This action might not be possible to undo. Are you sure you want to continue?
53-68 Published by: Springer Stable URL: http://www.jstor.org/stable/20114630 Accessed: 28/12/2009 15:53
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use. Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=springer. Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact firstname.lastname@example.org.
Springer is collaborating with JSTOR to digitize, preserve and extend access to Synthese.
of what appears to be Perhaps the clearest and most explicit development a narrowly Humean of language acquisition in recent philosophy theory is that of Quine, in the introductory to his Word and Object.1 chapters then a person's knowledge theory is roughly accurate, as a network of linguistic forms of language should be representable sentences - associated with one an let us say, to first approximation, to certain stimulus conditions. other and, in part, associated This for If the Humean mulation Quine presents as, I take it, a factual assertion. Thus he states - whether or "second as chemistry, that our "theories" "deliberate", of ordinary doctrine nature", as "the immemorial enduring middle as "a fabric of sentences sized objects" - can each be characterized to one another and to non-verbal stimuli by the variously associated mechanism of conditioned ledge (our total "theory", One difficulty that arises in interpreting such passages as these has to do with the relation between language and theory, where the latter term covers also general common-sense and belief. Quine's views knowledge about the interp?n?tration of theory and language are well known, but, even accepting them fully, one could not doubt that a person's language and his but 'theory' are distinct systems. The point is too obvious to press, it is, nevertheless, difficult to see how Quine distinguishes the two in his framework. In fact, throughout the discussion, he seems to use response" (p. 11). Hence the whole of our know in this sense) can be characterized in these terms.
the terms interchangeably. For example, in Chapter 1, he discusses the it by an example from learning of language in general terms, exemplifies chemical theory leading up to the statement just quoted, then seemingly describes work the "vast verbal that constitutes one's so constructed, the associative net of science ("and indeed every knowledge as both the "body of theory" that the world"), structure"
thing we ever say about one accepts and the language that one learns. Thus the discussion of how one constructs and uses a total theory of this sort concludes with the following statement: 53
Synthese 19 (1968-69) 53-68. ? D. Reidel Publishing Co., Dordrecht Holland
Beneath personal to evolve. the uniformity that unites and, us in communication for each alike, of us, nor, there the connections in a sense, does is a chaotic continue any finish
diversity two No
of connections, of us learn our
learning it while he lives. Since the comment fabric" merely summarizes our the discussion of how the "single (the latter
total theory is acquired constituting itself having been introduced to exemplify language learning), it seems that Quine must be proposing that a language, too, is "a fabric
stimuli variously associated to one another and to non-verbal the mechanism of conditioned Other parts of his exposition by response". reinforce the conclusion that this is what is intended, as we shall see in a moment. Nevertheless, interpretation of Quine's remarks is made difficult of sentences at points to use the terms 'language' and 'theory' a fundamen interchangeably, though obviously he must be presupposing tal difference between the two - he is, for example, surely not proposing because of his tendency
that two monolingual speakers of the same language cannot disagree on questions of belief, or that controversy as ir over facts is necessarily rational as an argument between a monolingual of English and speaker a monolingual speaker of German. states that he is considering a language as a "complex of present dispositions to verbal behavior, in which speakers of the same language have perforce come to resemble one another" (p. 27). Thus if a language is a network of sentences associated to one another and to Elsewhere, Quine external follows stimuli by the mechanism then it of conditioned response, that a person's disposition to verbal behavior can be characterized in terms of such a network. This factual assumption is far from obvious. of 'language' below.
