of signs, and the consequent distinction between uses of the language that accord, and those that fail to accord, with the rules, is not evident prior to the formulation of these rules themselves. But since this formulation is more or less arbitrary within the confines of what we actually say, it cannot claim to represent any unique determination of the actual underlying logic of the language under consideration. Nor can the specification of rules claim to offer new criteria, above and beyond those we have already formulated, for the logical correctness or legitimacy of particular inferences. As Quine would begin to realize more and more clearly, the facts of what we actually utter and do are all that is available to philosophical summary or reconstruction. Beyond these facts themselves, the actual form of the “rules underlying the language” must be taken to be either arbitrarily stipulated at the moment of reconstruction or be considered to be, antecedently to this moment, substantially indeterminate.
III
Already in 1934, therefore, Quine’s consideration of what is involved in understanding an existing language had led him to a conception of syntactical investigation that diverged sharply from Carnap’s constructivist treatment of languages as uninterpreted calculi. The introduction of specialized notation, whether conceived as constituting an autonomous language or simply as explicating the underlying logic of an existing one, could not, for Quine, help but depend on our antecedent grasp of ordinary patterns of usage characteristic of the language we already speak. Indeed, in introducing the Carroll problem, Quine had suggested some reason to doubt that the practice of a natural language can legitimately be treated as determined by a unique underlying set of rules at all.
Quine probably did not yet perceive the depth of the challenge this represented to Carnap’s understanding of languages as calculi. The decisive break would come sixteen years later, in Quine’s 1950 address at the Eastern Division of the American Philosophical Association.
In “Two Dogmas of Empiricism,” Quine argued for the untenability of the analytic/synthetic distinction and of the verificationist dogma of “reductionism” that he thought depended on it. The article is notorious.
Its thematic center is an accusation of circularity, directed at Carnap’s suggested procedure of determining analyticity by explicitly specifying semantic rules constitutive of a language. Over the period from 1934 to 1950, Quine had gained the courage to make this attack explicit; and he had realized that by questioning the motivation of a stipulative determination of analyticity he could also call into question the coherence of the notions of necessity, intensionality, and even synonymy or sameness of meaning, which, he now realized, are interdefinable with analyticity, if they are definable at all. Any of these notions might have a clear significance, if analyticity itself does. But according to Quine, the natural strategy of demarcating the class of analytic sentences in any language by specifying semantical rules is itself empty. This is the case, Quine argues, not only for natural languages, where the underlying rules themselves might be thought to be vague and inexplicit, but even for the artificial languages that Carnap clearly had primarily in mind.
It is, of course, possible,given
any selection of sentences as analytic, to specify semantical rules that determine those sentences, and just those sentences, as analytic. But this specification provides no more information, above and beyond that already present in the selection of sentences already made. In the case of an artificial language, where analyticity is already determinate, the specification of rules underlying this determinacy is empty. In the case of an existing natural language, on the other hand, the selection of a particular range of sentences as “analytic”, as a subset of those generally accepted as true, is arbitrary, and cannot be rendered non-arbitrary by the subsequent or concomitant provision of explicit rules. The explicitation of rules, whether conceived of as constitutive of a fully-formed artificial language or simply as an aid to the comprehension of an existing language, cannot determine what is, in the actual practice of speech, undetermined.
The appeal to pre-existing use that was already decisive, as we saw, in 1934, is explicit at various points in “Two Dogmas.” Quine makes it, for instance, in the course of rejecting the interdefinability of constituent terms as a criterion for the analyticity of a sentence:
There are those who find it soothing to say that the analytic statements of the second class reduce to those of the first class, the logical truths, by definition; ‘bachelor’, for example, is defined as ‘unmarried man’. But how do we find that ‘bachelor’ is defined as ‘unmarried man’? Who defined it thus, and when? Are we to appeal to the nearest dictionary, and accept the lexicographer’s formulation as law? Clearly this would be to put the cart before the horse. The lexicographer is an empirical scientist, whose business is the recording of antecedent facts; and if he glosses ‘bachelor’ as ‘unmarried man’ it is because of his belief that there is a relation of synonymy between those forms, implicit in general or preferred usage prior to his own work.
