REESE M. HEITNER
FROM A PHONO-LOGICAL POINT OF VIEW: NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
ABSTRACT. Though largely unnoticed, in ‘‘Two Dogmas’’ Quine (1951, Two Dogmas of Empiricism, Philosophical Review 60, 20–43. Reprinted in From a Logical Point of View, 20–46) himself invokes a distinction: a distinction between logical and analytic truths. Unlike analytic statements equating ‘bachelor’ with ‘unmarried man’, strictly logical tautologies relating two word-tokens of the same word-type, e.g., ‘bachelor’ and ‘bachelor’ are true merely in virtue of basic phonological form, putatively an exclusively non-semantic function of perceptual categorization or brute stimulus behavior. Yet natural language phonemic categorization is not entirely free of interpretive semantic considerations. ‘‘Phonemic reductionism’’ in both its linguistic (Bloch 1953, Contrast, Language 29, 59–61) and behavioral (Quine 1990, The Phoneme’s Long Shadow, Emics and Etics: The Insider/Outsider Debate, T. Headland, K. Pike and M. Harris, (eds.), Newbury Park, CA, Sage Publications, 164–167) guise is false. The semantic basis of phonological equivalence, however, has repercussions vis-a`-vis Quine’s critique of analyticity. A consistent rejection of meaning-based equivalencies eliminates not only analyticity, but imposes a form of phonological eliminativism too. Phonological eliminativism is the reductio result of applying Quinean meaning skepticism to the phonological typing of natural language. But unlike analyticity, phonology is presumably not subject to philosophical dismissal. The semantic basis of natural language phonology serves to neutralize Quine’s argument against analyticity: without the semantics of meaning, more than just synonymy is lost; basic phonology must also be forfeited. Let’s begin with the fact that even Quine has to admit that it is possible for two tokens of the same orthographic type to be synonymous, for that much is presupposed by his own account of logical truth. Paul Boghossian (1999, 343)
1. LOGICAL, ANALYTIC, AND SYNTHETIC TRUTH
Ours is a philosophical environment deeply colored by Quine’s arguments in ‘‘Two Dogmas’’. Yet it is important to recall that by way of introducing his critique of analyticity, Quine actually draws a three-way distinction among logical, analytic, and synthetic truths. Concerned as he is with the analytic/synthetic distinction, Quine (1951) briefly notes the logically formal status of tautological Synthese (2006) 150: 15–39 DOI 10.1007/s11229-004-6253-z
Ó Springer 2006
16
REESE M. HEITNER
substitutions to contrast them with the semantic nature of analytic substitutions, and advances from there to pursue a sustained critique of the latter. According to Quine, analytic truths are, in principle, no more immune to empirical revision than synthetic truths and synthetic truths, in this way, no less ‘‘analytically’’ invulnerable. There is no philosophically defensible distinction to be had between semantically based analytic truth and empirically based synthetic truth. Analytic truths holistically bleed into empirical truths, and empirical truths holistically bleed into analytic truths. From this perspective, ‘‘meaning’’ is just an elaborate network of firmly coagulated beliefs, deeper and more vivid in color, but different merely in degree rather than kind from other beliefs. Quine does not, however, question the basis of a distinction between logically tautologous substitutions and analytic substitutions. Indeed, while undermining a traditional analytic/synthetic distinction, Quine implicitly helps himself to the presumption that tautologous substitutions between two word-tokens of the same word-type, say, ‘bachelor’ and ‘bachelor’ – in contrast to analytic substitutions between two tokens drawn from two different word-types, say, ‘bachelor’ and ‘unmarried man’ – are ostensively free of the kind of semantic considerations that trouble analytic truths. But this is a false dichotomy. Basic word-type equivalence among token distinct word forms is itself dependent upon semantic considerations. Interpretive semantic information is required in order to comprehensively group even distinct word-tokens into equivalent word-types in natural language. ‘‘Phonology is’’, as one researcher concluded ‘‘a semantically conditioned science’’ (Ullmann 1957, 32). For philosophers, like W. V. Quine, who have a dim view of semantics and fault analyticity for its semantic dependence, the semantic basis of basic word-type equivalence is prima facie problematic. This is why in a series of publications, Quine (1953a, 1981, 1987, 1990) has attempted to avoid any appeal to semantics in his efforts to reconstruct basic phonological word-type equivalence on an exclusively non-semantic basis. In fact, alone among contemporary philosophers of language, Quine has repeatedly challenged the semantic basis of phonological word-type equivalence in favor of a non-semantic account based upon acoustic, phonetic, perceptual or, as a last resort, behavioral data. Consistent with his rejection of semantic-based analytic equivalence, Quine also rejects a semantic-based account of basic word-type equivalence. But ‘‘phonemic reductionism’’ – the attempt to account for natural language phonological categorization
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
17
(not real-time perceptual processing) independent of interpretive considerations of semantics – is philosophically appealing but false. Given the empirical failure of a non-semantic reduction of basic natural language word-type equivalence – and the philosophical short-comings of a behavioral reduction – a consistent disavowal of semantics prohibits Quine from acknowledging the legitimacy of basic word-type phonological equivalence anymore than he wishes to acknowledge the legitimacy of traditional analytic equivalence. In this way, a consistent Quinean rejection of semantics entails not only a rejection of analytic equivalence, but also a rejection of basic phonological equivalence. So much the worse for both phonology and analyticity, Quine might say. But whereas philosophers might be content to eliminate analytic equivalencies among synonyms, a consistent meaning eliminativism jeopardizes basic natural language phonological equivalence. Unlike analytic eliminativism, ‘‘phonological eliminativism’’ is the result of applying Quinean meaning skepticism to the phonological typing of natural language, no doubt why seeking to protect basic phonological equivalence from semantics has been just as important for Quine as eliminating traditional semantic-based analytic equivalence. The observation that basic phonological equivalence is semantically conditioned suggests a way of neutralizing Quine’s rejection of analyticty by way of reductio: either the semantic-based equivalencies of both analyticity and phonology are uniformly rejected, or the semantic-based equivalencies of both analyticity and phonology are uniformly tolerated. But whereas the former sacrifices natural language phonology in favor of a consistent semantic eliminativism, the latter move neutralizes Quine’s argument against analyticity. And this choice must be made independent of the virtue or vice of traditional analytic equivalence. Without recourse to interpretive semantics, linguists can no more comprehensively justify basic natural language word-type equivalence than philosophers of language can justify traditional analytic equivalence. It is unsurprising then, as Quine observes, that the behavior of synonyms – relations between distinct word-types – cannot be accounted for without appealing to something other than logical form, as even the foundation of basic word form is itself semantically conditioned. The failure to asemantically reduce natural language phonology constrains what can consistently be said about the semantics of analyticity without jeopardizing phonology. If it is meaning which dooms analyticity for Quine, then Quine’s argument applies equally against phonology. But a rejection of natural
18
REESE M. HEITNER
language phonological equivalence – phonological eliminativism – is more likely to be viewed as a prohibitively costly reductio than merely the price to pay for consistent meaning eliminativism. At stake, then, is the overall cogency of Quine’s attack on the semantic basis of analytic equivalence. Given that basic phonemic equivalence is also semantically conditioned, Quine’s eliminativism threatens much more than just analyticity. One may philosophically dismiss analyticity in virtue of a rejection of semantics, but such semantic eliminativism does not come for free. Quine’s semanticbased attack on analytic equivalence incurs the expense of eliminating basic natural language phonological equivalence. In fact, Quine’s semantic-based attack on analytic equivalence undermines the very linguistic resources by which logicians minimally recognize basic tautological truths equating type equivalent but no less token distinct linguistic forms, e.g., ‘bachelor’ and ‘bachelor.’ (In order to be tautologically equivalent – in order to be phonologically equivalent – two word-tokens do not have to be imperceptibly indistinguishable, see below.) Basic word-type equivalence is not only fundamental to everyday communication. Basic word-type equivalence relating type equivalent linguistic tokens is fundamental to identifying basic tautological truth. Without phonological word-type equivalencies, tautological truth itself dissolves for it is phonological equivalence which governs basic tautological relations among word-tokens of the same word-type. Tautological truths among word-tokens of the same word-type are really instances of basic phonological substitutions among word-tokens of the same word-type. In addition, then, to analytic and synthetic truth, there is strictly logical, tautologous truth. But whereas Quine perceives a dogmatic distinction between the analytic and the synthetic, there lurks another dogma of empiricist extraction: a distinction between logically formal, that is, ‘‘form-driven’’ semantically independent tautological substitutions and analytic, that is ‘‘content-driven’’, semantically dependent synonymous substitutions. Central, however, to exposing the defective nature of this distinction, and recovering from its philosophical liabilities, is a reassessment of basic phonological equivalence, indeed, Quine’s own account of natural language phonological equivalence. Notwithstanding Quine’s relatively recent and, from what I can tell, final remarks on the phoneme in a largely unknown article entitled ‘‘The Phoneme’s Long Shadow’’ (1990), still more light needs to be thrown on an issue basic to linguistics, but critical to the overall success of Quine’s attack on analyticity.
