Hanks' Univocality Objection to Theories of Interrogative Content

In his paper, “The Content-Force Distinction”, Peter Hanks claims that three prominent semantic accounts of interrogatives—those of Hamblin (1973), Karttunen (1997), and Groenendijk and Stokhof (1982, 1997)—are all inadequate because they fail to preserve the univocality of “knows that” (e.g. Jones knows that Smith is tall) and “knows whether” (e.g., Jones knows whether Smith is tall). I will not examine the three accounts of interrogatives that Hanks impugns here.  Nor will I attempt to assess the various replies to Hanks’ argument that may be offered on behalf of the three approaches in order to show that they can, in fact, preserve the univocality of “knows” across cases of “knows that” and “knows whether”. Instead, I will examine Hanks’ argument directly, and challenge the assumption that the term “knows” is univocal in the way Hanks suggests.

Hanks begins by noting that the term “knows” is genuinely ambiguous when used in its acquaintance sense (e.g., “Jones knows Wilson”) and in its propositional sense (e.g., “Jones knows that Smith is tall”). This is what explains the infelicity of the following sentence:

(1)  Jones knows that Smith is tall and Wilson.  

In short, it is ambiguity of “knows” in its acquaintance and propositional senses that explains the infelicity of (1).  Hanks draws the preliminary conclusion that one cannot form a conjunction with a genuinely ambiguous use of “knows” (in the manner illustrated in (1)) without infelicity. So goes the first stage of Hanks argument.

The second stage of Hanks’ argument consists in his contention that the difference between “knows that” and “knows whether” is best conceived of as the difference between a knowledge claim that modifies indicative content and a knowledge claim that modifies interrogative content, respectively. In support of this conclusion, Hanks invites us to consider two statements:

(2) Jones knows that Smith is tall.(3) Jones knows whether Smith is tall.

It is clear that (2) and (3) have different contents.  Suppose that Jones knows that Smith is not tall. In that case, (2) would be false, while (3) would be true.  But how are we to make sense of the difference between (2) and (3).  It is tempting to view (3) as a disjunction of two knowledge claims, along the lines of (4):

(4)  Either Jones knows that Smith is tall or he knows that Smith is not tall.

On the present suggestion, “knows whether” is elliptical for a disjunction of a pair of competing “knows that” claims.

However, Hanks maintains that the above analysis of “knows whether” is unsatisfactory.  This is because the commitment to compositionality requires that we see expressions like “whether Smith is tall” as making the same semantic contribution to different sentences in which it occurs. But this requirement cannot be satisfied if we conceive of (3) along the lines of (4).  For example, consider (5):

(5)  Jones asked whether Smith is tall.

If we assume that “knows whether” statements are an elliptical treatment of a disjunction, then (given the commitment to compositionality) we would have to interpret (5) along the lines of (6):

(6)  Either Jones asks that Smith is tall or Jones asks that Smith is not tall.

But (6) is clearly infelicitous.  Hence, interpreting “knows whether” as a disjunction comes at the cost of compositionality.

According to Hanks, our best chance of making sense of (3) in a way that satisfies the compositionality requirement is to hold that the embedded expression “whether smith is tall” is an instance of interrogative content.  How we understand interrogative content will of course depend on our theory of interrogatives, which is what is under dispute in Hanks’ paper. But the basic idea is this: the expression “Jones knows whether Smith is tall” is best interpreted along the lines of “Jones knows the answer to: is Smith tall?” This is the conclusion of the second stage of Hanks argument.

In the third, and final, stage of his argument, Hanks observes that the following claim is felicitous:

(7)  Jones knows that Smith is tall and whether Wilson is married.

According to Hanks, this suggests that “knows” is not ambiguous across “knows that” and “knows whether”. 

Given that “know that” is a knowledge claim with embedded indicative content, and “know whether” is a knowledge claim with embedded interrogative content, it follows that “knows” is not ambiguous across indicative and interrogative content. Hence, Hanks concludes that those accounts of interrogative content that fail to preserve the univocality of “knows” across indicative and interrogative contents are mistaken. 

Here is a recap of all three stages of Hanks’ argument and his conclusion:

Stage 1: One cannot form a conjunction with a genuinely ambiguous use of “knows” without infelicity. 

Stage 2: “knows that” and “knows whether” correspond with knowledge claims with embedded indicative content and interrogative content, respectively.  

Stage 3: One can form a conjunction with “knows that” and “knows whether” without infelicity.  