a Humean acquired? Evidently, will acquire substance only if such notions as 'similarity' are theory a prelinguistic in some way. Quine characterized therefore postulates with a built-in distance measure (and presumably innate) "quality space" the structure of this space will determine the content (p. 83-4). Evidently, a of the theory of learning. For example, one could easily construct theory of innate ideas of a rather classical sort in terms of a prelinguistic quality space with a build-in distance measure. Quine would, apparently, accept a very strong version of a theory of innate ideas as compatible with his framework. Thus he considers the possibility that "a red ball, 54
I return to other aspects of this concept How is knowledge of such a language
a yellow ball, and a green ball are less distant from one another in... It is difficult to the child's... quality space than from a red kerchief". see how this differs from the assumption that 'ball' is an innate idea, if we admit the same possibilities along other 'dimensions' (particularly, to be fairly abstract). In this respect, then, Quine seems to depart quite radically from the leading ideas that guided so far imaginable, empiricist theory and to permit just about anything as 'learning' of concepts In particular, consider the fact is concerned. these dimensions that a speaker of English has acquired the concept 'sentence of English'. Suppose that we were to postulate an innate quality space with a struc ture so abstract that any two sentences of English are nearer to one than a sentence of in terms of the postulated distance measure Then a learner could and any sentence of another English language. 'sentence of English' - he could, in other words, acquire the concept know that the language to which he is exposed is English and 'generalize' to any other sentence of English - from an exposure to one sentence. The same is true if we mean by 'sentence of English' a pairing of a certain another phonetic a quality interpretation. We could, once again, construct space sufficiently abstract so that the infinite set of English sentences could be 'learned' from exposure to one sentence, by an organ ism equipped with this quality space. and semantic if we allow
The handful of examples and references that Quine gives suggests that he has something much narrower inmind, however; perhaps, a restriction to dimensions which have some simple physical correlate such as hue or brightness, with distance defined in terms of these physical correlates. If so, we have a very strong and quite specific version of a doctrine innate ideas which now can be faced with empirical evidence. It might of
be thought that Quine adds empirical content to his account that "the child's early learning of a verbal response by his insistence of the response in association with depends on society's reinforcement the stimulations that merit the response..." (p. 82) and his general in sistence throughout 'reinforcement' Quine's fortunately, For example, he is willing to accept the possibility vacuity. than corroborative consists in no more reinforcement resemblance that learning to the child's requires effort is the sole reward" then, comes reinforcement, that learning of concept is based on reinforcement. But, un to near is reduced that "society's usage, whose
(p. 82-3). To say very close to saying 55
NOAM CHOMSKY that learning cannot proceed without
is "congenial... to Skinner's scheme,
data. As Quine
notes, his approach
does not enume
is correct, but it should also be added that "Skinner's scheme" is almost totally empty, in fact, if anything, even less substantive than Quine's version of it, since Skinner, as distinct The remark from Quine, does not even require that reinforcing stimuli impinge on - it is sufficient that the organism they be imagined, hoped for, etc. serves only a ritualistic In general, the invoking of 'reinforcement' function trying in such discussions to determine as these, and one can safely disregard it in content of what is being proposed. the substantive a classical empiricist conception of a non returns to
rate the rewards".
However, Quine vacuous sort in his assumptions about how with his view of language as a network
language is learned. Consistent of sentences2, he enumerates three possible mechanisms which sentences can be learned - i.e., by by of language can be acquired (p. 9f.). First, sentences which knowledge can be learned by "direct conditioning" to appropriate non-verbal that is, by repeated pairing of a stimulation and a sentence stimulations, under of sentences with conditions ; second, by association appropriate sentences (let us put aside the objection that in both cases, the associations should soon disappear, through extinction, under normal circumstances) ; third, new sentences can be produced by "analogical synthesis".3 The at first seems to offer an escape to vacuity, once again. third method
if the first sentence of this paper is derivable by analogical synthesis from "the sky is blue" (both involve subject and predicate, are generated with their interpretations by the rules of English grammar, and share then it is no doubt true that language can be many other properties), Thus of learned by 'analogical synthesis', by 'generalization' along a dimension the abstract sort suggested above But it seems clear (cf. p. 55). that Quine has nothing of this sort in mind. The one example that he is a case of substitution in a fixed context. And of one word he seems for a similar to imply one ('hand', that the process of simply serving to speed