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
19
Propelled by the observation that even basic natural language allophonic equivalencies among phonetically distinct variants of the same word (the basic ‘‘phono-logical truths’’ which underwrite all our garden-variety tautological substitutions between word-tokens of the same word-type) are, pace Quine, themselves dependent upon what Quine himself considers to be illegitimate semantic considerations an attack on analyticity on account of its semantic recalcitrance has a number of repercussions.1 2. PHONO-LOGICAL TRUTH: ‘‘FORMAL’’ SUBSTITUTIVITY
On the one hand, and within analytic philosophy of language, questions of linguistic equivalence are principally addressed through the test of truth-preservation under substitutional equivalence and through related considerations of propositional opacity and necessity. On the other hand, and within linguistics, issues of substitutional equivalence revolve around the notion of the phoneme: two phonetic segments are phonemically equivalent if their substitution never has any semantic effect.2 For example, the phonetic feature known as aspiration routinely occurs in English but never serves to distinguish one word from another (Clark and Yallop 1995). Analogous to italization in standard English orthography, aspiration – though phonetically present – does not affect any change. By contrast, the l/r distinction, though phonemically distinctive in English, is not distinctive in Korean or Japanese. This is why ‘‘two’’ productions of ‘The book is on the right’ and ‘The book is on the light’, if said in accordance with Japanese phonology, would have the same truth value: ‘light’ and ‘right’ are phonologically equivalent. In Japanese, ‘The book is on the right’ and ‘The book is on the light’ would be the same sentence said twice, but two different sentences according to English phonology. As Bennett (1988) points out, Where English is concerned, linguistic significance includes the respects that differentiate ‘can’ from ‘con’, but not those that differentiate a clumsily printed ‘can’ from an elegantly written one. There could be languages, however, where the difference between shapes like our printed ‘a’ and ‘o’ was a mere matter of handwriting, with no linguistic significance, while a shape like our handwritten ‘a’ represented a completely different letter. (404)
All natural languages are subject to this sort of variability, and with regard to speech, is technically known as known as language-specific ‘‘allophonic’’ variation. When the variation is predictable on the basis
20
REESE M. HEITNER
of phonotactic regularities, the allophones are said to be in ‘‘complementary distribution’’. When the variation is not phonotactically predictable, the allophones are said to be in ‘‘free variation’’. These details, however, should not obscure the observation that these two tests of substitutional equivalence – one from phonology and the other from philosophy – are only superficially distinct. In both cases, we are concerned with specifying the conditions for recognizing equivalence relations among identifiably distinct linguistic specimen. In natural language phonology, such substitutional equivalencies are extensive and routinely occur among distinct allophonic productions. Within the philosophy of language, however, matters point in the opposite direction: Fregeans maintain that coreferring names (and perhaps even synonymous words, see Mates 1950) provoke failures of substitution, while Quine views analytic substitutions with eliminative suspicion. Indeed, given the prominence afforded Quinean arguments attacking analyticity, little time need be spent reviewing the circular use of analyticity to prop up the intuitive equivalence between synonyms such as ‘bachelor’ and ‘unmarried man’ or ‘attorney’ and ‘lawyer.’ Suffice it to say that, unlike substitutions between ‘a’ and ‘b’ (substitutions between two different phonological types), strictly tautologous substitutions between say, ‘a’ and ‘a’ (two tokens of the same type) are presumably not subject to Quine’s skepticism as are ‘bachelor’ and ‘unmarried man’ or ‘attorney’ and ‘lawyer’ – pairs which exemplify different word-types. Unlike basic ‘‘phono-logical’’ truths spanning two distinct tokens of the same type, analytical truths can be converted into strictly formal tautological truths – and thereby immunized against meaning skepticism – only by circularly appealing to the interpretive content the two synonyms are assumed to have in common. As Quine (1960) explains in Word & Object (with respect to sentence synonymy), The general relation of intrasubjective sentence synonymy thus unsuccessfully sought is interdefinable with another elusive notion of intuitive philosophical semantics: that of an analytic sentence . . . . The interdefinitions run thus: sentences are synonymous if and only if their biconditional (formed by joining them with ‘if and only if’) is analytic, and a sentence is analytic if and only if synonymous with self-conditionals (‘If p then p’). (65; italics added)
Postulating propositions and meanings as the semantic glue which somehow binds the apparent equivalence between translations and synonyms is a philosophical solution that is, as Quine has derisively
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
21
argued, suspiciously over-designed for the problem. At best, ‘‘propositions’’ and ‘‘meanings’’ are merely labels which succeed only in papering over a problem. At worst, talk of ‘‘propositions’’ and ‘‘meanings’’ acts to self-perpetuate a non-existent pseudo-problem: ‘‘Pending a satisfactory explanation of the notion of meaning, linguists in semantic fields are’’ quips Quine (1953a, 47), ‘‘in the situation of not knowing what they are talking about’’. Deeply skeptical of the introduction of abstract objects, and weary of the many lightly veiled guises of meaning, but equally confident that no comparison of the actual physical appearances of ‘bachelor’ and ‘unmarried man’ yields a single ‘‘natural kind’’ category that subsumes them both, categorizing them as instances of a single, but inherently unspeakable, abstract object is scientifically unjustifiable and arguably incoherent. Asking synonymy, or the more technical notions of ‘‘intensional isomorphism’’ (Carnap 1947) or ‘‘synonymous isomorphism’’ (Church 1954) to prop up analytic equivalence are, according to skeptics, all fundamentally misguided attempts which merely substitute, as Quine would say, one ‘‘creature of darkness’’ with still others. Rereading the opening pages of ‘‘Two Dogmas’’, it is clear that Quine (1951) considers analyticity to fall short of the objective standard of the logically tautologous, that is strictly ‘‘syntactic’’, ‘‘formal’’ or as with the above passage from Word & Object, ‘‘selfconditional’’ substitutions based exclusively on form. Indeed, ‘‘the business of formal logic’’, says Quine (1953b) in his critical review of Strawson’s (1957) Introduction to Logical Theory,3 ‘‘is describable as that of finding statement forms which are logical, in the sense of containing no constants beyond logical vocabulary, and extensionally valid, in the sense that all statements exemplifying the form in question are true. Statements exemplifying such forms may be called logically true’’ (139). And as Carnap (1937) writes in his The Logical Syntax of Language: ‘‘A theory, a rule, a definition, or the like is to be called formal when no reference is made in it either to the meaning of the symbols (for example, the words) or to the sense of the expressions (e.g., the sentences), but simply and solely to the kinds and order of the symbols from which the expressions are constructed’’ (1). Simple substitutions among type equivalent tokens, i.e., substitutions between ‘bachelor’ and ‘bachelor’ are ostensively formal in this sense, for it is their form – not semantic content – that licenses their substitution.
