Conditions on Understanding Language1
Philosophers in general are uncomfortable, if not downright skeptical, about attributing semantic knowledge, particularly of a semantic theory, to ordinary speakers.2 Those who do not feel the pinch often adopt a two-pronged defense: they rebut skeptics with an array of distinctions (and hedges), contending that the skeptics' confusions arise because they ignore such distinctions,3 and, at the same time, argue that attributing such knowledge provides the best available working empirical hypothesis to account for linguistic comprehension.4 Though skeptical arguments abound about the relevance of semantics in explicating linguistic comprehension,5 a more acute challenge issues from Fodor and Schiffer: each offers an account of language understanding that excludes metalinguistic (semantic) knowledge, and therefore, knowledge of semantic theory.6
Both Fodor and Schiffer deny that there is any more to mastering a language than coming to have a capacity to go from what is heard to what is said:
A theory of understanding for a language L would explain how one could have an auditory perception of the utterance of a novel sentence of L and know what was said in the utterance of that sentence [Schiffer 1987, p.113, cf., also, p.262].
I assume that language perception is constituted by nondemonstrative inferences from representations of certain effects of the speaker's behavior (sounds that he produces, marks that he makes) to representations of certain of his intentional states, [in particular] a canonical representation of what the speaker said [Fodor 1984, pp.5-6].
Fodor and Schiffer deny that knowledge (or any other kind of epistemic/doxastic/psychological attitude) about the semantics of one's spoken language is causally relevant for effecting these transitions.
[W]hen we understand the utterance of a sentence we do not first come to the belief that it means such-and-such and then have that as our basis for thinking that the utterer was saying such-and-such [Schiffer 1987, my emphasis, p.262].
What really matters is this: For any perceptually analyzable linguistic token there is a canonical description (DT) such that for some mental state there is a canonical description (DM) such that ``DTs cause DM's'' is true and counterfactual supporting [Fodor 1989, p.8].
On this view, understanding consists in being a template for a causal network between the perceived linguistic sounds and shapes, and subsequent internal mental states. Understanding, adverting to jargon, consists in having a linguistic module in the mind. This module works (however it works) to set up a certain law-like covariation between instantiations of categories in the world and concepts in the mind.
The translation algorithm [from English into Mentalese] might well consist of operations that deliver Mentalese expressions under syntactic descriptions as output given English expressions under syntactic descriptions as input with no semantics coming in anywhere except, of course, that if it's a good translation, then semantic properties will be preserved [Fodor 1990, my emphasis, pp.187-8].
Schiffer concurs: no internally represented semantics is required for the use of a public language even if it has a semantics [Schiffer 1987, p.116]. This is the lesson of Schiffer's Harvey counter-example [Schiffer 1987, pp.192-207].
Harvey thinks in Mentalese...and his language processing uses not an internally represented [meaning theory] of English but rather an internally represented translation manual from English to Mentalese...Such a theory assigns no semantic values to the expressions of either language and in no sense determines a grammar (i.e., a meaning theory) for either language [Schiffer 1987, emphasis in the original, p.192f; p.262; cf., also, Schiffer 1994, p.304].
Since Schiffer contests the need for an internally represented compositional semantic, whether any epistemic relationship between a speaker and this semantics need exist cannot even arise. So, if the Fodor\Schiffer account, translationism, merely to label it,7 is correct, then the semanticist's is not, as Fodor likes to say, the only game in town.8
But how could translationism be right? Isn't it obvious speakers of Italian, for example, know semantic facts like (1)?
1. ``Sta nevicando'' means that it's snowing.
Translationists can agree that (1) is a truth Italian speakers know. But why, they wonder, must such knowledge be invoked in order to understand Italian? (1) is true only because speakers of Italian use ``Sta nevicando'' to communicate (``encode'' or ``express'') the thought that it's snowing [Schiffer, 1993, 1994, pp.303-04].9 So, if someone knows that (1), this must be because he knows his words can express this thought. From this it does not follow that such knowledge is (or need be) utilized in understanding ``Sta nevicando''. This is all Fodor means when he writes he is ``Gricean in spirit though certainly not in detail'' [Fodor 1975, pp.103-104; see, also, 1987, p.50; and Schiffer 1982, p.120; 1994, p.323]. Fodor -- and even Schiffer now -- are Gricean only inasmuch as both hold that whatever semantic properties natural language expressions have they inherent from (a language of) thought, Mentalese. So, contra Dummett [1978, p.97], translationists maintain that semantic knowledge about language is inconsequential; it plays no causal (and therefore, no rationalizing) role in linguistic competence.10
Still, aren't translationists postponing the inevitable? Suppose, as Harman believes, English speakers think ``in English''. Wouldn't it follow that understanding requires invoking knowledge about the semantics for English expressions? It would not. Even if one's lingua mentis is one's own public language, pace Fodor [1975, p.79ff] and Schiffer [1987, p.187], understanding still involves only the capacity to make transitions from what is heard to what is said, regardless of which language we think in. As Harman puts it, ``[w]ords are used to communicate thoughts that would ordinarily be thought in those or similar words'' [Harman 1975, p.271]. So, if we think ``in English'', then when someone who understands English hears an individual A utter ``It's snowing'', into her ``belief-box'' will go an English sentence to the effect ``A said that it's snowing''.11 No assumptions about the semantic properties of A's words need be invoked, much less knowledge about these properties?