Conclusion: Knowledge claims with embedded indicative content and embedded interrogative content do not involve a genuinely ambiguous use of “knows”.

I believe that the first stage of Hanks argument is unsound.  Consider the following sentence:

(8) Jones knows how to play the piano and that the concert is being held downtown.

(8) is no less felicitous than (7).  However, (8) features an ambiguous use of “knows”. The first conjunct is an instance of “know how” while the second conjunct is an instance of “know that”.  Or, for another example, consider (9):

(9) Jones knows how to play the piano and whether the concert is being held downtown.

(9) is also perfectly felicitous. But it too involves an ambiguous use of “knows”.  The first conjunct is an instance of “know how” and the second conjunct is an instance of “know whether”.

(8) and (9) illustrate that is possible to have a conjunction that features the ambiguous use of “knows” without infelicity. Hence, we cannot conclude from the fact that (7) is felicitous that it involves a univocal sense of “knows”.  The upshot is that it is unclear that three prominent theories of interrogative content Hanks mentions have the burden of preserving the univocality of “knows” across “knows that” and “knows how” since the claim that there is such univocality remains unestablished.

On Kripke's "Identity and Necessity"

In this blog post, I will like to consider the implications (if any) of the intelligibility of the following scenario for Kripke’s claim that identity statements involving rigid designators are necessarily true.  Suppose that the Messenger space probe were to send back pictures revealing that “Venus” is actually a binary planetary system—two planets orbiting around a common gravitational midpoint.  Moreover, as luck would have it, when viewed from the northern hemisphere on any given morning, one of the twin-planets (let’s call it V1) would stand perfectly in front of the other (let’s call the second V2), so that only V1 could be seen.  But in the evening, the planets would switch positions so that V2 would obscure V1 such that only the former would be visible.  The upshot of this is that the heavenly body that was visible (from Greece or Italy) in the morning and that was “tagged” with the name Phosphorus would actually be distinct from the heavenly body seen in the evening and which was “tagged” with the name Hesperus.  Now such a scenario is clearly fictional and may, given what we know about the laws of planetary motion, even be physically impossible.  However, it certainly strikes me as intelligible and I wish to consider what follows (in anything) from the simple fact that it is intelligible. 

Suppose, after considering the above scenario, I were to say something to the effect: “If Phosphorus and Hesperus turned out to be different heavenly bodies, then Kripke would be left without a cogent illustration for his identity thesis.”  Presumably, Kripke would hold that it is impossible for the antecedent of the above counterfactual statement to be true.  Moreover, the kind of impossibility that Kripke would have in mind would be logical and not merely physical.  In brief, Kripke holds that both names, Phosphorus and Hesperus, are rigid designators.  Even so, this does not impugn the intelligibility of anything said thus far since we do often employ counterfactuals in which the antecedent is logically impossible (such as in reductio arguments). 

In Kripke’s version of the Phosphorus-Hesperus scenario, the gravitational influence of a passing comet causes Venus to be moved from its normal position and Mars to take its place.  He then suggests that under such circumstances, the thing to say would not be that Phosphorus is not Hesperus (i.e., the two names designate different heavenly bodies), but that Phosphorus has been moved so that it no longer occupies its normal position in the night sky.  Now, it does not seem to me that this reply works in the case as I have described it.  This is because in the binary-planet case we are not simply imagining a case in which another heavenly body, say Mars, is occupying the position we once thought was occupied by Venus.  Rather, we are imagining a case in which what we took to be a single object, Venus, actually turns out to be two distinct objects.  What would Kripke say in response to the above example?

In attempting to answer this question it may be helpful to consider the following question:  “In the aforementioned counterfactual, which of the two heavenly bodies, V1 or V2, should we identify with Venus?”  Given the parity between the two cases (we can suppose that in astronomical observations V1 was identified as Venus just as often as V2) it would seem arbitrary to say that one, and not the other, ought to be identified with Venus.  Since both could not be Venus (personally I’m not convinced of this, but it seems like the position that Kripke would hold) then the only available option would be to say that neither heavenly bodies are to be identified with Venus.  If this is right, then the proper name “Venus” would cease to refer to anything.  (This conclusion seems to jive with Russell’s requirement for successful reference.)