'foot') synthesis is theoretically analogical dispensable, matters up (see p. 9). Therefore, we can perhaps conform to his inten tions by totally disregarding the knowledge this process, and considering attained
instead by a long-lived adult using only the first two methods of the knowledge attained by a young child who has used all three (there that can be said about the latter case until the notion being nothing 56
further that a child 'analogical synthesis' is given some content). Noting of nine and a man of ninety share knowledge of language in fundamental - each can understand an astronomical and use appropriately respects - it would number of sentences, for example seem, further, that little is lost in omitting 'analogical synthesis' from consideration entirely, even for the young child. Assuming that this interpretation of Quine's remarks is correct, we derive support for the conclusion that he regards a language as a finite network some associated of associated also to sentences, is just the structure that would arise from the two of language learning with substantive content. postulated this interpretation of Quine's remarks on language we can Against bring the fact that it is inconsistent with a truism that he of course stimuli, this mechanisms that a language is an infinite set of sentences accepts, namely, (with intrinsic meanings; A network derived by the postulated cf., e.g., p. 71). must be finite; it can, in fact, contain only the sentences mechanisms to which a person has been exposed (repeatedly, and under similar cir If we return to the definition of 'language' as a "complex cumstances). of dispositions at to verbal behavior", we reach a similar conclusion, least if this notion is intended to have empirical content. Presumably, a complex of dispositions as a is a structure that can be represented set of probabilities for utterances in certain definable 'circumstances' or 'situations'. But it must be recognized that the notion 'probability is an entirely useless one, under any known interpreta tion of this term. On empirical grounds, the probability of my producing some given sentence of English - say, this sentence, or the sentence or whatever - is indistinguish "birds fly" or "Tuesday follows Monday", able from the probability of my producing a given sentence of Japanese. of the notion of 'probability relative to a situation' changes Introduction on any known objective nothing, at least if 'situations' are characterized of course, raise the conditional of any grounds (we can, probability sentence as high as we like, say to unity, relative to 'situations' specified on ad hoc, invented grounds). Hence if a language is a totality of speech then dispositions (in some empirically significant sense of this notion), my language either does not include the sentences just cited as examples, or it includes all of Japanese. In fact if the "complex of is dispositions" on grounds of empirical observation, determined then only a few con ventional greetings, cliches, and so on, have much chance of being 57 as of a sentence' since
sociated to the complex defining the language, since few other sentences are likely to have a non-null relative frequency, in the technical sense, - we or set of observations in any reasonable corpus would, for example, expect the attested frequency of any given sentence to decrease without limit as a corpus One might increases, under any but the most artificial conditions. to sentences on imagine other ways of assigning probabilities so far as I can see, that avoid these diffi grounds, but none,
empirical culties. Hence if a language is a complex of dispositions to respond under a normal set of circumstances, it would be not only finite (unless it included all languages) but also extremely small. "speech lates the problem of "indeterminacy of translation" as resulting from the one language into another can be for translating fact that "manuals set up in divergent ways, all compatible with the totality of speech one another" (p. 27). As just noted, yet incompatible with dispositions, if we take the "totality of speech dispositions" of an individual to be for utterances under detectable by probability distributions then the thesis quoted is true, near-vacuously, since conditions, for a trivial set, all such probabilities will be empirically indis except on empirical grounds, within or outside of the language. tinguishable characterized stimulus the other hand, ifwe interpret the notions 'disposition' and 'situation' more loosely, itmight be argued that the problem is really quite different, that there will be so few similarities in what they among individuals are inclined to say in given circumstances that no manual of translation can be set up at all, compatible with such inclinations. Actually, Quine On in his exposition, these problems, by shifting his ground from of speech dispositions" to "stimulus meanings", that is, dis "totality to "assent or dissent" in a situation determined by one narrowly positions circumscribed experiment. He even goes so far as to say that this ar avoids bitrarily selected experiment provides all of the evidence that is available, in principle, to the linguist (equivalently, to the language learner - p. 39). to verbal response" under Clearly, however, a person's total "disposition is not the same as his "dispositions to be stimulus conditions arbitrary or to dissent from the sentence" under the particular to assent prompted conditions argue 58 of the Gedankenexperiment that Quine outlines. One might that by arbitrarily the "totality of evidence", Quine ir limiting Adding somewhat to the confusion is the fact that Quine in his use of the notion appears to vacillate Thus he formu dispositions".