But, one might wonder, how can merely tokening an English sentence in one's belief-box suffice for understanding? Mustn't we understand these internalized English sentences as well? Harman replies that we don't understand thoughts; we merely have or entertain them. Schiffer puts the point this way: ``...understanding a [language of] thought is simply a matter of thinking in it'' [Schiffer, 1994, p.322]. We don't say he's thinking it's raining but doesn't understand his thought. This could only mean he is unclear in his conception of rain or some such thing. So, nothing need be known in order to understand thought. Translationism, in short, maintains:
i. Understanding a natural language involves nothing more than having a (perceptual) capacity to hook up the natural language expressions with symbols of (the language of) thought.
ii. This skill requires no metalinguistic (semantic) knowledge?
My aim in this paper is to refute (i)-(ii).12 Before embarking, I'd like to say more than a few words about how the challenge of translationism illuminates anew what most philosophers argued was a misguided research program.
Structural Semantics - A Misguided Research Program?
In the 1960's and early 1970's, Fodor (along with Katz and Postal) practiced structural semantics (hereafter SS). SS theorists countenance properties and relations like synonymy, antonomy, meaningfulness, anomaly, logical entailment and equivalence, redundancy and ambiguity as a good initial conception of the range of semantics. Shunning details, SS theorists proceed by translating (or mapping) natural language expressions into (sequences or a set of) expressions of another language. There is no uniformity among them about the nature of this language or about how these translations or mappings are to be effected, but I beg no interesting questions by restricting attention to Katz and Fodor's SS proposal . The culmination of the various mapping rules and other apparatus within the Katz/Fodor framework results in theorems like (2):
2. ``Sta nevicando'' in Italian translates (or is mapped) into the language Semantic Markerese as S.
Mappings like (2) are constrained, and this is the raison d'etre for them as well as for semantic markers, such that synonymous expressions of a language L translate into the same (sequence or set of) expressions of Semantic Markerese, ambiguous expressions of L translate into different expressions of Semantic Markerese, anomalous expressions of L translate into no expression of Semantic Markerese at all and so on.
Once they surmised what SS was about, dissenting semanticists could not get into print fast enough to explain to Fodor and Katz just how confused they were about semantics. Davidson [1967, 1973], Vermazen , Lewis , Cresswell , Partee , among many others, each argued that since SS theories do not articulate relations between expressions and the world they cannot provide an account of the truth conditions for such sentences, and therefore, SS theories are not really semantic. Critics charged that the phenomena SS concerns itself with represent only a small portion of the full domain of semantics and, SS, they argued, cannot accommodate this full domain. As David Lewis put it:
...we can know the Markerese translation of an English sentence without knowing the first thing about the meaning of the English sentence; namely, the conditions under which it would be true. Semantics with no truth conditions is no semantics [1972:169-170].
Critics protested that even if an SS for Italian assigned an interpretation to every Italian sentence it would not specify what any expression of Italian means. On a Davidsonian conception, an adequate semantic theory for L must not only ascribe meanings to expressions of L, it must also ascribe them in a way that enables someone who knows the theory to understand these expressions. SS fails on this account because in the overall picture of SS there are three languages: the natural language, the language of Semantic Markers, and the translating (or mapping) language, which may be Semantic Markerese, the natural language, or some other language [Davidson 1973, p.129]. Since SS proceeds by correlating the first two languages using the third, one can understand its mappings, for example, (2), knowing only the translating (or mapping) language and not the other two. We can know that (2), perhaps on the basis of what Katz and Fodor tell us, without knowing what either ``Sta nevicando'' or its Semantic Markerese translation S means.
If someone understands Semantic Markerese, he no doubt can utilize (2) to understand the Italian sentence; but this is because he brings to bear two things he knows (2) does not state, namely, that Semantic Markerese is a language he understands and whatever information he has in virtue of which he understands S. It is this latter information an adequate semantics must characterize.