We have seen that, with his formulation of the Carroll problem, Quine had already suggested in 1934 that this appeal to antecedent use, indeed, tends to rule out any conception of the practice of a language as embodying any determinate set of syntactic or semantic rules at all, implicit or explicit. This point goes even further than the rejection of analyticity itself. For it implies not only that there can be no non-arbitrary sorting, by means of rules, of currently accepted sentences into analytic and synthetic but even that, more generally, the patterns of use characteristic of the acceptance and rejection of sentences in a language cannot be given any unique, explicit formulation in terms of rules at all. Nevertheless, in the period between “Two Dogmas” and his formulation of the indeterminacy result in 1960, Quine would make this second, stronger claim more and more explicitly. In 1954, Quine developed the argument of “Two Dogmas” more specifically, and brought it to bear more directly against Carnap, in “Carnap and Logical Truth”. Here, he directly addresses, for the first time, Carnap’s suggestion that the free propounding of an artificial language is analogous, in the sense in which it amounts to a determination of conventional rules, to the symbolic interpretation or regimentation of a natural language. The analogy, Quine maintains, fails. For the interpretation of an existing language by means of a set of rules is always, at least in part, aprojection
of the interpreter’s assumptions rather than a neutral determination of the real structure of the language under interpretation. We can see this, Quine argues, by considering the possibility of interpreting an alien language, one initially quite unfamiliar to us. He considers the case of an imaginary logical positivist, Ixmann, who wants to clarify the logic of science by developing an artificial language purged of metaphysical claims:
Ixmann’s answer consists in showing in detail how people (on Mars, say) might speak a language quite adequate to all our science but, unlike our language, incapable of expressing the alleged metaphysical issues … Now how does our hypothetical Ixmann specify that doubly hypothetical language? By telling us, at least to the extent needed for his argument, what these Martians are to be imagined as uttering and what they are thereby to be understood to mean. Here is Carnap’s familiar duality of formation rules and transformation rules (or meaning postulates), as rules of language. But these rules are part only of Ixmann’s narrative machinery, not part of what he is portraying… The threat of fallacy lurks in the fact that Ixmann’s rules are indeed arbitrary fiats, as is his whole Martian parable. The fallacy consists in confusing levels, projecting the conventional character of the rules into the story, and so misconstruing Ixmann’s parable as attributing truth legislation to his hypothetical Martians.
With this, Quine’s rejection of Carnap’s conventionalism about the formulation of languages is complete, and the appeal to antecedent usage that this rejection depends on is fully and explicitly formulated. The introduction of a corpus of rules, even in Carnap’s ideal case of the postulation of a wholly new language meant to show the emptiness of metaphysical questions concerning existence, can itself only be conceived as a projection onto the existing language under consideration. It would be a confusion of levels, Quine suggests, to consider the corpus of rules toaccurately
represent the real structure of the language as it is practiced, even when the language under consideration is just an imaginary one. The only intelligible criterion for the accuracy of an explanation of such a language, whether real or imaginary, is just that it provide an interpretation of its sentences in
our
language: that is, that we be able to translate each sentence of the language under consideration into a sentence of like truth-value in our familiar one. If a conventionally introduced corpus of rules - what Quine would later call a “translation manual” - can do this, it is adequate in every real respect. The purport of any such corpus to represent real distinctions, above and beyond the facts about which sentences are accepted as true and which rejected as false, of (for instance) analyticity or syntheticity, must be rejected as empty.
IV
When, in 1937, Carnap offered his first published response to Quine’s incipient criticism of conventionalism, he reacted with tolerance, apparently perceiving in Quine’s suggestions no deep challenge to his own views. InFoundations of Logic and Mathematics
, Carnap re-iterated the position ofSyntax
with some minor modifications. Here he goes on to consider directly, in all but explicit reply to Quine, the question of whether logic is a matter of convention. As inSyntax
, to assert the conventionality of logic simply means, for Carnap, to deny that there is “a distinction between objectively right and objectively wrong systems” of logical rules.
