Knowledge of Meaning

Notes from a graduate class with Ian Rumfitt on how we know what things mean.

Week 1: The thesis

1. The thesis

Wittgenstein: 4.024 Einen Satz verstehen, heißt, wissen was der Fall ist, wenn er wahr ist (‘To understand a declarative sentence means to know what is the case if it is true’).

• Frege (parent): name of a truth value $\approx$ formalized sentence ($\approx$ a function from circumstances of evaluation to truth values?)
  • Wiggins: this is about Frege’s Begriffsschrift, but can be generalized!
• Wittgenstein (siblings): ‘(4.022) A declarative sentence in use shows how things stand if it is true. And it says that they do so stand… (4.061) A declarative sentence in use is true if we use to say that things stand in a certain way and they do.’
• Davidson (child): Truth-conditions are (a type of) meaning.
• Strawson: ‘Generally harmless and salutary…not the end, but the beginning of our task’ (One thing to be true, another to be important) Meaning isn’t truth, but what beliefs (or wishes or instructions) you can use it to communicate. (‘Audience Directed Belief Expression’)
  • Rumfitt is aligned with Strawson here (speech act theory); but we’re aligned with Davidson (for the semantics, at least; the pragmatics we take to be built on top). Common knowledge of truth-conditions is just what allows performance of these communicative speech acts.
  • Full understanding requires, say, catching allusions; but knowledge of truth-condition is a central dimension.
• Obvious generalizations: to understand an imperative is to know under what conditions it is obeyed; to understand an interrogative is to know…?

Do we really want to say our philosophical opponents who disagree with us about the meanings to sentences don’t understand the sentences they’re uttering?

Why assume anything like this is true? I don’t know the truth-conditions in any transparent guise (say, with a disquotational schema)? What if it’s really, really context-sensitive? What if the truth-conditions are vague, and so unknowable?

2. Sentences

• On a broadly Strawsonian view: roughly, something with which you can perform a complete speech act. By contrast with some expression which (together) has type . Consider: ‘Hello, Jemima’; ‘Hurrah for Manchester United!’
• Frege’s context principle: only ask for the meaning of a word in the context of a sentence. ‘Only in the context of a sentence does a word mean anything’. (Wait, what? Seems quite weird, and in tension with Frege’s thesis.)

3. Individuation

• Austin: To perform a locutionary act is to (a) utter a phone (certain noises), (b) utter a pheme (certain lexical items in a grammatically okay way), and (c) perform a rhetic act (do so with a more or less definite sense and reference).
• Example: The same phone ‘flying planes can be dangerous’ corresponds to two English phemes.
• Also useful to treat lexical items as having their meanings necessarily, and ‘other meanings’ as different homophonous items.
• We identify sentences with phemes; in particular, those whose utterance may perform a complete speech act.

Is it open to the Strawsonian to individuate things in this way? It seems like you have to buy some realist-y notions. (Or, I guess I don’t understand how you can go syntax-first.) ‘Ann!’ seems to count (as a greeting from a friend, an imperative from a parent, …) without any obvious elided content (versus in answer to, say, ‘Who smokes?’). So, does ‘Ann smokes’ contain the subsentence ‘Ann’?

Response: in ‘Ann smokes’, ‘Ann’ isn’t used as a sentence; so it’s not a subsentence. But ‘Ann smokes, and Jane dances’ doesn’t use ‘Ann smokes’ as a sentence, yet ‘Ann smokes’ is surely a subsentence in a way that ‘Ann’ isn’t!

‘The proposition that arithmetic is incomplete was first proved by Godel?’ Does the ideology of propositions even make sense here, for necessary truths?

4. Classification

• Claim: it’s unfruitful to try and find differences in speech acts, say, in telling someone that something is the case, or to do something.
• Example: ‘In this house, we remove our shoes before entering’. (Similar: ‘You do it like this’.)
• In these cases, you’re committed to postulating some ambiguity, or something, between truth-apt uses and imperative uses.

Can we assimilate all imperatives to declaratives, where the imperative force is pragmatic? We derive ‘make it such that P’ (Altshuler) from a patently false ‘P’.

• Example: ‘A-company will parade at ten thirty tomorrow morning’. Same truth-conditions, but from a tour guide the speaker is responsible for the truth of the sentence, whereas from a general the audience is responsible for the truth of the sentence.

(Someone asks a discussion on why you can’t assimilate declaratives to imperatives. But declaratives-first is just, as a matter of fact, a much more successful and promising research program than the hypothesized imperatives-first alternative.)

I wonder what the imperative mood in other languages looks like. Could be a challenge to the assimilation!

