I want to try to make an approach to the "only sentences have meaning" view. (For now, I'll use the word "meaning" without saying what it is.)

We'll all agree, I think, that only statements (let's leave aside everything else for a bit) can have a truth-value. But we need to distinguish somehow between grasping the meaning of a statement and knowing whether it is true or false.

So what's the meaning of a statement? What do you know when you know what the meaning of a statement is? One widely discussed view is that when you know the meaning of a statement, you know what conditions must obtain for the statement to be true, so that should do for now.

Let's take two statements, "It is raining," and "It was raining." One can be true when the other is false, so they must not have the same meaning. The proposal is that we grasp the meaning of a statement by comparing it to other statements that we already know the meaning of. So let's try that here.

Suppose our stock of statements include, besides the two at issue, "It is sunny," "It is cloudy," "It is snowing," and the past tenses of all those. So maybe we learn to say "It is raining," when the other three present tense statements are false. In that sense, we know the meaning of "It is raining" and the other present tense statements comparatively.

But what about the past tense statements? How could you learn their meaning comparatively? Since the truth conditions of the present tense statements are exclusive, there's a clear path for comparison. But truth conditions for the past tense statements overlap with each other and with the present tense statements. It's just not clear to me how comparison takes hold here.

But what about the past tense statements? How could you learn their meaning comparatively? Since the truth conditions of the present tense statements are exclusive, there's a clear path for comparison. But truth conditions for the past tense statements overlap with each other and with the present tense statements. It's just not clear to me how comparison takes hold here. — Srap Tasmaner

I'm not sure how the issue of overlapping is relevant here. Surly not all truth conditions of past tense sentences overlap with each other - the truth conditions of "Frege was born in 1848" don't overlap with "Frege was born in 1850". Similarly, weather reports about the past presuppose either explicit or implicit reference to time and place. If I tell you out of the blue that is was raining without telling you when and where (and it's not clear from the context) then I haven't expressed with the sentence anything with definite truth conditions. So it seems to me that past and present tense sentences are analogous in this respect (and you said that present tense sentences don't pose a problem for me).

Another thing, you ask me how to explain the learning of the meaning of past tense sentences (the beginning of the quote). But this is not the question that I meant to answer, because it is not the challenge which (let's call him) the compositionalist poses to the contextualist (which is how I formulated the argument from the creativity of language). I think both parties agree that someone who can use only present tense sentences cannot learn to use past tense sentences on the basis of this knowledge alone - he must learn some new syntactic or semantic rules. The challenge is rather to explain simpler cases, which do seem troubling for the contextualist but not the compositionalist (because both of them have to explain how we learn sentences with new meaning, it doesn't come for free just because you accept compositionality). So for example, the contextualist should explain how can we understand new sentences in the past tense, even when we already understand some other sentences about the past. The compositionalist will say that it is simply a matter of reconfiguring the meanings of old words according to familiar rules, whereas the contextualist cannot say this since for him the smallest semantic or meaningful unit in a language is the sentence and not individual words. So the challenge here is to explain what it is, if not the meanings of the old known parts, that explains our ability to understand the sentence? (maybe this is not what you meant by your question, but just in case)

The past tense thing was just meant as a proxy for all the stuff we learn to talk about where we cannot directly check that the relevant truth conditions obtain. I think you hit a wall pretty quickly if all you have to go on is the truth and falsehood of statements.

Even the natural next step is blocked, which is recursively generating complex statements from simple ones using the logical constants. (And similarly for understanding such statements by analysing then into simple statements so coupled.) I don't see how you get the logical constants going at all.

If forced to choose, I'm saying the word is the basic semantic unit, not the sentence, so long as it's understood that the meaning of a word is the semantic contribution it makes to a sentence in which it is used. The statement is the unit of judgment, though, so "semantic" up there is really the wrong word. The word is the smallest unit of meaning.

If it matters, I think this is Frege's view as well, despite the way it is expressed in the Grundlagen, but that's considerably less important.

The past tense thing was just meant as a proxy for all the stuff we learn to talk about where we cannot directly check that the relevant truth conditions obtain. I think you hit a wall pretty quickly if all you have to go on is the truth and falsehood of statements. — Srap Tasmaner

Sorry I don't get your point. On my account you don't have to know the truth or falsehood of this or that particular sentence, only understand the truth conditions of some sentences (which is of course not the same as knowing whether they actually obtain).