relevantly establishes the thesis that alternative theories (manuals of trans with all of the evidence (though the general lation) exist compatible is nevertheless of translation thesis of indeterminacy certainly true, in a sense to which we return in a moment). But my point here is only that this kind of vacillation makes it still more difficult to determine what Quine means by 'disposition' It is easy to imagine a way or 'language'. out of the difficulties
by the implied if the notion finiteness of language and knowledge (or near emptiness, assume that of 'disposition' is taken very seriously). Thus one might in the widest of a 'universal grammar', sense, is an innate knowledge posed
property of the mind, and that this given system of rules and principles of infinitely many sentences (and the determines the form and meaning and belief) from the minute experiential infinite scope of our knowledge base that is actually available to us. I do not doubt that this approach is quite reasonable, but it then raises the empirical question of the nature a priori system; and, of course, any philosophical of this universal, conclusions that may be drawn will depend on the answers proposed for this question.4 Quine's attitude towards an approach of this sort is not It certainly seems inconsistent with his general point easy to determine. of logical of view, specifically, with his claim that even our knowledge that associate certain pairs truths is derived by conditioning mechanisms of logical relations of sentences (cf., e.g., p. llf.), so that our knowledge must sentences. as a finite network of interconnected be representable we can distinguish from causal ones, or either connections (How logical in our to be paired by accident type from sentences which happen is unclear, just as it is unclear how either sort of knowledge experience can be applied, but it is pointless to pursue this issue in the light of the strangeness of the whole conception.) Elsewhere, however, Quine appears to take the view that truth-functional logic might provide a kind of lend asserts (p. 13) that truth functions 'universal grammar'. Thus he "unverifiable without translation" analytical can be learned directly from the available evi hypotheses", dence. He gives no real argument for this beyond the statement, which appears quite irrelevant to the factual issue involved, that we can state themselves and hence truth conditions we can observe functional in terms of assent and dissent. underlying requires The inference from what to a postulated connectives of course structure assumptions truth involving that go beyond 59 to "radical
NOAM CHOMSKY alternatives evidence - mutually incompatible can easily be constructed. Hence Quine's matters of radical consistent with the evidence
to place these willingness indicates within the framework translation perhaps that he is willing to regard the system of truth-functional logic as avail as a basis for language-learning. If of experience, able, independently it seems quite arbitrary to accept this framework as innate schematism, so, to admit much else that can be
In imagined and described.5 of view of the unclarity of this matter, and the apparent inconsistency of the proposal discussed with Quine's explicit characterization just for acquiring them, I will 'theory' and 'language' and the mechanisms and not of this topic. put aside any further consideration are left with the fact that Quine develops his explicit notion of We framework 'language' and 'theory' within a narrowly conceived Humean intrusion of a rich system of innate ideas), and (except for the possible in a that he characterizes language learning ("learning of sentences") the conclusion this narrow interpretation, way consistent with although that a language (or theory) is a finite fabric of sentences, constructed detectable by training, or a set of sentences with empirically of being produced (hence a nearly empty set) is incompatible probabilities with various truisms to which Quine would certainly agree. of about the acquisition Quine relies on his empirical assumptions pairwise knowledge philosophical Fundamental and learning conclusions. of One language critical example to support some of his major will serve to illustrate.