22
REESE M. HEITNER
Indeed, if there were some way of explicating the analytic equivalence between such distinct word-types as ‘bachelor’ and ‘unmarried man’ in terms of their objectively physical qualities of shape and size – rather than their semantically interpreted properties – the interrelated issues of translation, analyticity and synonymy would resolve themselves in the same way they apparently do with respect to such substitionally equivalent word-tokens as ‘bachelor’ and ‘bachelor’ (like allophonic speech variants generally), without the unwelcome assistance from ontologically suspect quarters. After all, two tokens do not have to be indiscernibly identical in order to legitimately constitute the same linguistic type – though exactly how different they can be and still count as the ‘‘same’’ type is by no means clear and, in any event, subject to language-specific variation. Presumably ‘bachelor’ and ‘BACHELOR’ are both token exemplars of a single linguistic English word-type – not because of what they mean – but simply because ‘bachelor’ and ‘BACHELOR’ intrinsically share certain objectively identifiable perceptual features in common whereas ‘bachelor’ and ‘unmarried man’ obviously do not. Whether by some relatively simple measure of psychophysical similarity (Quine 1953a) or a more complex distributional comparison of phonetic features technically known as ‘‘complementary distribution’’ (Harris 1951; Bloch 1953), or even some sort of behavioral index (Quine 1987, 1990), any scientifically legitimate classification of linguistic items must, for Quine, be based upon some sort of ‘‘bottom-up’’, though undoubtedly liberal, measure of perceptual or behavioral similarity, not semantic considerations of meaning. After all, the force of Quine’s fire against analytic equivalence, if indeed it is to be directed only at analyticity rather than also undermining natural language phonology, presupposes that, unlike analytic equivalence, basic phonological equivalencies actually do yield to an explanation that is purely formal, that is to say, independent of any semantic considerations. Quine’s attack on analyticity is fueled by the observation that analyticity is semantically dependent. A semantic basis for phonological equivalence must be avoided if phonology is not also to be consumed. No doubt this is why Quine’s eliminative skepticism directed against loose and circular appeals to semantics to ‘‘explain’’ analyticity is matched only by his reductive confidence that basic phonological equivalencies are not, could not be, beset by any such semantic obscurities. If meaning is Quine’s target – and analyticity an intended causality – phonology becomes collateral damage unless it can be accounted for in non-semantic terms.
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
23
Apparently cognizant of the importance of providing a purely formal non-semantic account of basic phonological equivalence to contrast with the semantically dependent definition of synonymy, it is perhaps unsurprising that philosophers of language at the forefront of criticizing analytic equivalence have been equally forceful in maintaining that, by contrast, there exist entirely reductive nonsemantic explanations for the garden variety phonological equivalencies of the ‘bachelor’ and ‘bachelor’ kind. Indeed, both Quine and Nelson Goodman have been anxious to assure us that such nonsemantic reductions are ready and waiting. Notably Goodman’s (1972) notion of ‘‘replication’’ and Quine’s (1953a) half-hearted adoption of psychophysical acoustics as the basis of phonemic categorization are intended to guarantee that the ‘‘bottom-up’’ perceptual cues of visual shape and acoustic sound are what ground basic orthographic, and more importantly, basic phonological equivalencies, rather than anything related to semantics. A revealing example of this can, in fact, be found in the collaborative Goodman and Quine (1947), where only by force of nominalist confidence do Quine and Goodman help themselves to the ostensibly unproblematic reduction of linguistic types to inscriptional shapes via their notion of ‘‘shapepredication’’ (122). In fact when inveighing against the proliferation of abstract entities in their ‘‘Steps toward a constructive nominalism,’’ Goodman and Quine (1947) jointly aspire to reconstruct mathematical expressions as ‘‘concrete objects – as actual strings of physical marks’’ without the trappings of any platonistic devices of traditional mathematical syntax. In order to do this, Goodman and Quine emphasize that ‘‘the characters of our language are not these abstract shapes – which we, as nominalists, cannot countenance – but rather concrete marks or inscriptions . . .’’ (112). Taking another example, Ramsey (1950) writes that ‘‘tokens (like those of any sign) are grouped into types by physical similarity (and by conventions associating certain noises with certain shapes) just as words’’, adding ‘‘But a proposition is a type whose instances consist of all possible sign tokens which have in common, not a certain appearance, but a certain sense’’ (274; italics original). As for the role abstract propositions, meaning and semantics plays in explaining basic linguistic equivalence, Quine would have it that philosophers do not want it and linguists do not need it. But while the proposals suggested by Goodman and Quine reveal much more about their commitment to a semantic-free basis for individuating linguistic types rather than serious, no less accurate,
24
REESE M. HEITNER
empirical proposals (and both are happy to acknowledge this4), this is not to suggest that no empirically credible accounts of basic phonological equivalence are available. Indeed, specifying precisely how two word-tokens may differ but still exemplify the same word-type is among the professional pursuits of phonologists. In fact, one of the most significant objectives of early twentieth-century phonology, particularly as practiced in the United States among an intense if ultimately overshadowed group of so-called ‘‘neo-’’ or ‘‘postBloomfieldian’’ distributional linguists (Koerner 1978; Hymes and Fought 1981), was to discover the underlying objective reality governing the intuitive distinction between significant and insignificant sound differences in natural language. Viewing themselves as the, though largely symbolic, heir to Leonard Bloomfield’s linguistic leadership (Fought 1999a, b), their positivistically oriented linguistics was explicitly predicated upon the minimization (Harris 1951), if not out-right rejection (Bloch 1953), of semantics in an effort to identify the ‘‘objective’’ basis of phonological equivalence among phonetically distinct tokens without the aid of interpretive semantic considerations. To their credit, if not success, such asemantic methodological restrictions are far from trivial. For not only do different languages characteristically differ as to how they cross-classify precisely the same two sounds (spoiling any attempt to uncover a universal standard of phonological equivalence), languages internally generate both systematically rule-governed and random variations in the pronunciation of its own sounds. More specifically, the phonologist is saddled with the task of explaining why native speakers reveal subtle patterns of phonetic variation whereby the ‘‘same’’ sound is sometimes produced one way and sometimes produced another way – a phenomenon technically known as ‘‘allophonic variation’’ or ‘‘subphonemic variability’’.