And this assertion, Carnap continues to maintain, must be upheld,provided
we begin with the free stipulation of uninterpreted calculi, allowing the interpretation and meaning to be determined later. Carnap next reacted to Quine’s attacks in print two decades later, in the “Library of Living Philosophers” volume devoted to his work, a volume that also contained Quine’s “Carnap and Logical Truth.” In the brief response, Carnap again expressed puzzlement about the extent and intended force of Quine’s attack. In particular, he failed to see the reason for Quine’s apparent requirements, in “Two Dogmas” and “Carnap and Logical Truth,” that “analyticity” be given a general clarification, applicable to any arbitrary language, and that this clarification take the form of an empirical, “behavioristic” criterion. Carnap was especially puzzled in that he could find no argument, in Quine’s writings, to the effect that his actually suggested semantic and syntactic rules were not “exact and unobjectionable.”
In fact it is not surprising, given the extent to which Quine’s points about the arbitrariness of the stipulation of rules could thus be seem to be sympathetically absorbed by Carnap’s conventionalist doctrine, that Carnap never really saw Quine’s attack as having any great depth. But there was nevertheless a crucial difference in outlook and philosophical approach between the two philosophers, one that, as we have seen, appeared already in Quine’s first writings on Carnap. As we have seen, Quinealways
took it that the interpretation of any specialized logical notation, even one introduced as an autonomous, artificial language, would depend on the existing patterns of usage and agreed-upon understandings of terms and sentences in an already-understood language. Thus what was, for Carnap, only an optional starting point - the pre-existing meanings of the terms and sentences thatexplain
a logical calculus - was for Quine essential to the logical calculus having any interpretation at all.
Noting the extent to which Quine’s explicit results need not actually have been threatening to Carnap’s project, and the extent to which that project itself has subsequently been misunderstood, some recent commentary on the Quine/Carnap debate has attempted a partial rehabilitation of Carnap’s picture against what have elsewhere been taken to be Quine’s devastating criticisms. For instance, Creath (1987) argues that Quine’s arguments against conventionalism in “Truth by Convention” and “Carnap and Logical Truth” fail to attack any view that Carnap ever actually held.
Along similar lines, Ebbs (1997) argues that Quine’s attacks on conventionalism misses the pragmatic and programmatic spirit of Carnap’s suggestion that language frameworks be freely chosen. In particular, Carnap’s picture requires no metaphysically or epistemologically problematic picture of languages, and the logical truths within, them, as instituted or constituted by conventional, stipulative acts.
All that is required is what Ebbs calls Carnap’s “motivating insight”: that in order to settle philosophical and metaphysical disputes, we must explicitly “state rules for the use of linguistic expressions.”
But the rehabilitation of Carnap’s view can be, at best, partial. For although Quine did often present his attacks as bearing against a more general view than the one that Carnap actually held, his appeal to antecedent use provides, as we have seen, reason for doubting the wide freedom of choice that, according to the position Carnap actually did hold, the logician must enjoy. For it was a requirement for the cogency of Carnap’s view (his actual one as much as the other versions of conventionalism that Quine sometimes tended to attribute to him) that the logician’s freedom in creating new logical systems becomplete
: that, in other words, languages could reasonably be viewed as pure symbolic calculi, stipulated simply by laying down syntactical rules, without constraint by antecedently understood meanings. By contrast, Quine’s consideration of the role of antecedent use in providing an interpretation for whatever sign system we might create led him, from the start of his engagement with Carnap’s views, to doubt this key premise.
Ebbs argues further that the Carroll problem of infinite regress does not threaten Carnap’s view of linguistic stipulation, since investigators are already, in virtue of sharing a language, in a position to agree upon and take for granted some rules of inference, which they will then presuppose in determining and agreeing upon more specialized rules for the particular domain in need of clarification. But this begs the question against Quine by assuming that what is shared among native speakers of a natural language, as a presupposition for the possibility of communication, isalready
comprehensible as a set of agreed-upon rules, explicit or implicit. Though it is certainly true that investigators into a special area of language mustin some sense
antecedently share a language, if they are able to communicate at all, it is far from obvious that this sharing must amount to agreement upon any determinate set of logical or inferential rules, such as could help block the regress.