Shortly after the assault on SS by practically the entire philosophical community, Fodor ceased doing semantics. Most philosophers thought, certainly I did, he had acceded to his critics. However, in 1975, Fodor wrote he saw little difference, if any, between specifying meaning by translation and specifying it with truth-conditions:
We're all in Sweeney's boat; we've all gotta use words when we talk. Since words are not, as it were, self-illuminating like globes on a Christmas tree, there is no way in which a semantic theory can guarantee that a given individual will find its formulas intelligible...So the sense in which we can ``know the Markerese translation of an English sentence without knowing...the conditions under which it would be true'' is pretty uninteresting [Fodor 1975, pp.120-21].
In a critical response, I argued either Fodor misunderstood Lewis' objection against SS or the nature of truth-conditional semantics [Lepore & Loewer 1981]. I took Fodor to be misconstruing Lewis as saying that one must understand the language in which the canonical representation is expressed before one can utilize a semantic theory to determine what the represented sentences means and that's a problem every semanticist faces. This certainly is correct, but Lewis' point is not this obvious one. Instead, he is arguing that someone who understands a translation and knows it to be true need not understand the sentence of the translated language. We cannot understand (3) or (1) unless we understand English. But knowledge that (1), unlike knowledge that (3), requires no familiarity with English. Simply note that whereas (1) and (3) are grammatical, (4) is not:
3. ``Sta nevicando'' in Italian translates ``It is snowing'' in English.
*4. ``Sta nevicando'' in Italian means that it is snowing in English.
One need know no more English to know that (1) than Galileo knew for us truthfully to say that he said the earth moves.
With respect to the question whether language mastery requires semantic knowledge, none of this shows any more than that a semantics that specifies truth-conditions for sentences of a language L may serve to characterize (at least partially) knowledge sufficient for understanding L, while a semantics that specifies translation from L into another language by mapping structural descriptions of L into structural descriptions of the latter cannot (unless we add the assumption that the latter language is known). The conjectural inference from knowledge of truth conditions (partly) suffices to understand L to knowledge of truth conditions (partly) constitutes this understanding may seem natural, perhaps even good science, but Fodor, for one, balks. What's wrong with translationism, he asks?
Fodor's current skepticism about the utility of semantics for natural languages to account for linguistic comprehension is more than congenial with his early commitment to SS. He has repudiated much of the original SS program: its commitment to an analytic/synthetic distinction [Fodor and Lepore 1992]; its commitment to certain views about lexical decomposition [Fodor, Fodor, and Garrett 1975; Fodor, Garrett, Walker, and Parkes 1980]. But his early commitment to SS is of a piece with his denying the cogency of semantics for natural languages, at least qua theory of understanding.
Once Fodor gave up on the idea that understanding requires knowledge of a semantic theory, he began to see the semanticist's emphasis on natural language as sweeping all the interesting philosophical questions about content under the rug. For example, what bestows intentional (i.e., contentful) states on a cognitive system. Semantic theories, whether of the SS or truth-conditional variety, are useless here. Since natural languages, according to Fodor, merely shadow real intentionality, (philosophical) explication of semantic properties must focus on the mind, in fact, on the semantic properties of symbols of the mind. But explicating the semantic properties of thought, according to Fodor, is a metaphysical enterprise. Metaphysical questions merit metaphysical answers; not epistemic or psychological ones.
How does this make Fodor unrepentant? A truth-conditional account is no better than the original SS translation account not because both employ language, but rather because both leave unanswered interesting metaphysical questions about intentionality. Philosophers qua semanticists for natural language need not, nor should they be expected to, answer these questions. Fodor concurs. He concludes that semantics for natural languages, worse than boring, is worthless. This is his real challenge. Translationists challenge semanticists to supply a purpose for their endeavor. The rest of this paper assumes this challenge by advancing considerations that incline me to conclude that (a) and (b):
a. Understanding a natural language requires metalinguistic knowledge.
b. This metalinguistic knowledge must be semantic, and so cannot be merely translational.
Understanding and Rationalization
Maria utters to Massimo ``Sta nevicando''. Because translationism is compatible with Massimo believing Maria said it's snowing without his believing anything about the (causal) connection between what he heard and what he believes she said, it seems equally compatible with Massimo believing Maria said that it's snowing that Massimo foster only false beliefs about what Maria's words mean. But how, then, can the capacity to make correct transitions from what is heard to what is said alone suffice for linguistic competence? How can someone who has only false beliefs about what the expressions of a language L mean be linguistically competent with L? At least one author seems to think it's possible. Richard writes that someone ``might hold a false theory about competence, but still himself count as competent'' [1992, p.45]. If ``false theory'' means false beliefs about what words mean, then I disagree unless by ``count'' Richard means that the individual might never be exposed, something of no philosophical significance. So, is semantic skepticism compatible with translationism?