3. THE REDUCTIO THREAT OF ‘‘PHONOLOGICAL ELIMINATIVISM’’
To be sure, none of the phonotactically conditioned variants in ‘‘complementary distribution’’ or phonotactically unconditioned variants in ‘‘free variation’’ are as substitutionally radical as those which typify the analytic substitutions between synonyms or translations. In fact, for the native speakers who regularly produce and hear them, such allophonic discrepancies (like aspiration in English,
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
25
or the l/r distinction in Japanese) demand distinctions that typically border on the articulatory unproducible and audibly imperceptible (Strange 1995). But however seemingly negligible, neither is there a set of either invariant acoustic, perceptual or even exhaustive allophonic distributional cues to objectively signal phonemic equivalence among phonetically distinct tokens (see Bloomfield 1933). Basic phonological equivalence – the ‘‘phoneme’’ – cannot be defined in terms that are exclusively ‘‘formal’’ and independent of interpretive information: ‘‘phonemic reductionism’’ is false. The only feature which determines phonemic equivalence in the cases of naturally occurring free allophonic variation is the semantic equality of the tokens.5 But this is really nothing new. As Bloch and Trager (1942) explain, [A] purely phonetic description makes it impossible to distinguish the really significant features of the vocabulary and the grammar from the accidental and personal features which inevitably form part of every utterance; as a scientific procedure it is about as fruitful as it would be for a biologists to assign two cats to different species because one had more hairs on its tail than the other. (39)
Yet given the failure to reduce basic phonological equivalence to the non-semantic factors of phonetics, the observation that two identifiably distinct phonetic forms (like the orthographic variants of ‘bachelor’ and ‘bachelor’) are considered to be the same ‘‘type’’ provokes the same issue as does the philosopher’s quest to explain – or explain away – intuitions of ‘‘analytic equivalence’’ between forms as diverse as ‘bachelor’ and ‘unmarried man.’ The semantic basis of natural language phonological categorization induces Quine to forfeit – in addition to analytic equivalence – phonological equivalence, or concede the role meaning plays in grounding both natural language phonology and synonymy. For Quine, neither option is agreeable: ‘‘phonological eliminativism’’ is pyrrhic and the latter is un-Quinean. Yet Quine’s objections against meaning are one of principle, and this is what leads Quine to redouble his efforts to deliver on a nonsemantic account of natural language phonology. According to Quine, phonemic categorization need not be reduced to the phonetics of the speech stream. Instead, a behavioral reduction of phonological equivalence is sufficient. Indeed, following the earlier lead of some prominent ‘‘neo-Bloomfieldian’’ linguists, Quine pursues a behavioral reduction of basic phonological equivalence so as to avoid a reliance on semantics. But raw behavior stripped of semantic significance is no more a key to unlocking the phonemic structure of a language than
26
REESE M. HEITNER
raw speech stripped of its semantic significance is. In fact, Quine’s attempts to provide a behavioral reduction of phonemic equivalence are not empirical alternatives subject to experimental disconfirmation, but philosophical circumlocution susceptible to direct refutation: such behavioral reductions are inevitably false, trivial, or circular.
4. QUINE’S BEHAVIORAL REDUCTION OF PHONEMIC SUBSTITUTIONS
In an effort to avoid contaminating basic phonemic equivalence with semantics, Quine proposes to rely on the behavior of an informant. Rather than observe the language itself, Quine endorses a behaviorist reduction of linguistic equivalence whereby ‘‘two sounds belong to the same phoneme if substitution of one for the other does not affect a speaker’s disposition to assent to any sentence’’ (1987, 150). Yet this proposed analysis is either false or profoundly subjective as no room is made for the possibility of phonological error, either in the form of fleeting performance ‘‘slips of the ear’’ or systemic confusion. Surely the standards of phonemic categorization are not to be blindly determined by just anyone: the young, the foreign, or the hard of hearing. Yet determining who is a ‘‘competent’’ judge of phonological equivalence presumably requires some antecedent notion of what the standard is so that it may be selectively confirmed. But dispositional accounts of phonological equivalence have much deeper problems than encouraging a form of phonological individualism. Quine’s behavioral account does not solve the problem of providing a non-semantic account of phonemic equivalence; it merely surrenders to it. Basing phonemic equivalence on even the behavior of acceptably competent informants is not a solution because such is merely a diversionary tactic, successful only in postponing an inevitable semantic concession. It is important to remember that when confronting an alien tongue (or even re-approaching one’s own language), it is supposedly the job of the uninformed, untutored, linguist to objectively determine when two sounds are equivalent. Relying on a native informant for this information is not to answer the question, but merely to have someone else answer it for you. So while judging phonemic equivalence on the basis of overt behavior (instead of phonetic structure) may very well be, as Quine (1990) would have it, methodologically harmless, it is also trivial, as the question of phonemic equivalence has not been answered but only pushed one step
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
27
backwards. On what basis, then, does the informant make these distinctions? Given the failure of phonemic reductionism, even the most patient and phonetically attentive informant will inevitably answer these types of questions on the basis of interpretive semantics: whether the substitution affects a change in meaning or not. The so-called behavioral ‘‘pair-test’’ developed by Zellig Harris (1951) (and, interestingly, favored by a young Chomsky (1957)) necessarily presupposes that at least the informant is accessing semantic information, if the test is not to degenerate into an acoustic test of brute auditory discrimination. In fact, Harris (1951, 20) himself explicitly acknowledged, and Chomsky (1975, 102) would relunctantly concede, that much.6 So while superficially devoid of semantic considerations, these efforts to sanitize the phoneme by probing behavior rather than interpretive meaning represent attempts of philosophical slight of hand – behaviorists maneuvers which did not go unnoticed, or uncriticized some forty years before Quine would pursue the same tactic.7 After all, given that linguistic fieldworkers can (with the aid of sensitive audio and recording equipment) overcome any of their native language auditory biases, the only reason to rely on a native informant in the first place would be to surreptitiously access the very semantic information native speakers, unlike non-native speakers, possess. In fact, the only reason linguists need to rely on native informants is that they, unlike the field worker, already know the semantics of the language. But perhaps this explicit semantic interpretation of the native’s behavior is to, in a sense, beg the question against Quine. Is it not the case that the native informants may have learned to phonologically categorize their language merely on the basis of observing the behavior of others in their language community? So while informants typically have no explanation for their linguistic intuitions regarding what counts or does not count as a phonemically equivalent repetition other than semantic contrast any more than most people, this is arguably irrelevant. What really matters is the origin of these intuitions, not their introspective status. Is it not possible for non-native linguists to ‘‘go native’’ and personally figure out the phonemic structure of a language for themselves merely by observing the behavior of established speakers? Indeed, is that not how the natives themselves must have acquired their phonology? Were we not all originally ‘‘aliens’’ in our own language? In this sense, the answer is of course, yes: going native is always an option. Note, however, that
28
REESE M. HEITNER
this procedure, if true to the principles of a strictly non-semantic behavioral analysis, would not be anything like the typical language learning experience. This procedure is prohibited from leveraging any of the interpretive semantic significance of any one’s behavior (verbal or otherwise). Rather, the only tool that the behaviorist, qua behaviorist, has is behavior. So instead of learning phonemic categorization strictly on the basis of perceptual clues available from the speech stream itself (accessible, as it were, exclusively through a radio or loud speaker, see Heitner 2004 for a discussion of this), this behaviorist proposal is tantamount to merely watching the behavior of language users without, however, attributing any interpretive value to their (speech) behavior. So while such an analysis is surrealistically analogous to watching live drama without attributing meaning to the actors’ behavior and on this basis, trying to figure out what sounds count as the same and which count as different, Quine might contend that this ‘‘mind blind’’ analysis might, in principle, succeed in cracking the phonemic code. But this will not work. For now the question has become: what distinguishes one overt behavior from another? A classificatory standard distinguishing ‘‘assent’’ and ‘‘dissent’’ is required just as much as a classificatory standard distinguishing the phonemic and the nonphonemic. In order to be free from meaning, the behavioral cue distinguishing simple Quinean dissent from assent must be transparently accessible, or at least reducible in principle to some sort of a purely behavioral description – perhaps in terms of physiological movement. But dissent and assent, as relevant here, are fundamentally voluntary, intentional, acts subject to culturally-specific norms of expression. For this reason, judging which behaviors count as the ‘‘same’’ or ‘‘different’’ is subject to the same kind of initial uncertainty as which sounds count as the same or different. The question of ‘‘phonemic equivalence’’ has merely been transformed into a question of what is to count as ‘‘behaviorally equivalent’’, but this is not progress. Behaviorism is not a solution to our problem. Instead, behaviorism is a restatement of the problem. And this is why the short-circuited future, if not over-determined failure, of ‘‘phonemic reductionism’’ is not nearly as interesting as its philosophical repercussions vis-a`-vis Quine’s influential critique of analyticity. For once the appealing but false premise of phonemic reductionism distinguishing basic ‘‘phono-logical’’ equivalence from analytical equivalence is uncovered and rejected, the Quinean case against analyticity looks different.