One significant obstacle, indeed, to understanding the depth and force of Quine’s attack against Carnap is that there is a great tendency to take the picture of language that Carnap held as inevitable or obviously true. It can seem simply obvious that if speakers share a language, their agreement simply in speaking it must amount to agreement onsome
corpus of rules, explicit or implicit, in principle capable of formulation and explicitation. The impression that this much is obvious may explain, to some extent, the tendency of commentators to understand Quine to be attacking aspecific view
of the institution or significance of the rules constitutive of language, a view that Carnap never held, and then to object that (as Carnap himself appears to have thought) the attack misses its mark. But in fact the scope of Quine’s attack goes much deeper, to the extent of challenging the seemingly obvious assumption that languagemust
be explicable as a rule-based calculus itself.
By the time he formulated the parable of Ixmann, Quine understood clearly that any interpretation of the actual rules supposed to be constitutive of a language could only amount to theprojection
of interpretive assumptions, at home in the interpreter’s language,
onto
the language under interpretation. It is implicit in this, and in the motivation of most of Quine’s various attacks on versions of conventionalism, that there is no non-arbitrary way to describe a language as a rule-bound calculus that is both consistent with, and wholly determined by, the actual use and practice of that language. In this sense, the force of Quine’s attack is not even limited to conventionalist pictures of the adoption of the rules supposed to govern language; it holds force against
any
picture, conventionalist or not, that supposes that language is explicable in terms of such rules at all.
Though Quine may never have put the point just this way, his attack on Carnap therefore called into question the exceedingly general notion of logical, linguistic, syntactic or semantic rules as constitutive or explanatory of a language. Such rules, if the upshot of Quine’s critique is right, can only be stipulated against the presupposed background of the understood meanings of terms in an already-existing language, a background which
itself
is not capable of explicitation as a system of rules (on pain of a Carroll-style regress).
V
As we have seen, Quine’s attacks on Carnap, beginning in 1934, developed from the innocent-seeming thought that the meanings of special linguistic symbols and rules could only be interpreted against the backdrop of an already-understood language. But although he always appealed in this way to antecedent use, and understood this as something other than an explicit corpus of rules, it was not always clearwhat
, exactly, was the object of this appeal. It was this that the model of radical translation, in its description of the limits and scope of the range of facts accessible to an interpreter with no antecedent knowledge of the language under interpretation, attempted to make maximally clear. With the model, Quine found, as well, a way to express the surprising upshot of his critique of Carnap as a general result about language and meaning, the indeterminacy of translation.
The descriptive set-up of the scenario of radical translation, which Quine first explicitly formulated in the second chapter ofWord and Object
, is familiar enough to require only a brief rehearsal. In radical translation, a translator is charged with the task of making sense of a wholly unfamiliar language, unguided by clues of shared or cognate word forms or cultural cues.
The attempt will culminate, if it is successful, in the production of atranslation manual
systematically linking sentences of the foreign language with sentences in the translator’s own language, or providing systematic, recursive recipes for such linkages.
The evidence on which the interpreter must depend in arriving at a systematic translation is limited to what she can observe of the natives’ speech behavior, including their tendencies to use various utterances in the presence of various observable phenomena and events, and the natives’ responses of assent or dissent, when queried as to the use of a particular sentence in a given environmental situation.
From this meager evidentiary base, meant nevertheless to capture all of the evidence thatcould
, in principle, be accessible in radical translation, the interpreter must construct a systematic translation of each native sentence into a sentence of his familiar language. The result, which Quine suggests at the beginning of the chapter, is that translation is systematically indeterminate. For, as a detailed appeal to the radical translation scenario will show:
…manuals for translating one language into another can be set up in divergent ways, all compatible with the totality of speech dispositions, yet incompatible with one another. In countless places they will diverge in giving, as their respective translations of a sentence of the one language, sentences of the other language which stand to each other in no plausible sort of equivalence however loose.