Even the natural next step is blocked, which is recursively generating complex statements from simple ones using the logical constants. (And similarly for understanding such statements by analysing then into simple statements so coupled.) I don't see how you get the logical constants going at all. — Srap Tasmaner

Again, I don't see your point here... If as you say logical constants recursively generate sentences by combining other sentences, then how is the question about the meaning of single words supposed to arise here? I think that on the contrary, truth functional logic seems to be very congenial to the contextualist.

If forced to choose, I'm saying the word is the basic semantic unit, not the sentence, so long as it's understood that the meaning of a word is the semantic contribution it makes to a sentence in which it is used. The statement is the unit of judgment, though, so "semantic" up there is really the wrong word. The word is the smallest unit of meaning. — Srap Tasmaner

Perhaps there is a sense in which this whole debate is terminological. You could say that words have individual "meanings" on my account, but only if by "meaning" we understand something other than saying, e.g., that the meaning of the word 'cat' is cats (that is, that meaning can be explained by sub-sentential relation between words and things in the world (such as reference)). Rather knowing the "meaning" of a word is to know its logical function in different sentences, or seeing what is in common between different sentences that contain the same word (and I have in mind here Wittgenstein's "propositional variables" in the Tractatus, and it can be argued that Frege's object/concept/function distinction is a similar idea). So perhaps the real issue here is explanatory priority between words and sentences, rather than the question about the meaning of individual words (which can be understood in all sorts of different ways).

So what's the meaning of a statement? — Srap Tasmaner

A meaning of a statement is simply the mental associations you make with the statement, which includes representations, referents you take it to pick out (if you do), relations you take it to specify or claim, and so on.

What do you know when you know what the meaning of a statement is?

One assigns meaning.

One widely discussed view is that when you know the meaning of a statement, you know what conditions must obtain for the statement to be true, so that should do for now.

That's one thing that's the case when one assigns meaning to a statement (we could argue that it shoudn't count as a statement if that's not the case), but I wouldn't say that that's what meaning amounts to. That's rather analogous to saying that when one owns a car, one will have keys to start the car. But what it is to own a car is not exhausted by having keys to start the car.

Let's take two statements, "It is raining," and "It was raining." One can be true when the other is false, so they must not have the same meaning. The proposal is that we grasp the meaning of a statement by comparing it to other statements that we already know the meaning of.

That would only be the case in some instances. Not all.

But what about the past tense statements? How could you learn their meaning comparatively? Since the truth conditions of the present tense statements are exclusive, there's a clear path for comparison. But truth conditions for the past tense statements overlap with each other and with the present tense statements. It's just not clear to me how comparison takes hold here.

I'm not clear on what you'd find confusing here. But one way that you could comparatively grasp "It was raining" is by being familiar with "It was snowing", and then knowing the difference between rain and snow.

Sorry I don't get your point. On my account you don't have to know the truth or falsehood of this or that particular sentence, only understand the truth conditions of some sentences (which is of course not the same as knowing whether they actually obtain). — Fafner

You either analyse sentences into components or you don't. If only sentences have meaning, then their components are meaningless. You can't have your cake and eat it too. Your contextualist who notices "structural similarities" among sentences would be like the guy who reasons that "ball" and "balk" and "balm" must have similar meanings. If you don't get to use components, all you have is truth conditions.

Again, I don't see your point here... If as you say logical constants recursively generate sentences by combining other sentences, then how is the question about the meaning of single words supposed to arise here? I think that on the contrary, truth functional logic seems to be very congenial to the contextualist. — Fafner

It's the logical constants themselves that you don't get to have, because by definition they're meaningless.

I think I'm okay with your last paragraph, mostly at least, but I'm at work, so...

You either analyse sentences into components or you don't. If only sentences have meaning, then their components are meaningless. You can't have your cake and eat it too. Your contextualist who notices "structural similarities" among sentences would be like the guy who reasons that "ball" and "balk" and "balm" must have similar meanings. If you don't get to use components, all you have is truth conditions. — Srap Tasmaner

It can't be right that if something is composed of meaningless parts, then it itself must be meaningless (if this is what you meant). Surely, the letters of the alphabet from which words are composed are semantically meaningless, and yet the compositionalist claims that words are meaningful (and the same applies to sentences, they are also composed of meaningless letters). So if you are right, then even on the compositionalist view all words (and therefore sentences) are meaningless - and that can't be right.