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
29
Instead of eliminating merely synonymy, phonology too has been threatened. Indeed, the self-imposed philosophical price to pay for not being able to define, even in principle (and certainly never in practice), phonological equivalence in suitably non-semantic terms is to either push the basis of phonemic categorization beyond the pale of genuine scientific legitimacy and forfeit the legitimacy of phonological equivalence, or reinstate synonymous substitutions as another irreducibly semantic fact of natural language. A consistent rejection of substitutions which cannot be read directly from non-interpreted form entails not only the rejection of analytic equivalence but also phonemic equivalence. So either all putative equivalencies between phonetically distinct items – not just those that suspiciously hang between ‘bachelor’ and ‘unmarried man’, but those which relate simple free allophonic variants – would, by parity of reasoning, collapse in virtue of Quineaninspired arguments originally leveled merely against analyticity and synonymy. Or analytic equivalence must, notwithstanding its nonreductive status, be accorded the same status as phonology. The first option maintains a rejection of all non-formal substitutions at the expense of phonology. The second option maintains the integrity of basic phonological substitutions but effectively immunizes analyticity from meaning skepticism despite the fact that synonymy is not strictly recoverable or definable in purely formal non-semantic terms either. Given these two options, perhaps phonological eliminativism is the next big thing. After all, as Boghossian (1999) notes ‘‘Let’s begin with the fact that even Quine has to admit that it is possible for two tokens of the same orthographic type to be synonymous, for that much is presupposed by his own account of logical truth’’8 (343). But once basic phonological equivalence is acknowledged to be, in principle, no less dependent upon non-formal, interpretive considerations (as measured against the purely formal, non-interpretive criteria definitive of the strictly tautological), the observation that analytic equivalence is parasitic on non-formal semantic considerations is no longer philosophically surprising. In fact the semantic basis of phonological equivalence can be viewed as consistent with a form of post-reductionist nominalism.
5. POST-REDUCTIONIST NOMINALISM
At the risk of creating a philosophical caricature, it might be useful to step back and re-approach the interrelationship between phonological
30
REESE M. HEITNER
form and semantic content – words and what they mean – in philosophically exaggerated terms. As nominalist hyperbole would have it, individually distinct objects (‘‘particulars’’) are classified as the ‘‘same’’, not because of some hidden essence or participation in some sort of platonic form, but because they are each labeled with the same word. This would be the ‘‘nom’’ in nominalist. Indeed, the appeal of a hyper-nominalism like this lies precisely in its ability to give a deflationary account of object classification that does not invoke an essentialist ontology of universals or concepts. As Quine (1953c) puts it, There are those who feel that our ability to understand general terms, and to see one concrete object as resembling another, would be inexplicable unless there were universals as objects of apprehension. And there are those who fail to detect, in such an appeal to a realm of entities over and above the concrete objects in space and time, any explanatory value. (102)
To the question why various chairs are all categorized as chairs, nominalistically minded philosophers demand that we resist the temptation to look for the essence of chairhood, but merely note that all chairs are referred to by, or labeled with, the same word. (And to the question why they might all be labeled in the same way comes the answer that they are all somehow ‘‘similar’’ to each other.) Put very crudely, it is our naming practices that create the illusion of categorytypes and ‘‘universals’’ over and above the basic similarity matrices of token-particulars. The consistent nominalist does this for all categories where we might otherwise be tempted to infer the existence of universals or hidden essences. When confronted, however, with the question of how various word-token particulars are classified into word-type categories, nominalists cannot deploy this simple technique. They are in no position to say that various word tokens, like other particulars, are classified as the ‘‘same’’ in virtue of a single label used to refer to them, quite simply because there is no obvious practice of ‘‘labeling labels’’ by which to claim that linguistic tokens themselves fall into their categories in virtue of their labels (save for the linguistic use of phonemic slash bars, and perhaps Wilfrid Sellars’ parallel deployment of ‘‘dot quotation’’9). Nor does the tactic of relying on basic similarity work. Given the language-specific nature of free allophonic variation, similarity does not govern phonemic speech categorization. In this way, by tacitly relying on language as an instrument of object classification so as to avoid an ontology perceived to be
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
31
metaphysically otiose, nominalistically inclined philosophers tend to either implicitly assume, or as with Quine explicitly claim, that the classification of individual speech-tokens is something that more or less takes care of itself, or at any rate, does not presuppose semantic interpretation. Indeed, nominalist considerations like these implicitly encourage, as the late cognitive psychologist John Macnamara observed, a certain dismissive approach to basic phonology: The study of such phonological problems is of central importance to cognitive psychology. Failure to come to grips with them has led two main branches of contemporary philosophy – phenomenology and the logical analysis of language that stems from Frege – into seriously simplistic proposals about concept formation. The standard position is that the problem of how to categorize objects is solved for the child by the names he hears applies to them. That presupposes that the problem of categorizing those phonological entities that we call names is psychologically simpler than that of categorizing nonlinguistic objects. To examine this presupposition will require a detailed analysis of phonological learning. . . . (1982, 85)
Indeed, the classification of individual speech-tokens is taken for granted so that these ‘‘pre-packaged’’ labels can then be used to impose a nominalist organization among the world of objects and properties. It would be, as Quine (1981, 1987) himself has emphasized, circular to claim that words are classificatory instruments (or less tendentiously, distinct words are indicative of distinct objects and properties), if it were to turn out that the individuation of speechtokens themselves into type equivalence speech categories could not be taken for granted but was actually dependent on whether the speech-tokens themselves, as semantically interpreted, refer to the same object or property. In response to this perceived crisis, noted in the philosophical (Quine 1981, 1987), linguistic (Hockett 1942, 1954) as well as developmental (infant speech perception) literature (Werker and Tees 1992), the possibility – or rather the necessity – of classifying linguistic-tokens without appealing to semantic content became conventional wisdom. Awkwardly caught maintaining that speech tokens are classified into phonological types on the objective basis of their intrinsic properties of similarity, a more consistent nominalism resists the temptation to see either intrinsic essences – or even ‘‘similarities’’ – in chairs, dogs, bits of language or, as with Quine, even in behaviors. A more consistent nominalism looks around for what could be used as leverage to extrinsically or ‘‘instrumentally’’ categorize speech-tokens without positing something that all the tokens, in and
32
REESE M. HEITNER