Before evaluating the indeterminacy result, it is important to understand the underlying motivational assumptions of the radical translation scenario itself. Since Quine wrote, it has been standard in the interpretive literature to object to the radical translation scenario on the ground that it restricts the interpreter artificially by placing tendentious and unmotivated limitations on the form of the evidence to which he may have access. If the evidence is so restricted, commentators have argued, the indeterminacy result follows trivially, but fails to establish anything significant about the nature of meaning or language overall. The impression of an unmotivated and artificial limitation on evidence, indeed, is strengthened by Quine’s consistent tendency to describe the totality of facts available to the interpreter - and indeed all the facts that thereare
about the use of the language - in a physicalist, behaviorist language of stimuli and responses.
But in fact, as we are now in a position to see, the impression that the radical translation scenario depends on behaviorism is, though perhaps fostered by Quine’s own rhetoric, quite superficial.
For its significance is the same as that of the appeal to use that Quine had consistently presupposed: that any interpretation of a language presupposes, and cannot go beyond, the facts of antecedent usage in the practice of that language.
Though sometimes couched in their idioms, this appeal itself has no essential dependence on behaviorism. Rather, it simply formulates methodologically the thought that the interpreter who does not already know a language can only avail himself of such facts as he might reasonably be thought, in this position, to have access to. If we are to make sense of the interpretation of a language as comprising a set of rules by means of which we can understand it (whether an explanatory calculus, as for Carnap, or a translation manual, as for Quine), it is important that the statement of the facts available at the outset not include any information about any logical, deductive, or grammatical rules that will later on be used toexplain
these antecedently observable facts. In this sense, the interpreter’s evidentiary restriction involves nothing more than a limitation to what must, on any account, be considered to be accessible to a potential interpreter, independently of the interpretation he will provide. This limitation, significantly, involves no prejudicial or tendentious limitation to one or anothertype
of facts (for instance facts “about behavior” or “expressible in physical terms”). Indeed,anything
that could, in principle, be observed by an interpreter innocent of the interpreted language can be included in the evidentiary base. The requirement is restrictive only in prohibiting a circular presupposition of an interpretation, prior to any interpretation actually being formulated.
The force of the indeterminacy result is not that, then, the facts about meaning are indeterminate with respect to some other, more restricted set of facts; but, rather, that forany
uninterpreted fact (be it about a subject’s behavior, his inner constitution, or whatever) there is an open question about its meaning that can only be answered bysome
interpretation or other.
The result follows readily from reflection about the extent to which the knowledge embodied by a translation manual, and requisite for providing an interpretation of a language as a whole, must systematically outrun anything directly required by the totality of facts antecedently available to an interpreter. The point, as Quine had already suggested in his attack on Carnap, is that any explicative introduction of rules specifying the form of a language goes significantly beyond what can be considered to be genuinely inherent in that language itself. The slack is taken up, in interpretive practice, by what Quine calls “analytical hypotheses,” systematic assumptions not directly required by any fact of linguistic practice, but stipulated in order to achieve maximum simplicity and charity in interpretation.
But because the analytical hypotheses are not uniquely determined by any objective facts of the matter, there is significant room for variability and arbitrariness in their stipulation. The result is that two translation manuals of a single language into another one can differ and disagree to a large extent, while still legitimately claiming to embody equally all the genuine facts about the underlying language.
Quine’s exposition of the indeterminacy thesis proceeds by considering, in detail, the procedure that a radical interpreter might follow in arriving at a systematic interpretation of a language, meanwhile showing the particular points at which indeterminacy tends to arise. The interpreter will begin with sentences that are assented to only momentarily or for a short time upon the presentation of a stimulus. Quine calls these “occasion sentences;” his classic example of this is the one-word sentence “Gavagai,” which prompts assent upon the presentation of a rabbit. Even here, with the sentences most directly keyed to present stimulations, indeterminacy threatens. For instance, it is impossible to exclude the possibility that the native occasion sentence refers at least in part to another object, seen by the native on a particular occasion but missed by the interpreter. More generally, the native’s assent or dissent to a prompted occasion sentence may depend as much upon collateral information held by the native as upon the presence of the stimulus itself.
The possible role of collateral information may be minimized, to some extent, by comparing different speakers of the language in point of their willingness to assent to various observation sentences. But since significant collateral information may be shared by all competent speakers of a language, it is never possible completely to factor out the contribution it makes to the observable facts, or to eliminate the translational indeterminacy that results.