Let us suppose that in the binary-planet example, the proper name “Venus” ceases to refer.  What about the status of the proper names “Phosphorus” and “Hesperus”?  Stated in the terminology of possible worlds, a rigid designator is defined as one that designates the same thing in all possible worlds in which it designates.  Suppose that the names “Phosphorus” and “Hesperus” are both rigid designators, and that “Phosphorus is Hesperus.”  (Strictly speaking, there is no need for imagination here, since this is what Kripke takes to be the case in the actual world.)  Since they are rigid, they each designate the same thing in all possible worlds in which they both designate. That is to say that “Phosphorus is Hesperus” is true in all possible worlds in which “Phosphorus” and “Hesperus” both designate, and hence that “Phosphorus exists & Hesperus exists → Phosphorus = Hesperus” is true in all possible worlds, and is therefore necessarily true.  This, according to Kripke, is all we mean when we say that “Phosphorus = Hesperus” is a necessary truth. 

In my binary-planet example, the antecedent of the conditional (Phosphorus exists & Hesperus exists) is true and the consequent (Phosphorus = Hesperus) is false.  It therefore appears to represent a counterexample to Kripke’s claim that Hesperus and Phosphorus are rigid designators.  Now, I have already conceded that the situation described in the antecedent may be impossible.  Moreover, perhaps Kripke should be interpreted as allowing that even a rigid designator may designate different things in impossible worlds.  If so, then perhaps there is also an expectation on Kripke’s part, that we restrict ourselves to counterfactuals with possible antecedents whenever we unpack the definition of rigid designator in terms of possible worlds. 

However, on pain of circularity, such a move seems illegitimate at this stage.  Kripke is supposed to be offering us an argument for necessity of identity from his definition of rigid designation.  But if Kripke’s definition of rigid designation already required us to omit impossible worlds, it would be to argue for the conclusion that a certain identity is necessary from the starting assumption that the identity failing to hold is impossible.  But that would be to settle before hand the position one wished to achieve by argument.  If Kripke’s argument is to be cogent, it must be possible to begin with the initial supposition that the antecedent it possible.  Then if the argument is successful, we would be led to reject our initial supposition.  But once we grant that the antecedent is possible (i.e., that Phosphorus is not Hesperus), it is not clear that any argument can be given to show that the possibility that has been assumed for the sake of argument should be rejected.  In sum, without some independent argument showing that the binary-planet counterfactual, in which the names “Phosphorus” and “Hesperus” designate different objects, is impossible, it is not clear that Kripke’s argument that the names are rigid designators gets off the ground.

Language and Thought

What's wrong with Frege's argument for Senses (Brian Kim)

Brian Kim responds to Jackie Carter's post on Frege puzzles.

I agree with Jackie that much of the motivation for many crazy views about truth, concepts, and intentional content derive from their ability to explain Frege Puzzles. I for one would like to separate my theory of meaning from a theory of intentional content. Like Stalnaker, I think thought is prior to language. I think this is the minority view in philosophy these days so in order to motivate this view, one needs to undermine the main sources of motivation for the alternative - Frege's argument for the distinction between sense and reference. In fact, Frege's argument does show that there is a distinction between sense and reference, but what it does not show is that senses have anything to do with thought.

Here's one way to present Frege's argument:
So the cognitive significance of the sentence P: "Peter Parker is Spiderman" is different than Q: "Peter Parker is Peter Parker". After all they pick out different thoughts. Are thoughts the referent of sentences? Well if they were, then the referent of the sentences P and Q must be determined by the referents of its parts. Thus, if we substituted co-referential names, the referent of the sentences should stay the same. However, if we substitute 'Spiderman' for 'Peter Parker' in P, then referent of P would be the same as Q. Thus, the referent of P and Q cannot be the thoughts they express.

The problem is that Frege simply assumes that only sentence types pick out thoughts. Sure, if we only consider the referent of sentence types, then that must be determined by the expression types 'Peter Parker' and 'Spiderman'. Then if we agree with Kripke, then names (expression types) are rigid designators, the reference of the sentence types P and Q cannot differ. However, sentence tokens can refer to thoughts as well. On this view, the referent of sentence tokens P and Q must be determined by the reference of the expression tokens 'Peter Parker' and 'Spiderman'. However, these two tokens don't have to refer to the same thing. I can use the token 'John' to refer to whomever I want. To Peter Parker, Bruce Wayne, or the dog walking down the street. There aren't any constraints on the reference of tokens. In one context, the token 'Mike' will be taken to be a token of the expression type 'Mike' (Brent) and in another context it will be taken to be a token of type 'Mike' (Seifried). The context plays a large part in determining the reference of expression tokens while the linguistic meaning determines the reference of expression types. So now, P and Q as sentence token clearly can refer to different thoughts because the expression tokens 'Peter Parker' and 'Spiderman' can refer to different people depending on the context.

