No CrossRef data available.
Article contents
Inscrutability
Published online by Cambridge University Press: 01 January 2020
Extract
Pick what seems a correct scheme of reference for a language, German say. Perhaps you have picked one which tells us such things as that
‘Katz’ refers in German to cats (and nought else)
‘Hund’ refers in German to dogs (and nought else).
Your scheme, given that it also accounts for German devices of sentence compounding such as ‘und,’ ‘nicht,’ and ‘alle,’ assigns truth conditions to German sentences.
Quine reminds us that it is possible to permute your reference scheme in such a way that the result assigns the same truth values to all German sentences, even though the permutation changes the reference of all of German's terms. For example, we can take ‘Katz’ to refer to undetached cat parts, ‘Hund’ to undetached dog parts, ‘lauft’ to be true of undetached parts of things that run, ‘ist identische mit’ to name the relation undetached object parts stand in when they are parts of the same object, and so on. Quine observes that it seems possible to do this without effecting the ‘observational consequences’ of German theories: a theory in German will be seen, under one interpretation, to be confirmed by some evidence iff it is so confirmed under the other.
- Type
- Research Article
- Information
- Canadian Journal of Philosophy Supplementary Volume , Volume 23: Meaning and Reference , 1997 , pp. 165 - 209
- Copyright
- Copyright © The Authors 1997
References
1 For example: the theory we began with has ‘ein Katz lauft’ true iff a cat runs. The replacement theory has it true iff an undetached cat part is an undetached part of something that runs. The one condition obtains iff the other does.
2 I blithely speak of predicates referring to their extensions throughout.
3 Quine, W.V.Ontological Relativity and Other Essays (New York: Columbia University Press 1969), 50CrossRefGoogle Scholar
4 Quine, Ontological Relativity, 48Google Scholar
5 I will confuse use and mention as I do here whenever the confusion is easily resolved by those who are attuned to it in the first place.
6 Quine, Ontological Relativity, 82-3Google Scholar
7 Perhaps it will be said that there is an incoherence here: How can (say) it be true that theory T is committed to UFOs, and not some other entities (OFUs), while it is also true (from some other perspective about reference) that T is committed to OFUs, not UFOs? This would be incoherent if the notion of truth invoked were not language relative. But for Quine, ascriptions of truth are language relative.
It would be incoherent to hold that T is committed to UFOs, but there is a language we might speak in which one could truly say that T was not so committed. So we must say, on a view like that which I am ascribing to Quine, that what (some of) the sentences of this other theory say is not what they appear to say, when translated in the natural fashion (i.e., homophonically). We will say that, in the sense of meaning in which meaning determines reference, ‘refers’ in those languages does not mean what it means in our language- it cannot, since it has a different extension. (It does not follow that there is no sense of ‘meaning’ in which ‘refers’ in the different languages may not have the same meaning. For instance, if we allow that there is a sense of ‘meaning’ on which expressions with similar ‘inferential roles’ have similar meanings, we may allow that ‘refers’ has similar meanings in languages in which it has disparate extensions.)
8 Field, Hartry ‘Quine and the Correspondence Theory,’ Philosophical Review 83 (1974)CrossRefGoogle Scholar
9 There are a number of (in my opinion unsound) objections to such relativization due to Davidson, Donald (in ‘The Inscrutability of Reference,’ Southwestern Journal of Philosophy 10 [1979]CrossRefGoogle Scholar) which limitations of space forbid discussing here.
10 Davidson, ‘The Inscrutability of Reference,’ 231Google Scholar
11 This works only if no expression occurs in multiple languages. There are ways of individuating expressions — as constructions from sets of tokens, for example — which will achieve this.
12 Sometimes, the relativization is said to be a background theory. I don't think it makes much difference, as long at the second parameter itself provides a translation manual.
13 Once the five-termed REF(α, L, y, M1, M2) is in place, we can use it to understand ordinary talk of reference: if my use of ‘refers’ presupposes M, then my use of ‘α refers in L to y’ is true as I speak it provided REF(α, L, y, M, M).
Since the translation manual involved in the REF relation is simply a mapping of vocabulary to vocabulary, this account makes claims about reference trivial. If we individuate languages in terms of their expressions (so that it is necessary that ‘Katz’ is an expression of German), then “'Katz’ in German refers to the set of cats” says something necessary. What is not (wholly) trivial is the choice of how to extend an innocent language to one in which talk of reference is possible: as I shall argue below, different accounts of reference may have different empirical implications which will allow us to choose among them.