Of course, the role of collateral information, and the extent of the resulting indeterminacy, grows larger when the translator moves from occasion sentences keyed as directly as possible to present stimuli to more abstract sentences, held true not only under particular, distinct conditions of stimulation but more enduringly or abstractly. And even if the problem of collateral information could be solved in some unique way, indeterminacy would continue to threaten under another heading, what Quine would later call the “inscrutability of reference.”
The problem is that the determination of a translation, even for the basic designative terms of simple occasion sentences, will depend on some systematic sense of the overriding categorical structure of the language as a whole, of its most basic means of sorting individuals into ontological types. This structure is itself undetermined by anything that the translator can observe, antecedent to interpretation. Thus, for instance, even if “Gavagai” is successfully tied to presently evident rabbits, there is nothing in this observational tie to require that “Gavagai” actually refers to rabbits (individuated aswe
would individuate them); it may, for all we know, refer merely to temporal stages of more enduring processes. Or it may refer to what is conceived as a part of a single, spatiotemporally distinct particular.
These aberrant possibilities seem unusual from our perspective; but there is nothing in the interpreter’s fund of evidence to exclude them. And if they may, indeed, obtain, then the interpreter’s evidence does not suffice to establish that the native’s term “Gavagai” and our term “rabbit” are coextensive, even if the former term is used by the natives under every circumstance in whichwe
would use the latter.
It follows that, beyond a core of observation sentences whose translation is maximally determinate, there is a wide range of sentences which may equally well be translated in any of various, clearly different ways. No matter what types or categories of facts are introduced into the observational base, there is no way to minimize the range of indeterminacy, without circularly presupposing the interpretation which it is the radical interpreter’s task to provide. But because the radical translation scenario models our ordinary capacity to understand meaning, it follows that there must be an ineliminable indeterminacy in the very meanings of our ordinarily understood sentences and terms. Though the fiction of an interpreter of a wholly alien language is used to expound the result, the model of radical translation also captures, according to Quine, the epistemic conditions each of us are under in coming to understand utterances in our own language, and the indeterminacy result must also be taken to hold for it. As Quine puts it elsewhere, “radical translation begins at home.”
Having admitted that indeterminacy affects any intelligible notion of interlinguistic sameness of meaning, or synonymy, there is no way to prevent it from affecting the intralinguistic notion as well.
It follows that, on any intelligible sense of “meaning,” two speakers may speak and understand the same language, and yet diverge radically in the meanings they associate with its sentences.
The result, thus put, has an air of extreme paradox. If it is correct, it seems to follow that the vast majority of the sentences that we use everyday, in ordinary language, have no determinate meaning. When I use any one of these sentences, even one as plain as “there is a rabbit,” there is no determinate fact of the matter about what I mean. And this does not result simply from giving “meaning” a specialized or philosophically loaded sense; Quine’s claim is that indeterminacy of meaning arises forany
coherent notion of linguistic meaning, no matter how broad or general.
Perhaps because of its extreme air of paradox, commentators responding to the indeterminacy result have often attempted to find grounds, for instance in considerations overlooked by Quine about the conditions which must be satisfied for a speaker to master a language, on which it is possible to argue that the actual extent of indeterminacy of meaning, in the real practice of a language, is in fact significantly less than Quine suggests, or perhaps actually nonexistent.
But by seeing the real sources of the indeterminacy result in Quine’s sustained critique of the picture of languages as calculi, we can fully accept the result while at the same time perceiving the larger implications of the paradox it articulates. To a large extent, the paradoxicality of the result arises from the seeming poorness of its fit with our ordinary intuitions about the use of language. When somebody utters a sentence in my own language and I take myself to understand it, I generally have no sense of arbitrarily selecting one meaning or interpretation from a variety of systematically different possibilities. Nor does the abstract possibility of alternative translation manuals seem to pose any practical obstacle to the ordinary practice of communicating and understanding meanings. Indeed, there seems to be an obvious sense in which, in uttering a familiar English sentence meaningfully, Imust
, as a competent speaker of English, be said to understand and be capable of communicating its meaning.