Now what explains the cognitive significance of identity statements? Well P can be used to pick out different contingent propositions while Q will always pick out the necessary proposition if we are using the tokens 'Peter Parker' the same way. Of course, "Paderwski is Paderwski" can be used to pick out a contingent proposition.

What's the lesson here? Well the lesson is that Frege's assumption that only sentence types refer to thoughts would only be accepted by someone who thought that there was strong connection between thought and langauge. On this view, the way we pick out thoughts is highly dependent upon the nature of language (of expression types). Then of course using Frege's argument we would be forced to accept that thought is fine-grained and language like. However, if we think that sentence tokens plus context pick out thoughts, then we place a greater divide between thought and language. Of course, language is still important, but language is merely a means of picking out thought - it isn't a mirror image of thought.

McDowell's Truth Theory and Frege Puzzles (Jackie Carter)

In what follows, Jackie Carter offers a clear and concise summary of McDowell's proposed solution to the Frege Puzzle. Enjoy!

So, I think that McDowell's concept of truth seems somehow unsatisfying on its face.... [H]owever, I came to find it much more palatable -- in part because of its implications for philosophy of language and in particular the Frege Puzzle. Anyway, this is my effort to explain why McDowell's truth theory, which makes "true" a status internal to a world view, helps solve the elusive Frege Puzzle. (Disclaimer: my grasp of this feels rather tenuous.)

Recall that (briefly put) Frege's Puzzle is that "Hesperus=Hesperus" seems to have a cognitive value different from that of "Hesperus=Phosphorus." The problem is that if all the proper names do is pick out some referent (some object in the world), then the two identities seem to do the very same thing - the latter seems no more informative than the former. Intuitively, Frege thinks we should avoid this conclusion, so he proposes that names not only pick out referents but also have "senses." So, in the first identity, the sense of the name is the same on both sides, while in the second identity the senses are different. QED, the second identity is of greater "cognitive value" because it involves equating two different senses.

But then the question is what exactly is the "sense" of a word? Frege wants it to play the intensional role of meaning, though he leaves the concept of sense rather mysterious (which then leads Russell and Searle to offer some rather problematic explanations).

McDowell distinguishes between sense and reference as knowledge of truths on the one hand, and knowledge of objects on the other. So, someone who knows the sense of "Hesperus=Phosphorus" has knowledge of a truth, which someone who only knows "Hesperus=Hesperus" is lacking.

{McDowell writes, “Knowledge of the reference of a name…could reasonably be held to be knowledge which, in the context of appropriate further knowledge not itself involving the name, would suffice for understanding utterances containing the name – that is, precisely, knowledge of its sense” (McDowell's "On Sense and Reference," p. 163). }

But if we understand "sense" to be knowledge of truths, we run into problems if truth & truth conditions are purely extensional (as in a deflationary account like Rorty's or even Davidson's). Truth (and along with it, truth conditions) is a mere translational relation between object language and metalanguage, but the "sense" that a given individual associates with a word, sentence, or phrase is supposed to be intensional. Moreover, different people know different truths about different names; e.g., the truths that I know about the object "Hesperus" may differ from those of another. So, how can sense just be knowledge of truth conditions, which would be objective (external, the same for everyone).

The point here is when we say:

"Hesperus = Phosphorus" is true iff Hesperus is Phosphorus,

the right hand side (RHS), which captures the "sense" of the identity, may be different for different agents. For McDowell, truth is intensional, so when we disquote (ie, move over to the RHS), we are already talking about the conception of an agent (we are internal to the world view of an agent). Because truth and truth conditions are intensional, sense is intensional, which enables it to play to intensional role of meaning. That is, the notion of sense, knowledge of truths or truth conditions, can thus capture the intensional notion of meaning.

In short, McDowell’s understanding of truth as internal to the worldview of an inquirer allows him to make truth, and therefore sense, an intensional concept. For McDowell, an account of truth conditions cannot be decoupabled from an account of agents’ beliefs. The right hand side of a Tarski sentence gets at an agent’s worldview and tells us something about her beliefs.

For Davidson, by contrast, a truth theory must include both an extensional account of truth conditions and an intensional account that derives from empirical human behavior. For McDowell, there is no reason to separate these accounts. It is unclear how (and if) someone with an extensional, deflationary truth theory like Davidson's can explain meaning, i.e. sense. So the Frege puzzles seem to go unsolved if truth conditions (the RHS of T-sentence) do not get us to the intensional perspective.