Translationists, because going Gricean is an option, can say it is not. By virtue of being linguistically competent in their sense and by virtue of knowing what thoughts words express or encode, translationism can frustrate semantic skepticism. But this concedes nothing to those who insist semantic knowledge is partly constitutive of linguistic competence, since semantic knowledge, on this account, is of no consequence. So, even though it were impossible to be linguistically competent without having true beliefs about what one's words mean, nothing follows about the causal efficacy of these beliefs in rendering the transitions translationists identify as constitutive of understanding. Even if intuition inclines one toward authority about what one's words mean (perhaps because we have authority about what our thoughts are, according to translationism), no one is inclined to endow speakers with authority about the causal ancestry of their beliefs, in particular, their beliefs about what another says when he speaks (or even the causal history of beliefs about what their words mean). This, unfortunately, is really bad news for philosophers/cognitive scientists who hope to frustrate translationism on empirical grounds [e.g., Segal 1994, pp.116-17]. Even if the best psychological account available of linguistic comprehension attributes rich semantic knowledge to competent speakers, this cannot establish that translationism is false, since it too can attribute such knowledge. To defeat translationism, we must establish that semantic information has repercussions for understanding. Neither the impossibility of semantic skepticism nor empirical science can establish anything so strong. I will take a different tack.
What I want to argue is that translationism is inconsistent with Massimo having reasons for his new belief? Like Dummett, I want to maintain that ``any adequate account of language must describe it as a rational activity'' [1978, p.104, my emphasis]. Though Dummett's target is ``a causal theory such as Quine appears to envisage, representing [language mastery] as a complex of conditioned responses'' [1978, p.104], I want to cast my net wider to encompass translationism, a position, unlike Quine's, thoroughly cognitive. To this end I must show that the rationalizations speakers qua speakers have cannot be underwritten by translationism.
Why should the sort of rationalizing linguistic comprehension carries require ascribing metalinguistic (semantic) knowledge to speakers? Why isn't it secured already by the reliable connections between heard utterances and beliefs about what is said translationism presumes? The capacity for language comprehension produces correct internal states on the basis of what is heard; so, why aren't such states justified on this basis alone?
Someone's belief being justified and his having another belief which rationalizes his belief are distinct. Many perceptual beliefs are justified directly by experiences on which they are based, and in principle a belief can be justified simply by being the result of a reliable belief forming mechanism. One's belief that one is currently in pain is clearly not justified on the basis of other beliefs one has. The only explanatory story we are in a position to give is one which invokes a mechanism that connects reliably one's being in pain with one's believing one is. Translationists see linguistic comprehension in the same light [Fodor 1984].13
What's on offer is reliabilism [Dretske 1981; Goldman 1986; Nozick 1981]. Beliefs about what's said count as justified just in case processes that produce them tend, in the ``relevant'' set of counterfactuals, to be truth inducing. There is indeed a lawlike correlation between an Italian speaker's beliefs about what is said and the heard utterances that bring them about.
Also, no ``KK principle'' is invoked; being justified that p does not entail being justified that you are justified that p. This reliabilist feature serves translationism well. If someone's belief fixation processes may be reliable and constitute justification even though he does not realize they do, then translationists can deny speakers require special metalinguistic knowledge about the connection between what's uttered and what's said in order to secure whatever justification linguistic comprehension requires.
Much has been written, pro and con, about reliabilism, but in this context it's viability is not relevant. We want to know Massimo's reason for his belief that Maria said it's snowing when he heard her utter ``Sta nevicando''. That he has a certain faculty that, cateris paribus, delivers him from heard Italian utterances to true beliefs about what is said fails to reveal his reason. If Massimo knew he was so constituted, by virtue of learning Italian, that he reliably acquires true beliefs about what's said when he hears Italian utterances, then Massimo would have a reason for his belief. But drawing on such knowledge here is illegitimate. It undercuts reliabilism's appeal by resurrecting the KK principle.14
So, does Massimo have a reason for his belief even if he lacks beliefs about Italian? Imagine, as is consistent with translationism, that Massimo is clueless about why he believes (correctly, let's suppose) that p when he hears Maria utter something Italian. Nothing in his head justifies his belief. Massimo's condition is mildly pathological. Poor dupe, running around the world telling all he meets what others said but always lacking reasons for such attributions. Massimo can no better explain his belief that Maria said it's snowing than to say ``I don't know why I believe this. I just do. Didn't Maria say `Sta nevicando'?'' But someone who understood not one word of Italian and happened to find himself believing Maria said that it's raining could make the same case for himself. Massimo is not unlike someone who perpetuates ghastly deeds, but literally has (or should I say, can have) no idea why he persists. No degree of prodding or assistance could bring him to reconstruct reasons for his behavior. Just as we would withhold agency from him, we should withhold linguistic comprehension from clueless Massimo.