of themselves, must share, nor search for how they must be intrinsically similar. This is not circular but, as the linguist Kenneth Pike (1952) originally argued, cyclical as we are no more required to learn how to classify all of our phonemes all at once any more than we are required to learn how to categorize all objects all at once. Nor is this procedure linguistically invalid or developmentally unrealistic. Rather it is just another example of an opportunistic ‘‘bootstrapping’’ learning process governing both object classification and linguistic development, amenable to experimental manipulation and empirical confirmation (see Heitner 2004). There is, then, philosophically speaking, nothing specifically more mysterious about equating ‘bachelor’ with ‘unmarried man’ than equating free allophonic variants in natural language generally – at least when it comes to the issue of semantics. ‘Bachelor’ and ‘unmarried man’ do not obviously look or sound anything alike, but that is – as some ‘‘neo-Bloomfieldian’’ linguists eventually conceded with respect to free allophonic variation – surprisingly, if not disturbingly to them, irrelevant. Free allophonic variants do not naturally sound alike either. Just ask someone unfamiliar with the basic allophonic rules of English phonology. Indeed, there is no reason to expect the analytic equivalence between ‘bachelor’ and ‘unmarried man’ to be any more non-semantically accessible than the phonological equivalence of free allophonic variants generally. Both phonemic equivalence and analytic equivalence are semantically conditioned, dependent upon interpretive information semantically imposed upon the language rather than perceptually intrinsic to the language (or behavior). This basic symmetry between phonological and analytic equivalence, however, has been overlooked because Quine’s attack on analyticity highlights the semantics of synonymy without taking into account the semantics of phonology. Further obscuring the connection between analyticity and phonology is a widespread assumption among philosophers that phonology is just about the sounds of words of no philosophical relevance. But it is Quine’s own writings on the phoneme intended to free phonology from semantics which reveal that at least Quine was aware of the tactical importance of providing a non-semantic reduction of basic phonological equivalence. Without it, Quine’s argument against analyticity is an argument against phonology. Phonology – the basis of our everyday intuitions governing simple tautological truth – is semantically conditioned. The information which determines which linguistic forms are tautologically
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
33
equivalent is itself at least partially based upon interpretive semantics. For this reason, a linguistically more accurate, a developmentally more realistic, and a philosophically more consistent approach to synonymy must expose this widespread dualism between form-based ‘‘phonological’’ substitutions and meaning-based analytic substitutions – implicit in Quine’s attack on analyticity – as yet another dogma-like presumption within empiricist philosophy of language. Linguists have, in their own way, sought to recover from a semantically irreducible phoneme.10 But it is philosophers – ironically largely responsible for the ‘‘neo-Bloomfieldian’’ pre-occupation (see Koerner 1978, Bloomfield 1944) with ‘‘reducing the phoneme’’ in the first place – who have yet to come to grips with its full implications. ACKNOWLEDGEMENTS
This paper was originally presented at the December 2002 Eastern Division meeting of the American Philosophical Association. I wish to thank my commentator, David A. Truncellito, an especially helpful anonymous reviewer from Synthese, and my dissertation advisor, Prof. Michael Levin, Graduate Center, CUNY for improving the manuscript. NOTES 1
Note that the semantic basis of phonological equivalence challenges the distinction between the semantics of analytic truth and the presumptively non-semantic ‘‘formal’’ status of tautological truth – what Boghossian (1999) calls ‘‘logical truth’’. The semantic basis of phonological equivalence undermines Quine’s attack on analytic equivalence, but the semantic basis of phonological equivalence does not threaten logical relations generally. No wider notion of logical or formal truth, logical formality, deductive reasoning, propositional logic, predicate logic, etc. is globally threatened simply by conceding the semantic basis of phonological equivalence, and hence basic tautological relations. Our repertoire of logico-mathematical terms have an application far beyond that of merely capturing the basic tautological equivalence among token variants of the same type (even if all symbol use is characteristically dependent upon such type equivalencies among token distinct symbols). 2 There are a number on side issues which should be kept in mind. This standard phonological definition of phonemic equivalence does not ignore the resources of non-semantic analyses to account for phonemic equivalence in terms of phonetic similarity or allophonic patterns of complementary distribution. These non-semantic analyses, however, are incomplete. Semantic information is needed to distinguish ‘‘free allophonic’’ variation from ‘‘minimal pair’’ phonemic alternation. See Bloomfield (1933) for a paradigmatic illustration of the role semantics plays in
34
REESE M. HEITNER
recognizing phonemic equivalence. Note also the semantic basis of phonological equivalence does not prohibit the existence of natural language ambiguity in the form of homonyms. The claim that semantics is critical for determining basic phonemic equivalence does not entail that phonemic equivalence determines equivalence of semantic content (homonyms exist). Finally, phonemically equivalent wordtokens appear to be freely substitutable in all contexts, including belief, modal, and even quotational, as even direct quotation does not demand speech reports to be imperceptibly identical to the original (though some forms of quotation can, of course, be required to transmit with such artificial fidelity). 3 For those familiar with the debates between Quine and Strawson, it is important to distinguish Strawson’s argument from the present argument, though both are critical of Quine for relying on an artificial notion of logical truth independent of any semantic considerations. In response to Quine’s attack on analyticity, Strawson (1957) argues that Quine’s own account of logical truth is parasitic on the very notions Quine seeks to dismiss. Strawson does this by pointing out that logical substitutions require one to rule out homonyms and indexical expressions which vary in reference depending upon their semantic interpretation: only substitutions that keep the same reference are logically valid, using the example ‘If he is sick, then he is sick’ to stress that the first occurrence of ‘sick’ can be interpreted as mentally sick while the latter can be interpreted as physically sick, conspiring to make a potentially false conditional. How are such basic logical inferences to be protected from such ambiguity? ‘‘The apparently unavoidable answer to these questions’’ says Strawson, ‘‘is that typographical identity of sentences or recurrences of sentence-letters (as Quine calls them) represent or symbolize identity of statements or propositions or whatever we call those things which are true or false. (The terminology of ‘sentence-letters’ is objectionable in so far as it obscures this important fact.) If this apparently unavoidable answer is right then we have no adequate characterization of logical truth until we have an adequate criterion of identity of statements or propositions . . . . We may, that is to say, characterize the difficulties as follows. Some arise from the fact that typographical identity of referring expressions does not guarantee identity of reference; others arise from the fact that typographical identity of predicate expressions does not guarantee identity of sense’’ (19–20). Though Quine (1960) reports that this criticism is ‘‘an interesting one, which I cannot claim to have answered anywhere’’ (65; footnote 3), at no point does Strawson attempt to undermine Quine’s critique of analyticity by questioning the nature of ‘‘typographical’’ identity itself. So while Quine (1960) expresses reservations about ‘‘embracing the analyticity of the truths of logic as an antecedently intelligible doctrine’’ (65) (and refers the reader to his article ‘‘Truth by convention’’), he does this on the basis of a holistic rejection of a sharp demarcation between truths of logic and truths of fact. Both Quine, and it would seem Strawson too, however, take it for granted that ‘‘typographical identity’’ is an unproblematic notion which does not require the introduction of any semantic concerns in order to be defined. 4 Both Goodman and Quine freely share their own doubts regarding ‘‘similarity’’ and the viability of defining basic phonological equivalence in non-semantic terms. In his ‘‘Seven Strictures on Similarity’’, Goodman (1972) concedes (as his ‘‘Second Stricture’’):