are certain "analytical hypotheses" to knowledge that go is that the correctness the evidence. A crucial point, for Quine, beyond of analytical hypotheses, in the case of ordinary language and "common sense knowledge", is not "an objective matter" that one can be "right or wrong about". These analytical hypotheses "exceed anything implicit to speech behavior". Therefore, when we use in any native's disposition
these analytical hypotheses trivial cases) (as we must, beyond the most in translating, in learning a language in the first place, or in interpreting we "impute our sense what is said to us under normal circumstances, to the native mind". The imputation is of linguistic analogy unverifiably 'unverifiable' in the sense that alternatives consistent with the data are that is, it is 'strong verifiability' that is in question. "There conceivable; can be no doubt that rival systems of analytical hypotheses can fit the to perfection, of speech behavior and can fit the totality of totality 60
QUINE'S to speech translations
dispositions compatible ent control"
determinacy To understand Quine
as well, and still specify mutually in sentences insusceptible of independ of countless (p. 72). These remarks Quine puts forth as the thesis of "in of translation". behavior
to bear in mind that the thesis clearly it is necessary between the construction of analytical distinguishes sharply
of "stimulus meanings hypotheses on the basis of data and the postulation on the basis of data. The latter, he states, of observation sentences" involves only uncertainty of the "normal inductive" kind (p. 68). The same the inductive inference involved in translation of sentences containing and understanding) truth 'learning' (similarly, functional In these cases, induction connectives. leads us to "genuine which are to be sharply distinguished from the "analytical hypotheses", to which reference ismade in the discussion of indeterminacy hypotheses" about of translation. induction', Hence serious epistemological and problem, formation' or 'theory construction', which does involve such Such a distinction can no doubt be made; its point, however, is less than obvious. It is not clear what Quine is presupposing when he over the "normal uncertainty of induction" as within the range passes 'hypothesis a problem. If clarified, this would add more content to his of knowledge, of the a by specification empirical theory of acquisition on which 'normal induction' and the notions of relevant priori properties and sufficient evidence are based. It would then be necessary for him to of radical translation. that the mind is natively endowed with justify the empirical assumption the properties that permit 'normal induction' to 'genuine hypotheses', but not 'theory construction' with some perhaps narrowly constrained class of "analytical hypotheses". to the thesis of indeterminacy of translation, there can statement about analytical hypotheses be no doubt that Quine's surely is true, though the question arises why it is important. It is, to be sure, that if a system of "analytical undeniable goes beyond hypotheses" evidence then it is possible to conceive alternatives compatible with the as in the case of Quine's about evidence, just "genuine hypotheses" stimulus meaning and truth-functional connectives. Thus the situation sense knowledge", in the case of language, or "common is, in this no different from the case of physics. Accepting respect, terms, Quine's 61 To return which Quine involves has no in mind a distinction between 'normal is true, apparently,
NOAM CHOMSKY for the purpose of discussion, we might say that "just as we may meaning the terms of some fully speak of the truth of a sentence only within or conceptual scheme, so on the whole we may meaningfully speak theory the terms of some particular of interlinguistic synonymy only within system of analytical
of a sentence
(p. 75). But, Quine
To be thus reassured is to misjudge
only ... In within
the parallel. In being able to speak of truth
theory, truth is not much hampered; fixed most of
for one is always working with
tentative. short, are
conveniently the meaning
theory, however the parameter
of a foreigner's
the time. Not
so the analytical hypotheses
always ready to wonder
remark without reference to any one set of analytical hypotheses, indeed even in the absence of any; yet two sets of analytical hypotheses equally compatible with all linguistic behavior can give contrary answers, unless the remark is one of the limited sorts that can be translated without recourse to analytical hypo theses (p. 75-6). Thus what the case of physics from the case of language distinguishes is that we are, for some reason, not permitted to have a "tentative theory" in the case of language (except for the 'normal inductive cases' mentioned concerning above). There can be no fixed set of analytical hypotheses language in general. We need a new set for each language (to be more