Diagnoses: Even if reliabilism secures some sort of justification, it does not secure one sufficient to underwrite linguistic comprehension. If Massimo believes Maria said it's snowing when he hears her utter ``Sta nevicando'', we expect him to have beliefs about Maria's utterance that play a rationalizing role. If such rationalization is integral to language understanding, where could it spring from if not from knowledge (or belief or other propositional attitudes) about the sounds and shapes of the language itself.15
Massimo, in our imagined scenario, really is clueless. He has no idea why he believes Maria said it's snowing when he heard her utter ``Sta nevicando''. So, even though he makes the transition, he has no reason¾conscious, unconscious, tacit, explicit, implicit, or any other sort. I am claiming that if one understands a language, he must have reasons that rationalize his transitions. To echo Davidson, ``nothing can count as a reason for holding a belief except another belief'' [1986, my emphasis, p.123]. I know no argument that defends this position tout court, but it seems right in the case of language comprehension. What about linguistic comprehension provides reason for the belief that it's snowing when this understanding combines with the belief that Maria uttered ``Sta nevicando''? Additional beliefs or knowledge Massimo has about Maria's utterance that non-speakers lack. In short, if translationism is right, it behooves us to ask about another's reason for what he believes on the basis of linguistic shapes and sounds (he believes) he perceives. But then nothing less than appeal to other mental states about what he perceives can rationalize the belief.
Suppose I'm right both about the failure of reliabilism to deliver required rationalizations for beliefs about what's said and suppose even further that such rationalizations require intervening attitudes about words heard. Question: why can't the translationist rebut, ``OK, but why isn't this just the denial of semantical skepticism, something I've already conceded?'' Answer: what I'm assuming is that rationalizations for beliefs parallel rationalizations for actions, that is, both require causally efficacious intervening attitudes. So, a belief that p (partly) rationalizes a belief that q only if the belief that p is (partly) causally responsible for the belief that q. Needless to say, this isn't uncontrovertial.16
There is an important feature of my discussion I have not flagged. Not all that long ago, Chomsky spent too much time defending the psychological/epistemic status of grammatical theories he postulated to account for linguistic comprehension. My argument against translationism circumvents these hairy issues. It's insignificant for the purposes of this debate what the nature of the relationship is between a speaker and whatever metalinguistic information is essential for understanding. If I'm right, then there must be a relationship. Whether it amounts to tacit or implicit or explicit knowledge, or whether the relationship is not knowing but ``cognizing'', or whether it is a completely different doxastic relationship is irrelevant. What's essential is, if I'm right, there must be some such epistemic/psychological/doxastic relationship toward semantic information that stands between the heard utterance and the acquired belief about what is said if the latter is not be an unrationalized psychological state. We can leave open its nature. 17
Why Semantic Knowledge?
Suppose translationism is wrong. It doesn't follow that knowledge about meaning must be invoked to account for transitions from heard utterances to beliefs about what's said. Take Schiffer's story. Why doesn't it suffice to say that what underwrites Harvey's understanding his language is his knowing a translation manual from his public language into his language of thought?
A translation manual from L to L' is a finitely axiomatizable theory that correlates words and structures of L with words and structures of L' so as to entail theorems that correlate L sentences with their synonyms in L'. Such a theory assigns no semantic values to the expressions of either language and in no sense determines [a meaning theory] for either language...Harvey works in the following way. His internally represented translation manual determines a function that maps each English sentence onto its Mentalese synonym, and he is so ``programmed'' that when he has an auditory perception of an utterance of a, then straight-away there enters his belief box the Mentalese translation of ``The speaker in uttering `a' said that a''... [Schiffer, 1993, p.244].
It's easy to get snowed by Schiffer's technical jargon and by what appear to be merely heuristic devices, for example, appeals to Mentalese, translation manuals, the belief-box, talk about translating public language sentences into Mentalese. We want to know Massimo's reason for his belief that Maria said it's snowing when he believes she utters ``Sta nevicando''. According to Schiffer, Massimo understands Maria's utterance if he is caused to believe Maria said that it's snowing (i.e., token in his belief-box a sentence of Mentalese which expresses what ``Maria said that it's snowing'' does in English) when he hears her utter ``Sta nevicando''. If an epistemic relationship toward some internal psychological state is required for linguistic comprehension, why can't it be knowledge of the correct mapping from the Mentalese counterpart of ``Maria uttered `Sta nevicando''' to a Mentalese counterpart of ``Maria said that it's snowing''?