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
35
Similarity does not pick out inscriptions that are ‘tokens of a common type’ or replicas of each other. Only our addiction to similarity deludes us into accepting similarity as the basis for grouping inscriptions into several letters, words and so forth. The idea that inscriptions of the same letter are more alike than inscriptions of different letters evaporates in the glare of such counterexamples as those in Figure 1. a m
d w
A M
Figure 1. One might argue that what counts is not degree of similarity but rather similarity in a certain respect. In what respect, then, must inscriptions be alike to be replicas of one another?. . . We have terrible trouble trying to say how two inscriptions must be alike to be replicas of one another – how an inscription must resemble other inscriptions of the letter a to be itself an a. I suspect that the best we can do is to say that all inscriptions that are a’s must be alike in being a’s. That has the solid ring of assured truth, but is hardly electrifying. Moreover, notice that to say that all a’s are alike in being a’s amounts to simply to saying that all a’s are a’s. The words ‘‘alike in being’’ add nothing; similarity becomes superfluous. (438–439) Interestingly, Goodman goes on to state (in his ‘‘Third Stricture’’) that ‘‘Similarity does not provide the grounds for accounting two occurrences performances of the same work, or repetitions of the same behavior or experiment’’ (439, emphasis added), an observation relevant to Quine’s attempt to provide a strictly behavioral reduction of phonemic equivalence below. As for Quine, after outlining a psychophysical account of phonemic equivalence, Quine (1953a) notes that ‘‘There are abundant reasons to suspect that neither this oversimplified account nor anything remotely resembling it can possibly provide an adequate definition of the phoneme; and phonologists have not neglected to adduce such reasons. As a means of isolating other points of comparison between grammar and lexicography, however, let us make the unrealistic assumption that our grammarian has some such non-semantical definition of the phoneme . . .’’ (53). This concession, however, does not stop him from then immediately re-asserting his philosophical conviction that ‘‘The basic point of view is that the class K is objectively determinate before the grammatical research is begun . . .’’ (53). 5 In his Language Bloomfield (1933) repeatedly repudiates any such non-semantic reductions of the phoneme: ‘‘Only by finding out which utterances are alike in meaning, and which ones are different, can the observer learn to recognize the phonemic distinctions’’ (93). It is ‘‘[f]or this reason even a perfected knowledge of acoustics will not, by itself, give us the phonemic structure of a language. We shall always have to know which of the gross acoustic features are, by virtue of meanings, ‘the same’, and which are different for the speakers’’ (1933, 128). 6 Chomsky (1975) – apparently contradicting or at least retracting an earlier claim that ‘‘there is no warrant for interpreting the responses studied in the pair test as semantic in any way’’ (1957, 99) – grudgingly concedes that ‘‘in a certain sense the ultimate criterion remains the speaker’s intuition about linguistic form, since only this can tell us which behavioral tests are to the point. We might hope that some
36
REESE M. HEITNER
more general account of the whole process of linguistic communication than we possess now may permit us to reconstruct the criteria of adequacy for linguistic theory in more convincing and acceptable terms. But for the present, it seems that we must rely, at least to some extent, on the speaker’s intuitive conception of linguistic form’’ (1975, 102). This acceptance of speaker intuition, however, actually represents a significant concession for Chomsky, as he was, at least initially, no more satisfied with an intuitive (i.e., semantic) characterization of phonemic equivalence than the ‘‘taxonomic’’ Bloomfieldian linguists he doggedly criticized (see Chomsky 1957, 1975). In fact, allusions to Chomsky represent somewhat of an occupational hazard for historians of linguistics (see Steinberg 1999) with Chomsky once favorably calling on W. V. Quine of all people for additional theoretical support (see Chomsky 1975, 85–86). See also Joseph (2002) and the references therein for more details about Chomsky’s ‘‘quirkiness as a historian’’ (165). 7 As Clark and Yallop (1995) emphasize in a standard text book in phonology, ‘‘It is now generally agreed that the classic attempt to produce phonological descriptions that would make no reference to the meanings of words, let alone to native speakers’ intuitions or insights is indeed inconsistent. Harris’ ‘Methods in Linguistics’ [sic] (1951) represents the claim that it is possible to discover phonemes purely by examining the distribution of phonetic segments. . . But Harris’ analysis in fact assumes the investigators ability to judge whether two utterances in a language are intended to be different words or whether they count as alternative ways of saying the same word. It can be argued that Harris’ and others’ efforts to define ‘objective’ analytic procedures constantly presuppose access to native speakers intuitions into their own language’’ (83). 8 This quote from Boghossian (1999) continues with: ‘‘As we saw in the passage I quoted above, Quine describes a truth of logic as: ‘a statement which is true and which remains true under all reinterpretations of its components other than the logical particles.’ Clearly, the idea isn’t that such a statement will remain true no matter how the non-logical particles are substituted, but rather that it will remain true provided that the non-logical particles are substituted in a uniform way, with multiple occurrences of the same word receiving the same substitution in every case. But what should we count as the same here?. . . So even Quine has to admit – what in any event seems independently compelling – that two tokens of the same type can express the same meaning’’ (343–344). But Boghossian’s argument seems incomplete. What’s important is not so much ‘‘that two tokens of the same type can express the same meaning’’, but that in order to first recognize and classify tokens as logically substitutional, semantic information is required. The observation that two tokens happen to express the same semantic is relatively harmless. The important point is that without the ‘‘same semantics’’ (that is quantifying over meanings) distinct tokens cannot be comprehensively equated even as tautologously ‘‘phono-logical’’ equivalent; simply identifying re-occurrences of the same word-type is not possible without some reliance on semantics. 9 In fact, ‘‘labeling labels’’ is arguably precisely what phonologists do when they label phonetically distinct token-words as the same phonological word-type. Moreover, the use of phonological slash bars (/top/) is the meta-language notation used to represent ‘toph’ (aspirated) and ‘top’ (non-aspirated) as equivalent despite their phonetic differences. But this notational technique may also be implemented at a higher level to concretely refer to the single linguistic type phonologically
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
37
distinct synonyms (and translations) uniformly instantiate. Indeed, this is the purpose of Sellars’ dot quotation. See Sellars (1950, 1967) and Heitner (2003). 10 Not so much, however, by accepting a semantically defined phoneme, but simply by moving on to different concerns. The Chomskyan framework of generative phonology neither provides, nor aspires to provide, any account of basic phonological classification (semantic-based or otherwise). So while fundamentally rejecting the ‘‘phoneme’’ as an independent unit of linguistic analysis and representation, Chomsky is content to treat phonological equivalence, and its opposite, phonological contrast, as ‘‘primitive’’ notions (Chomsky 1964, 83). See also Postal (1968) who also wishes to treat phonemic categorization as primitive rather than engage in any substantive discussion regarding the basis of these language-specific phonological categories.