Grice on Meaning

Grice’s paper, “Meaning”, represents a significant shift from approaches to the meaning of utterances that look to the meaning of the words used to one which looks, instead, to the content of the mental or psychological states of speakers. Grice begins by drawing a distinction between two senses of ‘mean’ as it occurs in sentences of the form:

(*) x means that p.

Where x ranges over objects which have meaning and p over declarative sentences. Grice illustrates the two senses of ‘mean’, which he calls “natural” (N) and non-natural (NN), with the following two examples:

(N): Those spots mean measles

(NN): Those three rings on the bell (of the bus) mean that the bus is full.

Grice maintains that sentences like (N) are factive, while sentences like (NN) are not. For example, Grice notes that it would be contradictory to say:

(N*): Those spots mean measles, but he hasn’t got measles

This is because in the case of natural meaning, sentences which have the form ‘x means that p’ entail ‘p’. It should be noted that strictly speaking the schema ‘x means that p’ entails ‘p’ does not apply to (N*) since the left conjunct does not contain an appropriate substitution instance for ‘that p’. Thus, (N*) should be rewritten as follows:

(N**): Those spots mean that he has measles, but he doesn’t have measles

The same holds true for (N). This minor incongruity in his example aside, I take Grice’s general point regarding the factivity of natural meaning to be well placed.

By contrast, the non-natural sense of ‘mean’ is non-factive, so that the schema ‘x means that p’ does not entail ‘p’. This is easily seen when one considers that from the fact that the conductor rings the bell three times it does not follow that the bus is full. At best, it only follows that the conductor thinks that the bus is full. Generalising from the examples given above, the difference between natural meaning (henceforth ‘meaningN’) and non-natural meaning (henceforth ‘meaningNN’) is this: it is not consistent with something’s having a meaningN that what it meansN is false; but it is consistent with something’s having a meaningNN that what it meansNN is false. (Grice discusses additional differences between the two cases, but I take this to be the main one).

Grice distinguishes between natural and non-natural meaning in order to avoid confusion. However, it is the latter that he is primarily concerned with since examples of meaning that involve language (Grice’s main focus) are typically cases of meaningNN. Grice attempts to show that ultimate source of an utterance’s meaningNN is the mental content of the speaker. He attempts to do this in two steps:

(Step 1) Occasion meaning: give a definition of single utterances couched entirely in terms of the speakers’ intention to produce certain effects in their audience.

(Step 2) Timeless meaning: give a definition of expression meaning couched entirely in terms of the definition of idiolect meaning given in (Step 1).

If Grice’s overall project is successful, then we could eliminate the semantic notion of timeless expression meaningNN in favour of the psychological notion, thereby showing that it is the mental states of speakers, rather than the meaningNN of expressions, that are the ultimate source of an utterance’s meaningNN. After consider two alternative definitions, which he rejects as insufficient, Grice settles on the following definition of idiolect meaningNN:

(IM): A specific utterance φ meansNN that p, if and only if, in performing it, the utterer intends:
(a) that an audience will come to believe that p, and
(b) that this audience will recognise intention (a), and
(c) that the recognition in (b) will cause belief in (a)

The timeless meaningNN of an expression, in turn, is defined as follows:

(TM): An expression θ meansNN that p within a certain linguistic community if and only if most utterances of θ by members of that community meanNN that p.

One difficulty with the above definitions, (IM) and (TM), is that they say nothing about word meaningNN. Even if we grant that Grice is right to claim that a sentence means what it does because of regularity in the meaning of utterances using it, it is not clear that the meaning of words can also be defined in terms of such regularities. After all, we often do use words in novel combinations, and the meaning of each individual word may change slightly given where it appears in the sentence and the other words around it. Moreover, it is not clear how, give Grice’s claim that the meaning of a sentence has to do with how it is regularly used, it is not clear how I can utter a completely novel sentence, “She was erratic like a lunatic chimpanzee on a merry-go-round!”, and it still be meaningful. Of course, if we did have a theory of word meaning, we would simply need to show how the meaning of the novel sentence arises from the meaning of each of the words. However, the above objection may not be as serious as it first appears. While not explicitly a theory of word-meaning, word-meaning can be seamlessly integrated into Grice’s account of timeless meanings. On this picture, word-meaning is simply a function of how words, as opposed to sentences, are commonly used by a particular linguistic community.