Indeterminacy thus seems to have no effect on ordinary linguistic practice; it is perfectly possible to say something, and mean something determinate by it, without having any particular systematic translation manual in mind at all. It can seem difficult or impossible to square these obvious features of the phenomenology of ordinary language with the claim that there is, when I utter a normal, declarative sentence, no genuine fact of the matter about what I mean. But it is this claim that the indeterminacy result implies; and hence it can seem that the only reasonable way to react to it is to find hitherto unnoticed grounds, implicit in our understanding of linguistic practice, for denying that the result could be true.
But we can put the result in a different perspective by placing it against the backdrop of Quine’s longstanding appeal to antecedent use, and reflecting on the way in which this appeal provided grounds for his emerging critique of Carnap’s picture of languages as calculi. For seen against this backdrop, the indeterminacy of meaning is, in effect, the product of two separate and somewhat (though not completely) isolable factors.One
of these factors is the totality of facts about the ordinary practice of a language, captured in Quine’s formulation as the totality of facts antecedently accessible to the interpreter. But another factor, equally crucial to the result, is introduced by the attempt to schematize or specify meanings by formulating them explicitly in a translation manual. That meanings so specified must systematically outstrip any determinacy actually present in the facts they purport to represent and systematize is a key thought of Quine’s, from early in his dialogue with Carnap. But this point implies no threat to the evident determinacy of these facts in themselves. If speakers are confined to the realm of anunreflective
linguistic practice, anddebarred wholly from reflecting about any systematic principles or rules underlying their use of language,
no troubling impression of indeterminacy need arise. Ordinary communication proceeds untroubled, without any need to work out or specify an entire interpretation or translation.
The indeterminacy only emerges as part of the reflective practice of explicating and specifying meanings, a practice that the radical translator’s activity of translation explicitly models. It is only within the ambit ofthis
general reflection that the possibility emerges of translating one and the same utterance in two radically different ways. Without it, the fact of indeterminacy remains, but it need not be considered to introduce anything paradoxical into the phenomenology of ordinary, unreflective practice.
But in practice, it will, of course, be impossible to make this a clean separation. As we saw in the last chapter, the possibility of systematic reflection about the ground and basis for linguistic meaning is inscribed in a language as soon as it contains the predicate “means” itself. Indeed, as soon as a language includes expressions for such notions as “meaning,” “truth” and “language” the reflective activity of explicitation that would culminate in a formal calculus or translation manual has already implicitly begun. A language purged of these expressions, and hence debarred from the possibility of systematic reflection on the basis of linguistic meaning, would scarcely be recognizable as a (human) language at all.
To construe the indeterminacy result as an artifact of reflection on the form of a language is not, then, to limit its significance to the abstract, theoretical activity of linguists and philosophers. In the ordinary, everyday practice of clarifying and reflecting on meanings, a practice which presupposes the concepts which, if fully explicated, would yield a systematic understanding of the structure of the language as a whole, indeterminacy and conflicting interpretations may arise at any point. But since it can be taken to be essential to human conversation that it
always
involve at least the possibility of raising questions of meaning, or of interpreting and criticizing what has been said with reference to an understanding of a language as a whole, this practice is none other than ordinary interlocution. Its ambiguities and indeterminacies are those of language as such, anywhere and everywhere it plays a role in human relation.
In interpreting the indeterminacy result as arising from the specific instabilities of a structuralist picture of language such as Carnap’s, moreover, it is important not to lose sight of the depth of the sources of this picture in our everyday thinking about language, and the genuine difficulty of resisting it. In the course of any systematic attempt to reflect about language as a practice it seems just obvious that this practice must, on some level of description, be guided by systematic rules of grammar and inference that can, at least in principle, be recovered by theoretical reflection. This seemingly obvious assumption forms a large part of the basis of projects, throughout the analytic tradition, that see themselves as clarifying or making explicit the underlying logical, semantic, grammatical, or pragmatic form of language. Carnap himself never questioned it, always assuming (despite the large amount of room his conventionalism allowed for arbitrariness and stipulation in the reconstruction of a language) that the explication of a language, or an area of a language, in terms of a specialized calculus could genuinely clarify and account for real, underlying relations of justification and inference within that language.