The same applies to what you said about logical connectives.

If words are the smallest units of meaning, then letters are meaningless. If sentences are the smallest unit of meaning, then words are meaningless. But it is clearly possible to hold that words are meaningful and can be combined into larger meaningful expressions, including sentences. I so hold.

The way I construe the meaning of a word, the logical constants are meaningful.

[Edit: wrote "connectives" instead of "constants."]

It can't be right that if something is composed of meaningless parts, then it itself must be meaningless (if this is what you meant). Surely, the letters of the alphabet from which words are composed are semantically meaningless, and yet the compositionalist claims that words are meaningful (and the same applies to sentences, they are also composed of meaningless letters). So if you are right, then even on the compositionalist view all words (and therefore sentences) are meaningless - and that can't be right.

The same applies to what you said about logical connectives. — Fafner

Letters are not meaningless; their meaningfulness consists in their relationships to sounds that can be made by the human voice and the diverse but phonetically constrained ways in which they can combine to form words. Words are not meaningless; their meaning consists in both their individual references and the relationships of similarity and difference between their various references as well as the diverse but constrained ways in which they can combine to form phrases and sentences.

I would say it's semantics all the way down.

Letters are not meaningless; their meaningfulness consists in their relationships to sounds that can be made by the human voice and the diverse but phonetically constrained ways in which they can combine to form words. Words are not meaningless; their meaning consists in both their individual references and the relationships of similarity and difference between their various references as well as the diverse but constrained ways in which they can combine to form phrases and sentences.

I would say it's semantics all the way down. — John

There's something to this, so long as, as you follow the chain up from phoneme to morpheme to word to sentence, you manage to mark the boundaries. (There's also a side trip some words take through "singular term.") The boundaries are important. It is arguably one of Frege's greatest legacies that he recognized the importance of the crossover from word to sentence.

It might be reasonable, helpful, or desirable, to use a different word for what we get at each step that we didn't have at the one before. The use of "meaning" at the word stage and the sentence stage does lead to confusion, but the usage is pretty well entrenched. There may be a darn good reason.

We don't have to stop at sentence, either. One next step that works for a lot of cases is language-game. You could also look to pragmatics, maybe speech-acts; whatever you call it, this would be the point at which you're looking not just for what a sentence means, but what is meant by its utterance. There is one more next step up from sentence that I think deserves special treatment, and that's inference.

When I say "meaningless" I mean semantic content, and save few rare exceptions (such as 'a' like in "a man"), letters don't have any semantic function analogous to words. If we put aside ambiguity (and a couple of other complications), then when you have two sentences with the same word, you know the word has the same semantic function (i.e., it's not a coincidence that we use the same word both in "the cat is on the sofa" and "I have a cat" etc.). Whereas if you have sentences with the same letters it tells you absolutely nothing about the meaning of the sentences since it's a pure coincidence that in English e.g., 'cat' and 'car' have two letters in common, and therefore from this you can't infer anything about their semantic content (it would be absurd to infer that 'cat' and 'car' must mean similar things because they share some letters - what about other languages where they don't?).

About the meaning of words, as I said in another post, it's a bit of a terminological dispute. The really interesting question is in what sense words have "meaning" and what is the relation between the meaning of words and the meaning of sentences, and which one is more basic. We can talk about the meaning of words in some sense, but on my view (as someone who takes the context principle seriously) the meaning of words is parasitic on the meaning of sentences, and that means that atomistic semantic theories (that try to explain the meaning of sentences on the basis of the meanings of their words) are false.

Yes I agree that letters do not have meaning which relates directly to the meaning of sentences as words do; I was just pointing out that they constitute their own layer of meaning as the 'basic pieces of the game' which also express our physicality and emotionality (as the sounds we can make). This is certainly not any kind of determinate propositional meaning, to be sure; it is more a kind of indeterminate phenomenological meaning. I think the more determinate meaning grows out of the basic meaningfulness that signs, including letters, words and sentences, but also everything else in the environment, have for us.

You could also look to pragmatics, maybe speech-acts; whatever you call it, this would be the point at which you're looking not just for what a sentence means, but what is meant by its utterance. There is one more next step up from sentence that I think deserves special treatment, and that's inference. — Srap Tasmaner

Here you are thinking of the intentionality of utterances and the inferences we make about the intentions of speakers as another layer 'above' the 'literal' meanings of sentences?