5. Declaratives

• ‘No one would assign truth-conditions to “let’s go”‘. (But why not? ‘[You] let us go’ is true iff the audience does let them go.) I thought the proposal was to assimilate all of these apparently non-declarative things to pragmatic derivations from declaratives? But The answer here was that we need to keep in mind even distinctions between, say, ordering and telling to and inviting in order to classify sentences.

(We ran out of time before finishing the handout.)

Week 2: The Thesis (cont’d)

1. Declaratives

• Frege: ‘What is true is independently of our recognizing it as such’ (‘implausible to suggest that sentences would exist independently of us’: shouldn’t this be about propositions, though?)
• McDowell: ‘The truth-condition of a sentence may obtain, or not, independently of our capacity to tell that it obtains, or that it does not’
• Wiggins: ‘Every statement which lacks (possesses) truth lacks (possesses) it independently of a sepaker’s means of recognizing it […] something potentially resistant, and outside the act of judgement itself, upon which the mind can gain purchase and go forward in judgement or denial’ (‘so, the source may be the judgment of others’: so Wiggins says this is necessary, but does he say that this suffices, such that just the judgments of others suffice?)
• Frege:
• • ‘Botticellis _La Primavera ist ein altes Tafelgemalde’ (‘It’s old’)
  • ’[…] ein shones Gemalde’ (It’s beautiful’)
• Dummett’s Test: can it serve as the antecedent of a conditional?
• • ‘If Reagan was a KGB agent, no evidence to that effect will ever be found’
  • On Frege-Geach: ‘reflects the peculiar power of the context to demand interpretation in terms of truth-conditions’
• Not reliable (!?): ‘if it works, it works only for English and not for very close relatives’. In Gibbardese, ‘Act $A$ is the thing to do’ expresses a plan to perform $A$: but, ‘packing is the thing to do’ can still serve as an antecedent. (Surely the test isn’t meant to be definitive, such that artificial counterexamples could be cooked up? But maybe that’s the point. )
• ‘If the cup broke if it was dropped, then it was fragile’ / ’[…] If (it’s the case that) Kripke was at the conference if strawson was, then Anscombe was there’

2. Neo-Ramsey

• Frege: Thoughts are not private, else disputes would be futile. (Really beautiful passage!)
• • But: ‘Let’s go to Quod’ / ‘No, let’s go to the Old Parsonage’ (But does Frege require that this is sufficient for being truth-evaluable, or just necessary? Further, maybe we do take these to be truth-evaluable after all…)
• Ramsey: $S$ is [true/false] iff $\exists P(S$ expresses the thought that $P\wedge [P/\neg P])$
• • Moral: propositions are just those things manipulable by the propositional calculus
  • For instance, Lewis’s argument about conditionals relies on the logical equivalence between $X$ and $(X\wedge C)\vee (X\wedge \neg C)$.
• Claim: the ‘morass’ of debate around the truth-conditions for indicative conditionals (and generics) matches their borderline embeddability within the propositional calculus.

3. Real declaratives

• Dummett: A sentence may be used to make a claim if those who understand it know what would justify an assertive use of it. (Conditionals, moral statements are fine. But of course we who think they’re just straightforwardly propositional agree.)
• • Non-classical negation, according to which the negative utterance ‘Not B’ expresses a second-order claim to be entitled to make a claim incompatible with the utterance of ‘B’
  • (Does a knowledge norm on assertion get you something like this?)
  • These are complicated contraries, but not negations (I think that’s right)
• Gibbard: Negation is disagreeing (entails not-agreeing) with a claim. But fails to validate disjunctive syllogism
• • Either packing now is the thing to do, or by now it’s too late to catch the train anyway; but it’s not too late to catch the train, so packing now is the thing to do.
  • Failing to rule ‘packing now is the thing to do’ (keeping it as a live option) is not going to entail deciding ‘packing now is the thing to do’ outright. (And he thinks the first premise is just ruling out rejecting both disjuncts, the second is rejecting the second.)
  • (But if disagreement is super weak, then the negation story doesn’t work out: to negate something, I should have to think of it)

Week 3: Context Dependence

1. Context-sensitive sentences

Apparently, some sentences are true regardless of context: ‘Two is a prime number’, ‘Nothing can travel faster than light’, perhaps ‘Torture is wrong’. But it just seems wrong that the English sentences always express their usual surface readings (which are necessarily true).

2. Higginbotham

Rumfitt (with Higgenbotham) wants to call utterances true or false, abbreviating that the speaker, in producing it, thereby speaks truely [falsely].

What Fred told Mary has no spatiotemporal location; but the rumour which swept through the common rooms of Oxford last week does; and both can be true.