We could understand the claim that REF(α, L, y, M1, M2) in another way, however: We could interpret it as entailing the claim that Ml and M2 are empirically adequate manuals for translating from the relevant languages. I do not think anything in what follows depends on which understanding we adopt, and so I will stay neutral on whether this understanding is preferable.
14 As noted above, the strong thesis is not a consequence of the weak one. Indeed, the weak thesis is consistent with the claim that there is only one adequate reference scheme for German in English – i.e., only one mapping of simple German terms onto English terms which, from an idealized perspective, preserves reference. For all that the weak thesis demands is that there be different stories to tell, from the idealized perspective, about what German and English speakers are referring to. This might be, even if there is, in English, only one adequate story which can be told about German reference.
15 Quine, W.V.Pursuit of Truth (Cambridge, MA: Harvard University Press 1990), 31-2Google Scholar
16 Surely there can be no serious objection to the idea that we can observe such interactions. If a duck quacks and that causes a goose to honk, there is a straightforward sense in which we observe that duck's quacking brought about the goose's honking. What is good for the goose, I would insist, is also good for interactions between speakers and the duck's quacking.
17 To what part of the permutation argument have I objected? It depends on what counts as a semantic property. If sentential properties like being typically accepted because one heard a duck quack are semantic, then the answer is that I'm objecting to premiss (b), which is false if what I said about principles P1 and P2 is correct. If the honorific ‘semantic’ is restricted to properties and relations such as being true in L, and referring to x in L, then my objection is that the argument isn't valid, since there are relations beyond these to which we can appeal to criticize a semantic theory.
18 One finds in Davidson and Quine epistemological arguments which might be thought to be helpful in establishing (a). For example, if one accepts that semantic facts are determined by something ‘accessible’ to the child learning a language, one might be able to argue to (a). I shall ignore these; they have, in any case, been criticized elsewhwere. (See, for instance, Friedman's, Michael ‘Physicalism and the Indeterminacy of Translation,’ Noûs 9 [1975]CrossRefGoogle Scholar.)
19 After writing an early version of this section in 1991, I discovered that Unger, Peter (in Philosophical Relativity [Minneapolis, MN: University of Minnesota Press 1989], Chapter 1)Google Scholar has used the fiction of a very twin earth for purposes somewhat similar to those I use it.
20 It should be noted that it is consistent with this that once I's manual is in place, there will be facts, about the relations between I and English and vte English speakers, which are relevant to translating from their languages into other ones. This will be relevant in the next section.
The last two paragraphs were prompted by objections from Vann McGee and Alex Byrne.
21 Quine, Ontological Relativity, 36-7Google Scholar
22 Here is what Fodor says about temporal parts reference schemes:
Notice that, in the normal course, a rabbit and its ears are contemporaries, so that a time slice that includes the one generally also includes the other. But ‘rabbit’ and ‘rabbit's ear,’ unlike ‘time slice of a rabbit’ and ‘time slice of a rabbit's ear’ are mutually exclusive. So the deviant ontology fails. [Fodor, JerryThe Elm and the Expert (Cambridge, MA: MIT Press 1994), 123, n.3]Google Scholar
It seems that Fodor is reasoning along the following lines:
A rabbit-ear-stage is (normally) part of a rabbit-stage. So, (normally) there are things which are both rabbit-ear-stages and rabbit- stages.
But this argument is no better than one which identifies cars and carburetors, because the latter are parts of the former.
Some speakers- rabbit/rabbit body dualists of a sort- will blanch at the idea that
A3: a is a rabbit and a is a rabbit body
might be true. For such speakers, ambiguation of ‘is a rabbit,’ so that it sometimes refers to MIT rabbits, and sometimes to maximal sums of gen-identical MIT rabbits, is necessary. Since such speakers seem to be placing two objects in one place at one time, the resort to ambiguation of some sort for them is necessary anyway.
23 Appropriate adjustments in interpretation- for example, for the tenses- are understood.
24 Evans alleges a problem for an intepretation of ‘rabbit’ as a predicate of rabbit stages: It cannot deal with the tenses. Evans considers two Tarskian accounts, in which tense is treated as quantification over time. I will only discuss the second. Here, ‘tu’ names the time of utterance; At(x,tu), I take it, is something like a existence predicate, true of a stage and a time iff the stage is contemporaneous with the time. Then, taking some inessential liberties with Evans’ presentation, the account runs so:
T. 1. x satisfies is F iff At(x, tu) and x satisfies F
2. x satisfies was F iff for some time t'before tu and some y co-membered with x, At(y, t’) and y satisfies F
3. x satisfies ‘warm’ iff x is a stage of a thing warm while the thing is warm.