precise, for each speaker of each language), there being nothing universal about the form of language. This problem, then, is one that faces the sense the child learning a language (or acquiring "common linguist, and interconnection between these processes), given the knowledge", own language.6 in his the person who hears or reads something Quine supposes an innate quality space with a built-in to certain "obvious" that is, apparently, correlated certain kinds of inductive operations physical properties. Furthermore, in this quality space) are based on (involving, perhaps, generalization To summarize, distance measure innate properties of the mind, as are also, perhaps, certain elements of the child (or the linguist these properties, truth-functional logic. Utilizing can form certain genuine hypotheses, which radical translation) doing about stimulus meanings might be wrong but are at least right-or-wrong, and truth-functional quisition another results 62 of knowledge) and to certain connectives. this, language-learning (ac Beyond of sentences to one is a matter of association stimuli a process which through conditioning, interconnected of sentences, or, perhaps, a
in a certain network
to respond. Language certain system of dispositions learning is a matter It is impossible to make of "learning of sentences". significant general statements about language or common-sense theories, and the child has no concept of language or of "common sense" available to him prior to his training. In this respect, the study of language is different from, let us tentative the framework of a say, physics. The physicist works within theory. The linguist cannot, nor can the psychologist studying a of the 'common sense' variety, just as the child can 'conceptual system' 'tentative
theory' that guides him in learning from experience. noted above, this is a relatively Apart from difficulties of interpretation of a classical empiricist doctrine. It involves, at every clear formulation or may which may certain not be step, empirical assumptions seem to regard evidence does not as but for which Quine true,
Let us briefly consider these empirical assumptions. It is, first of all, not at all obvious that the potential concepts of ordinary language are in terms of simple physical dimensions characterizable of the kind Quine appears to presuppose or, conversely, that concepts characterizable in terms of such properties are potential concepts of ordinary language. 'house' is characterized, It is a question of fact whether the concept for a speaker of a natural language, as a 'region' in a space of physical or, as Aristotle dimensions, suggested, in terms of its function within a matrix of certain human needs and actions. The same is true of many other concepts, even the most primitive. Is a knife, to a child with normal or an object experience, an object of such and such physical properties, or is it defined by an amalgam that is used for such and such purposes; of such factors, say as an object meeting certain loose physical conditions that is used for a certain sort of cutting? How would we in fact identify an object looking exactly like a knife but used for some totally different purpose in some other culture?7 This is as much an empirical question as in terms of a region in a the question whether concepts characterized can be acquired in the way a child of simple physical dimensions space but acquires his concepts. There is much to be said in this connection8, it is enough to note, in the present context, that Quine's empirical as - more sumptions may well be (I believe, certainly are) far too strong - and that too strong in the wrong direction they embody correctly, certain quite gratuitous factual assumptions. 63
NOAM CHOMSKY consider the idea that 'similarity' in a sense appropriate the kind of 'similarity' needed for an empirical theory is definable in terms of distance in a certain space of There
Furthermore, for psychology,
of generalization, physical dimensions. Two
is nothing obvious about this assumption. of a three-dimensional two-dimensional projections object may be that has an appropriate 'similar', in the relevant sense, for an organism
and an in concept of the three-dimensional object and its properties of projection, tuitive grasp of the principles there is no di although sort along which such stimulations match. We mension of the presupposed could easily design an automaton which would generalize from one such to another, but not from one of these to a projection of presentation the first in some simple object that matched of course, describe the behaviour of this could, physical automaton in terms of a more abstract quality space, just as we could that learned English from a single sentence in describe an automaton these terms - see p. 55, above. But this is only to say that it is an empirical problem, quite open for the time being, to determine what are the innate of mind that determine the nature of experience and the properties dimension. We content of what comes to be known sentences" on the basis of (or independently the entire notion a scene as rather of) this experience. of As far as "learning seems almost unintelligible. some other three-dimensional
Suppose like the view from my study window, except for the lake in the distance. Am I capable of this because I have learned the sentence: "This scene is rather like the view distance"?To and other any useful point, but from my study window, except for the lake in the say this would be as absurd as to suppose that I form this in sentences of ordinary life by "analogical substitution", sense of this term. It seems hardly necessary to belabor the
is concerned, that I describe
surely it is clear that when we learn a language we are not a "behavioral or acquiring sentences" through "learning repertoire" we somehow develop certain principles (unconscious, training. Rather, of course) that determine the form and meaning of indefinitely many sentences.