REFERENCES Bennett, J.: 1988, ‘Quotation’, Nous 22, 399–418. Bloch, B.: 1953, ‘Contrast’, Language 29, 59–61. Bloch, B. and G. Trager: 1942, Outline of Linguistic Analysis, Waverly Press, Baltimore, MD. Bloomfield, L.: 1933, Language, The University of Chicago Press, Chicago. Bloomfield, L.: 1944, ‘Secondary and Tertiary Responses to Language’, Language 20, 45–55. Boghossian, P.: 1999, ‘Analyticity’, in C. Wright and B. Hale (eds.), A Companion to the Philosophy of Language, Cambridge: Blackwell, pp. 331–368. Carnap, R.: 1937, The Logical Syntax of Language, Patterson, Adams & Co., NJ, Littlefield. Carnap, R.: 1947, Meaning and Necessity, The University of Chicago Press, Chicago. Chomsky, N.: 1957, Syntactic Structures, Mouton & Co., The Hague. Chomsky, N.: 1964, Current Issues in Linguistic Theory, Mouton, The Hague. Chomsky, N.: 1975, The Logical Structure of Linguistic Theory, The University of Chicago Press, Chicago. Church, A.: 1954, ‘Intensional Isomorphism and the Identity of Belief ’, Philosophical Studies 5, 65–73. Reprinted in Propositions and Attitudes, N. Salmon and S. Soames, S. (eds.), 159–168. Clark, J. and C. Yallop: 1995, An Introduction to Phonetics and Phonology, Blackwell, New York. Fought, J.: 1999a, Leonard Bloomfield: Critical Assessments of Leading Linguists. Vol. I: Biographical Sketches; Vol. II: Reviews and Meaning; Vol. III: Phonology, Morphology and Languages, Routledge, New York. Fought, J.: 1999b, ‘Leonard Bloomfield’s Linguistic Legacy’, Historiographia Linguistica 26, 313–332. Goodman, N.: 1972, Problems and Projects, The Bobbs-Merrill Company Inc., Indianapolis. Goodman, N. and W. V. Quine.: 1947, ‘Steps Toward a Constructive Nominalism’, The Journal of Symbolic Logic 12, 105–122. Harris, Z.: 1951, Structural Linguistics, University of Chicago Press, Chicago. Heitner, R.: 2003, ‘Brackets and Slashes; Stars and Dots: Understanding the Notation of Linguistic Types’, Language Sciences 25, 319–330.
38
REESE M. HEITNER
Heitner, R.: 2004, ‘The Cyclical Ontogeny of Ontology: An Integrated Developmental Account of Speech and Object Categorization’, Philosophical Psychology 17, 45–57. Hockett, C.: 1942, ‘A System of Descriptive Phonology’, Language 18, 3–21. Hockett, C.: 1954, ‘Two Models of Grammatical Description’, Word 10, 210–234. Hymes, D. and J. Fought: 1981, American Structuralism, Mouton Publishers, The Hague. Joseph, J.: 2002, From Whitney to Chomsky: Essays in the History of American Linguistics, John Benjamins Company, Amsterdam. Koerner, E. F. K.: 1978, ‘Bloomfieldian Linguistics and the Problem of Meaning: A Chapter in the History of the Theory and Study of Language’, in E. Koerner (ed.), Toward a Historiography of Linguistics: Selected Essays, John Benjamins, Amsterdam, pp. 155–176. Linsky, L.: 1952, Semantics and the Philosophy of Language, University of Illinois Press, Urbana, IL. Macnamara, J.: 1982, Names for Things: A Study of Human Learning, The MIT Press, Cambridge, MA. Mates, B.: 1950, ‘Synonymity’, in L. Linsky (ed.), Semantics and the Philosophy of Language, 111–136. Pike, K.: 1952, ‘More on Grammatical Prerequisites’, Word 8, 106–121. Pike, K.; 1958, ‘Interpenetration of Phonology, Morphology, and Syntax’, in E. Sivertsen (ed.), Proceedings of the Eighth International Congress of Linguistics, Oslo University Press, Oslo, 363–387. Postal, P.: 1968, Aspects of Phonological Theory, Harper & Row Publishers, New York. Putnam, H.: 1961, ‘Some issues in the theory of grammar’, in R. Jakobson (ed.), The Structure of Language and its Mathematical Aspects, American Mathematical Society, Providence, RI: pp. 25–42. Quine, W. V.: 1951, ‘Two Dogmas of Empiricism’, Philosophical Review 60, 20–43. Reprinted in From a Logical Point of View, 20–46. Quine, W. V.: 1953a, ‘The Problem of Meaning in Linguistics’, From a Logical Point of View, 47–64. Quine, W. V.: 1953b, ‘Mr. Strawson on Logical Theory’, The Ways of Paradox and Other Essays, 135–155. Quine, W. V.: 1953c, ‘Logic and the Reification of Universals’, Reprinted in From a Logical Point of View, 102–129. Quine, W. V.: 1960, Word and Object, MIT Press, Cambridge, MA. Quine, W. V.: 1966, The Ways of Paradox and Other Essays, Random House Inc., New York. Quine, W. V.: 1980, From a Logical Point of View, Harvard University Press, Cambridge, MA. Quine, W. V.: 1981, ‘Use and Its Place in Meaning’, Theories and Things, Harvard University Press, Cambridge, MA, 43–54. Quine, W. V.: 1987, Quiditties: An Intermittently Philosophical Dictionary, Harvard University Press, Cambridge, MA. Quine, W. V.: 1990, ‘The Phoneme’s Long Shadow’, in T. Headland, K. Pike, and M. Harris (eds.), Emics and Etics: The Insider/Outsider Debate, Sage Publications, Newbury Park, CA, pp. 164–167.
NEUTRALIZING QUINE’S ARGUMENT AGAINST ANALYTICITY
39
Ramsey, F.: 1950, Foundations of Mathematics and other Logical Essays, The Humanities Press, New York. Salmon, N. and S. Soames: 1988, Propositions and Attitudes, Oxford University Press. Sellars, W.: 1950, ‘The Identity of Linguistic Expressions and the Paradox of Analysis’, Philosophical Studies 1, 24–31. Sellars, W.: 1967, ‘Abstract entities’, in Philosophical Perspectives, Charles C. Thomas, Publisher, Springfield, IL, pp. 229–269. Steinberg, D.: 1999, ‘How the Anti-mentalistic Skeletons in Chomsky’s Closet make Psychological Fiction of his Grammars, in S. Embleton, J. Joseph, and H. Niederehe (eds.), The Emergence of the Modern Language Sciences: Studies on the Transition from Historical-Comparative to Structural Linguistics in Honour of E. F. Konrad Koerner, Vol. 1: Historiographical Perspectives, John Benjamins, Amsterdam, pp. 267–282. Strange, W.: 1995, ‘Cross-language Studies of Speech Perception: A Historical Review’, in W. Strange (ed.), Speech Perception and Linguistic Experience: Issues in Cross-Language Research, York Press, Timonium, MD, pp. 3–45. Strawson, P.: 1957, ‘Propositions, Concepts and Logical Truths’, The Philosophical Quarterly 7, 15–25. Twaddell, W. F.: 1935, ‘On Defining the Phoneme’, Language Monograph, No. 16. Reprinted in Readings in Linguistics I: The Development of Descriptive Linguistics in America (1925–1956), M. Joos (ed.), The University of Chicago Press, Chicago, 1957, pp. 55–79. Ullmann, S.: 1957, The Principles of Semantics, Jackson, Glasgow. Werker, J. and R. Tees: 1992, ‘The Organization and Reorganization of Human Speech Perception’, Annual Review of Neuroscience 15, 377–402. Graduate Center, CUNY New York, NY U.S.A. E-mail:
[email protected]