The picture of language as a calculus cuts so deeply in ordinary and philosophical thinking, indeed, that Quine himself, despite his sustained critique of it, also does not seem to completely escape its influence. Other regions of his thought, less closely connected to the dialogue with Carnap, tend to re-instate it, at least in part; and its vestigial influence on Quine’s thinking may explain why he never posed the indeterminacy result explicitly and specifically as a critique of it. For instance, Quine held, beginning inWord and Object
, that a logical “regimentation” of specific regions of language could clarify their inferential structure and ontological commitments.
The famous holist picture of language as an interconnected “web of belief”, surrounded at the outer perimeter by experience, with which he ends “Two Dogmas of Empiricism,” seems to suggest that the total state of language, diachronically revisable and changeable though it may be, could be portrayed, at least at any specific moment, by a determinate calculus of rules relating currently accepted propositions, both to each other and, holistically, to the empirical world. And the naturalist vision of epistemology that he celebrated beginning with “Epistemology Naturalized” can seem to suggest that general principles of the grammatical and inferential practice of a language could be determined purely empirically by means of reflection about the physiological route from sense-stimulation to the fixation of beliefs and their expression in behavior.
In each of these cases, the appeal to some notion of language as a calculus is less complete and explicit than Carnap’s conception, but it remains in the background nonetheless. These vestigial remnants of the picture of language as a calculus need not imperil the more general recognition that we have located in Quine’s critique of Carnap, to the effect that it is impossible to foreclose the indeterminacy that is a necessary result, once we conceive of language as a calculus. But their seeming irresistibility, as soon as systematic thinking about language begins, can start to explain why Quine himself never formulated this general recognition in these explicit terms.
More broadly, by understanding how Quine came to articulate a fundamental criticism of the picture of languages as calculi on the basis of his ongoing appeal to use, we can derive a striking general lesson about the role of the interrelated notions of rules, use, and practice in our ordinary understanding of language. It is anessential
part of this understanding that words and expressions are describable as similar, identical or different in meaning, and that this description, when offered, could be underwritten by a description of similarities, identities, or differences in the regularities of use. The most radical and surprising implication of Quine’s indeterminacy thesis is that this assumption of regularity is ungrounded in anything we could discern as a description of the facts. The set of assumptions of the determinacy and identity of meanings that make possible not only our ordinary reflection on meaning but the ordinary conversations in which this reflection plays an essential part stand revealed, then, as mythologies. Nevertheless they remain operative in what we regularly grasp as our regular “practice” of using language, and continue to essentially determine what we do and say within it. As soon as we begin to reflect on our practice of using words as such, the possibility of describing meanings as the “same” or “different” emerges as an essential part of this practice; but the effect of Quine’s result is to show that nothing describable as part of this practice grounds this possibility of determining sameness and difference of meaning.
At an earlier stage of its pursuit, the analytic tradition’s reflection on language had been explicitly directed against the mythology of “ideas” or psychological items as the underlying basis for judgments of identity or difference of meaning. With Quine’s indeterminacy result, this reflection reaches its most radical conclusion. In the more radical application that Quine’s indeterminacy result exemplifies, the critique bears not only against the earlier psychologistic conception but also against the pervasive mythology of meaning as grounded in regularly describable “usage” as well. It remains that the assumption of a substantial basis, in practice, for judgments of the identity and difference of meaning play a pervasive and practically ineliminable role in the simplest situations of intersubjective life. The startling effect of Quine’s result is to show the impossibility of any attempt to discharge this assumption by reference to the facts of use. If my assumption that an interlocutor will go on using a word in the “the same way” I do, or that he means the same thing with his utterances or inscriptions that I would mean in using auditorially or lexicographically similar tokens, indeed has a basis to which I can appeal, this basis is (as we might put it) nothing other than the fact that we share a language; and this fact is not further explicable in terms of facts of linguistic usage or reference more primitive or basic than it itself. This fact grounds every possibility of human linguistic communication, and of the application of linguistic criticism to the circumstances and practices of human life. But within the systematic attempt, engendered already with the first word of language’s reflection on itself, to comprehend its system and schematize its principles, it emerges as itself groundless, the essentially elusive core of human mutuality itself.
10%
Author: Paul M. Livingston