sense of knowledge of language (or "common A description as an associative net constructed by conditioned response is knowledge") in sharp conflict with whatever the use of the term Similarly, present 64 dispositions to verbal evidence we have about these matters. of 'language' to refer to the "complex in which speakers of the same behavior,
EMPIRICAL ASSUMPTIONS one another" noted earlier seems rather per (pp. 57-58) have
language have perforce come to resemble verse. Assuming even that the problems been overcome, makes
what point can there be to a definition of 'language' that brain lesions, eye injuries, language vary with mood, personality, nutritional and belief, in the way in which level, knowledge gullibility, to respond" will vary under these and numerous other "dispositions irrelevant in much conditions.9 behaviorist What is involved here is a confusion to be found
To mention discussion. just one further example, consider Quine's remarks on synonymy in his 'Meaning in Linguistics'.10 Here he proposed that synonymy in approximate consists "roughly likeness in the situations which evoke two forms and approximate likeness in the effect on the hearer". to refer If we take the terms "situation" and "effect" objective involving emotional to something that can be specified in terms of as Quine would surely intend (say as physical properties, or observable stimulus conditions and observable behavior
then the qualifications in the characteriza state, respectively), seem misplaced, tion of synonymy for there is not even just quoted likeness in the conditions that are likely to elicit (or to approximate serve as occasion for) synonymous or in the effects of such utterances, utterances. that I see someone about to fall down the stairs. Suppose What would be the probability of my saying :"Watch out, you'll fall down the series of steps, arranged one behind and above the other, in such a way as to permit ascent or descent from one level to another"; and what would the effect on the hearer be in this case? Or consider the likely circumstances and effects of "I'll see you the day after tomorrow", "I'll see you four days after the day before yesterday". This is not a matter of exotic examples; (hence synonymy) use or effects on hearers, in general. It is crucial to distinguish langue a person does or from parole, competence from performance.11 What is likely to do and what he knows may be related, in some way that cannot, for the moment, be made precise; the relation is, however, surely in part a factual and not a strictly conceptual can provide one. Performance use can provide evidence about meaning. as evidence about competence, Only
it is simply that the meaning of a linguistic expression cannot be characterized in terms of conditions of
in physics we may 65
NOAM CHOMSKY work theory, in studying language or interpreting what we hear), (or learning language, or translating, to make general statements this is not possible, since it is impermissible sense theories", about language or, more generally, about our "common on and since innate properties of the mind can impose no conditions language and theories?12 This is simply classical empiricist doctrine 'dogma' would, by now, be a more accurate term. It is difficult perhaps to see why this dogma should be taken more seriously than any other. It receives within the framework of a tentative
no support from what is known about language learning, or If it held true of humans, they from human or comparative psychology. would be unique in the animal world; and there is no evidence for this In general, it seems to me correct to say particular type of uniqueness. that insofar as empiricist doctrine has clear psychological it is content, in conflict with the not inconsiderable information that is now available. In any event, returning to the present theme, the particular that Quine makes about the mental processes and structures the basis for human status no assumptions that provide
and have language learning are quite unwarranted, the many assumptions that can be imagined. among special They can be justified only by empirical evidence and argument. Philos are no more persuasive based on these assumptions ophical conclusions on which the assumptions are without force. these conclusions present in a psychological context, Interpreted than the evidence of radical translation amounts rest; that is to say, for the
thesis of in then, Quine's to an implausible and quite determinacy unsubstantiated empirical claim about what the mind brings to the prob of language (or of knowledge lem of acquisition in general) as an innate This claim seems to me of only historical interest. Interpreted property. the possibility of develop a version of familiar skeptical ing linguistic theory, Quine's thesis is simply to the problem of arguments which can be applied as well to physics, to his "genuine hypotheses". veridical perception or, for that matter, context, a native speaker's that serious hypotheses concerning or concerning of English, the essential properties of human knowledge - the innate schematism that determines what counts as lin language on the basis structures are developed guistic data and what intellectual of these data - will "go beyond the evidence". If they did not, they would be without interest. Since they go beyond mere summary of data, it It is quite certain 66 in an epistemological as a claim about
be the case that But why
the data. concern?