If this is possible, using L does not require knowing anything about an internally represented meaning theory for L. So, the inference that knowledge of meaning is required for language comprehension is still not sanctioned. However, elevating Schiffer's suggestion to an object of knowledge doesn't work; seeing that it doesn't shows why an adequate account of linguistic comprehension requires reference to connections between language mentioned and language used, i.e., semantic information.
Schiffer's choices are between a function that maps structural descriptions of Italian expressions into structural descriptions of Mentalese expressions, in which case it is a translation manual (and Davidson's translation argument kicks in); or a function that maps structural descriptions of Italian expressions, that is, language mentioned, e.g., Maria uttered ``Sta nevicando,'' into language used, that is, that Maria said that it's snowing. Everything turns on how we understand the locution a certain sentence is tokened in Massimo's belief-box.18
Massimo hears Maria utter a certain sentence; he knows what she uttered translates into a certain Mentalese sentence; whatever corresponds in his lingua mentis to ``Maria said that'' concatenates with this translation and the entire product goes in his belief-box. That suffices for him to understand Maria's language. What I'm doubting is that we can specify this knowledge in a way that both avoids the standard translation argument and does not itself draw upon semantic information, i.e., meaning, truth, or satisfaction conditions.
The standard Fodorian reply that, for Mentalese, questions about understanding cannot arise won't work here. It may be illegitimate to ask in virtue of knowing what does someone understand one's mentalese? [Cf., Lycan 1984, p.237f.] But that is not our question. We're asking, in virtue of what knowledge does one understand his public language? The suggestion that it is in virtue of knowing a translation manual from, say, Italian, into Mentalese won't work if the mapping is from structural descriptions into structural descriptions. But suppose it maps structural descriptions into, say, well, what? It cannot be propositions or states of affairs. That would be entirely useless to Massimo; what about a sentence used in the language of thought or a sentence-in-use in the language of thought. I want to argue that a ``translation manual'' in this sense determines a meaning theory, that is, a semantic theory.
Suppose a function F in effect maps a set S of structural descriptions of sentences of L into a set P of sentences (not structural descriptions) of L' such that F(s) (in S) = p (in P) iff (if X assertively utters s in L, then X says that p). Then F determines a semantic theory for L.19
T is a truth theory for a language L in a metalanguage L' iff for every structural description s of a sentence of L, T implies a true sentence of L' such that:
s is true in L iff p
(where p is replaced by a sentence of L', and each s translates whatever replaces p). F determines an adequate truth theory T for L in L' iff for any structural descriptions s1 and s2 of sentences of L, F(s1) = F(s2) only if T implies (for some sentence p of L') (a) and (b):
a. s1 is true in L iff p
b. s2 is true in L iff p
But F(s1) = F(s2) only if there is some p in L' such that both s1 and s2 ``translate'' p. But this establishes that some truth theory for L implies (a) and (b).
If F were merely a translation manual in a traditional sense, this result could not follow. That's what Davidson's translation argument establishes. That ``Sta necicando'' translates ``It's snowing'' cannot determine that ``Sta nevicando'' is true iff it's snowing. The disquotation principle behind this inference is not innocent. It assumes that ``It's snowing'' is true iff it's snowing. In the current context that assumption is question begging since it's exactly what I'm trying to defend.
None of this establishes, of course, that knowledge of F provides the rationalization I claim (partly) constitutes linguistic comprehension, merely that it determines something that could provide this warrant. That I know p, and p determines something q such that if I knew that q, my belief that r would be rationalized does not imply that knowing p alone rationalizes my belief that r. Moreover, even if knowledge of a semantic theory for L suffices for understanding L, it's still open whether knowledge of non-semantic mechanisms might suffice for understanding L as well. So, I have not established that semantic knowledge is necessary. And, so I have not established that semantic theoretical knowledge is necessary since it is consistent with what I argued that one need not know any more than the meaning theorems that issue from an adequate semantic theory. Massimo belief that Maria said it's snowing is justified if he merely believes ``Sta nevicando'' means that it's snowing and Maria assertively uttered ``Sta nevicando''. That he need also know that an object a satisfies ``neve'' iff it's snow in order for the transition to be justified requires further argument. These are rather significant loose ends; I believe they can be tied up. I'll leave that for another occasion.
So, where are we? We began with what I still think is the greatest challenge to those of us who find semantics for natural language not only interesting but valuable, viz., how to refute translationism? To this end I argued that translationism leaves speaker's beliefs about what others say unrationalized. Going externalist about rationalization, as reliabilism recommends, seems misguided, at least to me. Internalist accounts both invoke attitudes about the words we hear and treats those attitudes as causally responsible in effecting beliefs about what's said. Traditional translation manuals mapping structural descriptions into structural descriptions are no help here; and any ``translation'' manual that takes words mentioned into words used (or words-in-use) sneaks in just the semantic information we are trying to redeem. There are obviously many missing steps and many of the steps taken lack anything like an ironclad defense, but I hope, at least, the dialectic is sufficiently precise. Let me end on a different note.