Presumably these are truncations of claims to the effect that relative to an utterance, or to an utterance time, x satisfies a wff iff such and such obtains. Otherwise, T seems incoherent, since (for example), T.1 defines a temporally unrelativized notion of satisfaction as a temporally varying relation. (Evans's idea, I take it, is that a stage satisfies a tensed predicate only at the times it exists, though it ‘satisfies’ an untensed predicate such as ‘warm’ or ‘rabbit,’ ‘absolutely’ or at every time.) I will henceforth so understand the proposal, and take it in a slightly generalized form, as defining satisfaction, of a wff by an object, relative to an arbitrary time. So understood, and writing u[Ф]t for relative to t, u satisfies Ф, the proposal runs
T. (1) u[is F]t iff (At(u,t) and u[F]t)
(2) u[was F]t iff ∃t'∃u‘(t'<t and u'is co-membered with u and At(u', t’) and u‘[F]t’)
(3) u[warm]t iff u is a warm stage.
What is wrong with T? According to Evans,
We are supposing that an object satisfies the tensed predicate ‘was warm’ iff it is a stage latter in the life of some object than some stage which satisfies the simple predicate. But this does not get the truth conditions right. ‘A rabbit was running’ may be true even though there is no stage of a rabbit latter than some running stage - the running stage might have been the last. (Evans, Gareth ‘Identity and Predication,’ Journal of Philosophy 72 [1975], 361)CrossRefGoogle Scholar
But there is an obvious response to this. Note first that the form
Quantifier phrase+tense+predicate
apparently suffers from a scope ambiguity, between subject and tense, as witnessed by sentences such as ‘The president was a Republican.’ In standard tense logic, accounting for this is straightforward, since predicates are treated as open sentences and tenses as sentence operators; the two readings may then regimented somewhat so
P1. was( the presidentx (x is Republican))
P2. the presidentx ( was(x is Republican))
If there is an ambiguity here, there is one in ‘a rabbit was warm’ as well, between readings which, on a first pass, might be regimented so
R1: was(a rabbitx (x is warm))
R2: a rabbitx (was (x is warm))
A subtlety arises, since in T the tenses are (apparently) being treated as expressions which modify adjectives (and perhaps noun phrases) to yield a verb phrase. In the context of a proposal like T, either the tenses need to be syntactically ambiguated (so that there is a ‘was’ which applies to adjectives and a ‘was’ which applies to sentences), or we need to reformulate T.l and T.2 so that ‘is’ and ‘was’ are understood as applying to something sentential (with the status of ‘x is F’), or we need to reformulate the syntax in some more dramatic way.
However, the issues here are purely syntactic, not semantic. If we think of the tenses in proposal T as applying to sentences- and thus think of the ‘F’ in T1 as ranging over something with the status of open sentences - we can regiment the ambiguity in ‘a rabbit is warm’ as above. If we then extend T in the obvious way, by adding
(4) u[a Nx(F)]t iff ?u‘(At(u',t) and u‘[N]t and u‘[F]t]
(5) u[rabbit]t iff u is an MIT rabbit
we assign readings R1 and R2 these truth conditions: Relative to t,
R1: For some t’ before t, there is a u which exists then which is an MIT rabbit and is a warm stage.
R2: For some u which exists at t and is an MIT rabbit, there is an earlier coslice which is a warm stage.
R1 of course is the reading which Evans says T cannot capture.
I believe that the treatment given here, and its obvious relation to a tense logical treatement refutes Evans's odd claim (‘Identity and Predication,’ n. 15) that “treating tenses as operators requires an ontology of persisting things.”
Considerations of space prevent me from discussing other claims which Evans and Fodor make about ‘deviant’ reference schemes. I must, however, remark that Evans's objections to re-interpretations which look upon predicates as designating properties seem to me completely without merit, ignoring as they do the possibility of treating a language in a way parallel to the way we treat the first order predicate calculus when we interpret it by mapping its sentences onto the sentences of an (interpreted) version of predicate functor logic.
25 The exact criteria vary between Davidson and Quine, and vary for each across writings; see, for instance, Quine, Word and Object (Cambridge, MA: MIT Press 1960)Google Scholar, Pursuit of Truth; Davidson, ‘Radical Interpretation,’ Dialectia 23(1973)Google Scholar. So far as I can see, such variation is not relevant to the present argument.
26 Two responses which I should perhaps mention, are these: (a) One could shrug and say that it is in fact indeterminate what one says- this is a consequence of inscrutability - and so it must, strictly speaking, be indeterminate whether I understand you, given that I picked one of a number of (mutually exclusive) tenable stories to tell about what you said. (b) One might challenge the inference from
it is determinate that I understand you, by using a scheme which interprets you as saying that Jones is surfing
to
the truth value of my use of ‘you said that Jones is surfing’ is determinate.