consistent with there are competing assumptions or all of this occasion should any surprise
in Modern from 'Some Empirical of Language', Excerpted Philosophy Assumptions inHonor to appear in Essays P. Suppes, and of Ernest Nagel (ed. by S. Morgenbesser, of the editors. M. White), Printed here by permission forthcoming. 1 1960. W.V. Press, Cambridge, Mass., Quine, Word and Object, M.I.T. 2 of his remarks that is, the interpretation that is discussed above. Accepting, 3 states that "the learning of these wholes Elsewhere, largely Quine (sentences) proceeds of parts" and that "as the child progresses, he tends and assembling by an abstracting to build his new sentences of inter from parts" (p. 13). For consistency increasingly pretation, methods we must enumerated that suppose are intended refers to "analogical since the synthesis", to be exhaustive. If something else is intended, to vacuity, the innate basis for the "abstracting" until this three then and *
London, a mental
into Meaning in his Inquiry and Truth, Allen and Unwin, as expressing of real logical form and of logical words 1940, with his concept a structure to presuppose that would avoid at least these reality, does appear of Russell's and interesting But a discussion intricate very obvious quite problems. to these questions, is impossible the scope within approach though a useful undertaking, of this paper. 5 The reasons for this choice into a much more take us too far afield, would general later in the book, about the scheme of dis thesis, developed of reality" the true and ultimate structure in "limning (p. 221), traits of reality worthy of the name" (p. 228).
again reduces is specified. "assembling" 4 It is that Russell, interesting
consideration of Quine's course that one must use
and in describing "all 6 as a of "interlinguistic that Quine is using the concept Recall, again, synonymy" not only but also learning in the first for discussing of language device translation, a language. of what is said to him by one who knows place and interpretation 7 Cf. and Choice', 'Goodness Proceedings of the Aristotelian Society, Philippa Foot, I am sure, that we 35 (1961) 45-60. She comments, Volume Supplementary correctly as looking would like knives, but being something describe such objects else. exactly See 8 also the remarks by J. Katz on such words as 'anesthetic' in his 'Semantic Theory
and theMeaning of "Good"', Journal of Philosophy 61 (1964) 739-766.
Consider, showing transfer. human. for example, the experimental differences between apes and The difference that has been produced purportedly to carry out cross-modal in ability to the 'linguistic to the is sometimes attributed tags' available in Two and G. Ettlinger, Sense Mod Responding 'Opposite evidence humans in Brain Mechanisms and G. Ettlinger, 205-6; Grune F. L. Darley, and Stratton, New York that suggests itself is simply that the 'concepts' or disjunction in terms of conjunction defined is the general procedure in concept-formation
alities', (8 July, 1966) and Language, ed. by Underlying Speech and London, 1967.) Another possibility used in the experimental situation, being of elementary physical properties (as
(Cf. A. Moffet, 3732 Science, No.
are entirely artificial to the 'concept and mismatched space' of the tested human his own system of concepts however, subject, imposes (since he what the experiment understands is about, the conditions of the experi etc.). Under the distinction between the artificial of the experimenter and the natural ment, concepts of the subject might well be undetectable. itmight Hence be that no difference concepts experiments), animal. The between has transfer in cross-modal about apes and humans (and nothing linguistic tags) shown by such experiments, is shown and that what is merely that an yet been animal use of concepts cannot make to the reasonable that are mismatched (or human) of his system of concepts. innate structure 9 Of the cited conditions, the one that might as relevant be regarded is "knowledge Thus it makes sense to argue and belief". a change that under certain in conditions, of leads But language. to a difference it is senseless to hold that surely of disposition to verbal behavior,
belief may entail a modification wherever difference of belief
a difference there is necessarily of language involved. 10 In From a Logical Point of View, Harvard 1953. Press, Cambridge, University Mass., 11 The issue is vs. abstraction, not simply one of observation but rather one of signifi cant vs. pointless A set of dispositions to respond idealization. is a construction postu lated on the basis of evidence, to charac that attempts grammar just as is a generative terize 'knowledge of a language'. In Quine's and terms, the first is based on "genuine" on "analytical" but only in a sense of "genuine" that is divorced hypotheses, its ordinary meaning that seems to me (or else on the basis of a value judgment It would be more accurate to say that setting up a "complex of quite unsupportable). a pointless to respond" is merely has no in dispositions step, since such a structure the second from so far as is known. teresting properties, 12 as noted for the constraints Except, earlier, imposed by the structure forms of space, the system of truth-functional logic, certain primitive the capacity to form arbitrary associations. of the quality and
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue reading from where you left off, or restart the preview.