The most common reason for resisting the idea that speakers have semantic (theoretical) knowledge is that such knowledge is not ``within the ken of plain folk'' [Schiffer 1987, pp.255-261]; not something of which we have ``conscious access'' [Foster 1975, p.2]; not something we can ``literally credit'' to speakers [Dummett 1974, p.110]. Nothing I'm recommending requires Massimo to have explicit representations or be able to consciously reconstruct pieces of practical reasoning from perceived sounds to extra-linguistic belief [Lepore 1982; Higginbotham 1983; 1987; George 1989]. Massimo may be unequipped, incapable, or unskilled. But if his beliefs about what's said are rationalized , it makes sense for us to articulate his reasons. My argument, if any good, establishes that some relationship toward metalinguistic states or information about one's language is required for linguistic comprehension. I don't have a clue what the psychological make-up of this relationship must be like; but no one should take a critical stance on these issues without, at least, having a fairly developed account of concepts against which to evaluate such attributions.Bibliography
Baker, G.P. and P.M.S. Hacker, Language, Sense and Nonsense, Basil Blackwell, Oxford, 1984.
Bruner, J., ``On Perceptual Readiness,'' Psychological Review, 65, 1957:14-21.
Chomsky, Noam, Knowledge of Language, Praeger, 1986.
Cresswell, M.J., ``Semantic Competence,'' in M. Guenthner Ruetter and F. Guenthner (eds.), Meaning and Translation, Duckworth, London, 1978:9-27.
Davidson, D., ``Truth and Meaning,'' Synthese 17, 1967. Reprinted in Davidson 1984.
Davidson, D., ``Radical Interpretation,'' Dialectica, 27, 1973. Reprinted in Davidson 1984.
Davidson, Inquiries into Truth & Interpretation, Oxford University Press, Oxford, 1984.
Davidson, D., ``A Coherence Theory of Truth and Knowledge,'' in Truth and Interpretation, 1986:141-154.
Davies, Martin, Meaning, Quantification, Necessity, Routledge and Kegan Paul, London, 1981.
Devitt, M., Designation, Columbia University Press, New York, 1981.
Dretske, F., Knowledge and the Flow of Information, MIT Press, Cambridge, MA, 1981.
Dummett, Michael, ``What is a Theory of Meaning (I)?, Mind and Language, ed., S. Guttenplan, Clarendon Press, Oxford, 1974:97-138.
Dummett, Michael, ``What do I Know When I Know a Language?'' (1978), reprinted in The Seas of Language.
Dummett, Michael, The Seas of Language, Clarendon Press, Oxford, 1993.
Evans, G., ``Semantic Theory and Tacit Knowledge,'' in S.H. Holtzman and C.M. Leich, eds., Wittgenstein: To Follow a Rule, Routledge and Kegan Paul, London, 1981.
Evans, G. and J. McDowell, eds., Truth and Meaning, Duckworth, Oxford, 1975.
Fodor, J, The Language of Thought, Crowell, New York, 1975.
Fodor, J.A., The Modularity of Mind, MIT Press, Cambridge, 1984.
Fodor, J.A., Psychosemantics, MIT Press, Cambridge, 1987.
Fodor, J.A., ``Why Must the Mind be Modular,'' in Reflections on Chomsky, ed. A. George, Basil Blackwell, Oxford, 1989.
Fodor, J.A., ``Review of Stephen Schiffers's Remnants of Meaning,'' in A Theory of Concent and Other Essays, MIT Press, Cambridge, 1990.
Fodor, J.A., ``Reply to Schiffer,'' in Meaning and Mind, eds. B. Loewer and G. Rey, Basil Blackwell, Oxford, 1991.
Fodor, J.A., Fodor, J.D., and Garrett, M. ``The Psychological Unreality of Semantic Representations,'' Linguistic Inquiry, 6, 1975:515-531.
Fodor, J.A., M. Garrett, Walker, and Parkes, ``Against Definitions,'' Cognition, 8, 1980:263-367.
Fodor, J.A. and E. Lepore, Holism: A Shopper's Guide, Basil Blackwell, Oxford, 1992.
Foster, J., ``Meaning and Truth Theory,'' in Evans and McDowell, 1975.
George, A., 1989, ``How not to become confused about linguistics,''
Goldman, A., Epistemology and Cognition, Harvard University Press, Cambridge, 1986.