Actually, couldn’t both be the same, say, that Alice loves Beth? And that seems to have spatiotemporal location. But plausibly all the ascription of spatiotemporal location is mistaken, here. Or we could say that the rumour is just what Fred told Mary, or that what Fred told Mary was just the remour, or that what Fred told Mary later swept through the common rooms of Oxford.)

With puns, you leave it open whether they mean P or Q

I think there’s usually a surface reading!

‘In uttering “This is red”, an English speaker does not say that the unique object referred to by the component demonstrative pronoun is red’.

I mean, they certainly don’t say “the unique […]”, but it seems obvious that they do say that the unique […]! I might contradict them by saying ‘the unique object […] is not red.’ (Maybe my judgments have been ruined by coarse-grained views of propositions.)

Week 4: Davidson & Lewis

1. Indirect Questions

Rumfitt claims, with many, that ‘tell’ (and ‘ask’) are veridical / factive with respect to interrogative but not declarative complements. But consider:

1. Did Ann tell anyone that she’s dropping out to take a job in finance?
2. Did Ann say to anyone that she’s dropping out to take a job in finance?
3. The meteorologists tell us where it’s going to rain, but are often wrong.
4. # The meteorologists know where it’s going to rain, but are often wrong.

(1) but not (2) seems to license the inference that Ann’s dropping out to take a job in finance; meanwhile, (3) seems fine, while (4) sounds horrible. It’s also nice to have a uniform (or reductive) semantics for both types of complements.

Anyway, we can contrast interrogative and relative pronouns:

5. Mary knew what Fred had in his bag.
6. Mary hid what Fred had in his bag.
7. Hilbert already knew what Weyl was going to present to the seminar.

This difference is grammatically marked in formal German: Wer hart arbeitet, verdient eine Belohnung vs. Lacon fragte, wer der Maulwurf sei.

Consider the following argument:

What I immediately see depends upon the light conditions, but no material thing depends upon the light conditions. So, what I immediately see is not a material thing.

The first premise sounds plausible because we naturally hear it with interrogative ‘what’ (perhaps because it’s obviously false with relative ‘what’). But the conclusion is heard with relative ‘what’.

2. Davidson

Claim: One know what a sentence means just in case one knows what is the case iff the sentence is true.

With the biconditional read materially, this is clearly insufficient. Read strictly, this seems quite bad. But I’m sympathetic to intensionalism. The basic case: intensional semantics is the gold-standard research program, and so shouldn’t be abandoned; if you allow impossible worlds into your model, your semantics for logical terms basically aren’t compositional, and your model lacks too much structure to be very useful.

Rumfitt wants to revive a rule-following approach to semantics.

3. Lewis

(8a) A man came to the door yesterday

(8b) A certain man came to the door yesterday

These have the same truth-conditions, but they embed differently:

(9a) Mary heard that a man came to the door yesterday.

(9b) Mary heard that a certain man came to the door yesterday.

(10a) Mary was surprised that a man came to the door yesterday.

(10b) Mary was surprised that a certain man came to the door yesterday.

(11a) If a man came to the door yesterday, the butler told him to go away.

(11b) If a certain man came to the door yesterday, the butler told him to go away.

Rumfitt writes that this is a problem for modeling semantics with function composition, or even for the supervenience of meaning on truth-conditions. But that’s not right (and I was able to convince him and the class): a natural hypothesis is that ‘a certain X’ emphasizes the indefinite, making us more readily hear the reading in which it scopes over the whole sentence. Of course functions deliver different results if applied in different orders — doubling and adding five is not generally the same as adding five and doubling.

A more interesting case concerns stress patterns:

(12a) Jane gave ME the tickets.

(12b) Jane gave me the TICKETS.

(13a) Mary was surprised that Jane gave ME the tickets.

(13b) Mary was surprised that Jane gave me the TICKETS.

(14a) Mary did not know that Jane gave ME the tickets.

(14b) Mary did not know that Jane gave me the TICKETS.

(15a) Jane gave ME the tickets by mistake.

(15b) Jane gave me the TICKETS by mistake.

Again, Rumfitt asserts that such “cases should lead us to reject the functional model of compositionality”. But again, I think this is too hasty (and the class seemed to pretty widely agree). There’s a plausible Gricean story, which goes something like: the choice to say it in this way (rather than in that way, which would’ve been more natural) tells me more information. This sort of ‘manner implicature’ depends on an alternative being salient; the inference in (13b) is blocked if the speaker is (say) French and so emphasizes the last word of every sentence. (Class was mostly English and German, no there were no French people to object. Why are there so many Germans in philosophy of language? And, like, so many Italians in metaphysics?)