The motivation is that determinate understanding is simply a matter of having an acceptable interpretation scheme. But such schemes will, because of inscrutability, issue in claims about what is said which don't have a determinate truth value.
In my view, (b) isn't coherent. It requires that we make sense of a situation in which we can correctly say something like this: Richard understands your utterances. In particular, he understands your utterance of ‘Jones is surfing’ to say that Jones is surfing. But it is not true that your utterance says that.
(a) is coherent. But it is, in my opinion, without credible motivation. First of all, it simply does not follow from the inscrutability thesis- i.e., from the relatively uncontroversial thesis that there are different schemes of reference one could give for your and my languages- that it must be indeterminate whether or not I understand you. Suppose that our respective languages are so related that, while it may be indeterminate what our uses of ‘Jones’ and ‘is surfing’ refer to, it is determinate that whatever one refers to, the other does. This is consistent with the weak thesis. And it seems to imply that, if I interpret your ‘Jones is surfing’ with my own, I understand you.
Furthermore, the response in fact constitutes a retreat from the official position of Davidson and (perhaps) of Quine, according to which there really are semantic and intentional phenomenon, such as understanding. It is one thing to say, as Quine and Davidson appear to, that there are different, but equally good ways for one person to understand another's linguistic behavior. It is quite another to say that there is no such thing as understanding another's linguistic behavior, or that it is never ‘strictly speaking’ true that I understand you. It is one thing to say, as Davidson and Quine do, that there are different acceptable accounts of what a person is referring to, since we can make changes in what we say the person believes to adjust for the variations in the account we give of their language; it is another to say that no matter what we might say, about what a person believes and what he means, there is a perspective, adequate to all the facts there are, from which it must be said that we have gotten it wrong.
27 The objection that follows and its example are based upon comments by Brian Loar, though he should not be held responsible for the form I've given it in. Other members of the Rutgers Philosophy Department pushed similar lines of objection.
28 This last sentence is, I think, ambiguous, and may have some readings which don't in any interesting sense follow from the fact that A's expressed a belief with ‘dogs bark.’ But there is a reading which does follow.
29 This is probably too simple minded. If I utter ‘no rabbit is crossing the road,’ I am disagreeing with someone who utters ‘a rabbit is crossing the road.’ But it is not clear that I am denying the sentence ‘a rabbit is crossing the road.’ I certainly didn't utter it.
So far as I can see, this sort of issue, as interesting as it is, is irrelevant to the issues at hand, and so I propose to ignore it.
30 I grant this is hyperbolic, though I think it is true.
31 Davidson, ‘Radical Interpretation,’ 134-5Google Scholar
32 It might be said that the response in (F) undercuts the motivation in (B). (B) offers a partial explanation for how it is that my interpreting you in a certain way R can cause my terms to have certain semantic properties. The explanation is in terms of such interpretation having certain intentional effects, or, at least, being accompanied by certain intentional states ('committments,’ which are, inter alia, combinations of intentions and beliefs). But (F) claims that the proposal I am making does not require us to see semantic facts as being determined by anything other than non-intentional phenomena and ‘very basic’ intentional phenomena. But wouldn't the intentional states involved in the committements be ‘non-basic’?
Assuming it is not going to completely discredit the passage from Davidson cited in the text, a response to this worry would need to show that the commitments in question are determined by non-intentional and ‘very basic’ intentional phenomena. I think this could be done, if we count as ‘very basic’ intentional phenomena such things as wanting (and intending) to be able to argue with and convey information to others. (The sort of desire I have in mind is general, not specific- wanting that for many p, I am able to argue about p, not wanting to be able to argue about p, where p is some fixed proposition.) For one might argue that, (1) all else being equal, people who have a common goal and are behaving in a way w such that one would, on reflection, recognize w as being the most likely way, given how things stand, to achieve the goal have a committment to continue to behaving in way w. But (2) people who are engaged in mutual intepretation have a mutual goal (being able to argue and inform) and a way of behaving (mutually interpeting) which satisfies this condition. So, (3) all else being equal, such people are committed to continuing to mutually interpret in the way in question.
If some such argument is sound, it derives the existence of the prima facie committment from the existence of general goals and the fact of mutual interpretation. And if such are derivable therefrom, they are determined thereby. So if what's said in (F) is correct and the general goals are ‘very basic’ intentional phenomena, the worry is put to rest.
33 Thanks, for comments, objections, and discussion, to Jody Azzouni, David Braun, audiences at Yale and Rutgers Universities, and an MIT reading group.