Hacking, Ian, Why does Language Matter to Philosophy?, 1975.
Harman, G., ``Language, Thought, and Communication,'' in Minnesota Studies in the Philosophy of Science, vol. 7, ed. K. Gunderson, University of Minnesota Press, Minneapolis, 1975.
Higginbotham, J.,''Is Grammar Psychological?'' in L. Cauman, et al (eds), How Many Questions? Essays in Honor of Sidney Morgenbesser, Hackett Publishing, Cambridge, MA, 1983.
Higginbotham, J., ``The Autonomy of Syntax and Semantics,'' in J. Garfield, ed., Modularity in Knowledge Representation and Natural Language Understanding, MIT Press, Cambridge MA, 1987.
Higginbotham, J., ``Is Semantics Necessary?,'' Proceedings of the Aristotelian Society, 87, 1988:219-241.
Higginbotham, J., ``The Place of Natural Language,'' 1995, in On Quine, Leonardi, P and Santambrogio, M., eds., Cambridge University Press, Cambridge, 1995.
Hornstein, N., Logic as Grammar, MIT Press, Cambridge, 1984
Katz, J. & J. Fodor, ``The Structure of a Semantic Theory,'' Language, 39, 1963:170-210.
Larson, R. and P. Ludlow, ``Interpreted Logical Forms'', Synthese 95, 1993:305-356.
Larson, R., and G. Segal, Knowledge of Meaning, MIT Press, Cambridge, MA, 1995.
Lepore, E, ``What Model Theoretic Semantics Cannot Do,'' Synthese, 54, 1983:167-187
Lepore, E. ``In Defense of Davidson,'' Linguistics and Philosophy 5, 1982:277-294.
Lepore, E., ed., Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, Basil Blackwell Press, Oxford, 1986.
Lepore, E., and B. Loewer, ``Translational semantics,'' Synthese, 48, 1981:121-133.
Lewis, David, ``General Semantics,'' Synthese, 22, 1970:18-67.
Loewer, B. and G. Rey, eds., Meaning and Mind: Fodor and his Critics, Basil Blackwell, Oxford, 1992.
Lycan, W.G., Logical Form in Natural Language, MIT Press, Cambridge, 1984.
McDowell, J., ``In Defense of Modesty,'' in Barry Taylor, ed., Michael Dummett: Contributions to Philosophy, , Martinus Nihjof, 1988.
Nozick, R., Philosophical Explanations, Harvard University Press, Cambridge, 1981.
Partee, B., ``Montague Grammar and Transformational Grammar,'' Linguistic Inquiry, 6, 1975:203-300.
Partee, B., ``Semantics: Mathematics or Psychology?'' in Baurle, R., and E. Egli, eds., Semantics from Different Points of View, 1979.
Quine, W.V.O., ``Methodological Reflections on Linguistic Theory,'' Synthese, 1970.
Richard, Mark, ``Semantic Competence and Disquotational Knowledge,'' Philosophical Studies, 65, 1992:37-52.
Schiffer, Stephen, ``Intention-Based Semantics,'' Notre Dame Journal of Formal Logic, vol 23, no. 2, 1982:119-156.
Schiffer, Stephen, Remnants of Meaning, MIT Press, Cambridge, 1987.
Schiffer, Stephen, ``Actual-Language Relations,'' Philosophical Perspectives 7, 1993:231-258.
Schiffer, Stephen, ``A Paradox of Meaning,'' Nous 28, 3, 1994:279-324.
Segal, G., ``Priorities in the Philosophy of Thought,'' Supplementary Proceedings of Aristotelian Society, 1994:107-30.
Smith, Barry C., ``Understanding Language,'' Proceedings of Aristotelian Society CII, 1992:109-141.
Soames, S., ``Semantics and Semantic Competence,'' Philosophical Persepctives, 3, Philosophy of Mind and Action Theory, ed. J. Tomberlin, Atrascadero, Ridgeview Publishing Company, 1989:185-207.
Stich, S., From Folk Psychology to Cognitive Science, MIT Press, Cambridge, 1983.
Thomason, R. (ed.), Formal Philosophy: Selected Papers of Richard Montague, Yale University Press, New Haven, CT, 1974.
Vermazen, Bruce, ``Review of Jerrold J. Katz and Paul M. Postal, An Integrated Theory of Linguistic Description, and Jerrold J. Katz, The Philosophy of Language, Snthese,'' 17, 3, 1967:350-65.
Wright, Crispin, ``Theories of Meaning and Speaker's Knowledge,'' in Philosophy in Britain Today, ed., S.G. Shanker, State University of NY Press, Albany, 1986:267-307.