As Srap Tasmaner suggested, it's better not to use the word 'meaning' here, or any of its cognates, to describe what you say since it will only cause confusion. It doesn't show that "it's semantics all the way down" as you claimed.

It's not really worded well there. What I meant was there are several choices for where to go after sentences, and I gave some examples. I don't know if there is one next step, but there are candidates that I think are important.

One of those is pragmatics, and yes I think of what is meant by an utterance of a sentence as something other than what the sentence (literally, as they say) means, and that depends on loads of stuff, including but not limited to the intentionality of the speaker. (Sometimes pragmatics has been taken as the theory of indexicals, but there's some reason to think they ought to come in earlier, at the sentence level. It's tricky, right?)

One other place to go from sentences is inference, and here I'm talking about reasoning and logic. For one thing, here the concept of utterance doesn't seem particularly valuable, and you're looking exactly at what a sentence means, not what someone might mean by it. (Sentences connect to each other to form arguments, in a way reminiscent of words connecting to form sentences.)

That sounds a little dogmatic, because of course one of the things we do a lot of, and why I found the Comey testimony interesting, is reason about utterances, and quite often it's the difference between what was said and what was meant by what was said that's at issue. But here we have the utterance as object. The utterance of your reasoning about some utterances -- maybe you're into rhetoric there, or something.

Now that I think of it, maybe "rhetoric" is the traditional term for my pragmatics level. It's acquired some crust I wouldn't want to disturb though.

I think it does. 'Meaning', like many other words, is polysemous. As long as we can clarify what we mean to say, and where the areas of determinability and indeterminability lie, why should we be excessively anal about terminology? What would you say to the word 'significance' as a substitute? Perhaps I should have said 'it is semiotics all the way down' instead; would that make you happier?

Here's an example to clarify my view of the context principle.

I mentioned the logical constants several times. Let's suppose someone asks me, "What does 'and' mean?" I seriously have no idea what to say. Can you explain that without using the word? (I'm tempted to check and see how dictionaries handle this, but of course dictionaries have to rely on cycles of words to define each other mutually.)

You have to show someone how to use the logical constants. You train them. (The sequent calculus carries an echo of this in its introduction and elimination rules.) At some point they know how to use them, and understand how the logical constants modify the truth conditions of statements.

Frege's point is that once you know that, there's nothing left to know about their meaning. There's no meaning besides how they're used in sentences and how that use changes truth conditions. That use is their meaning.

Are we talking about natural languages, like English, or some sketch of an "ideal language" in a logician's notebook? I'll assume we're talking about natural language, or perhaps language in general.

This much seems obvious to me:
1. Compositionality has its limits. Phonemes and letters don't have standalone meanings. The same word or phrase can have various meanings. The same sentence can have various meanings. The same proposition can be expressed in various sentences composed of entirely different parts.

2. Context does a lot of the work for us in constraining all that variability. Relevant aspects of context may include the intentions of the speaker; the shared associations, habits, and beliefs of the speech-community; relations among speaker's intention, utterance, and perceptible features of the environment; and so on.

3. Conversations like this one will remain mired in confusion if we proceed by assuming that everyone has roughly the same idea of what "meaning" means in them.

If a Fregean decrees that "words have no meaning", he's not rejecting the assertions in ordinary English that "words have meaning" or "words are meaningful", nor is he affirming the assertions in ordinary English that "words are meaningless" or "words have no meaning". Rather he's constraining the use of the term "meaning" in the special context of his philosophical discourse.

Of course words have some sort of linguistic significance, of course they mean something, of course they have meaning.... Of course it's often the case that we can tell whether or not someone "knows what a word means". The Fregean should acknowledge all this when he's speaking ordinary English among ordinary-English speakers; it's his burden to find some way to paraphrase these commonplace insights into his own peculiar idiom, and to find some way to make his own peculiar insights intelligible in the common tongue.

In that light, I just don't see what the tension between "compositionality" and "context" is supposed to be. It seems clear to me that the meaning of an English utterance depends in part on the meaning of the English sentence uttered; in part on the speaker's intention in uttering that sentence; and often in part on other features of context.

The analyst is free to isolate the English sentence and consider its "meaning" in abstraction from any context of utterance. It seems clear this abstract "meaning" can be analyzed in terms of the arrangement of linguistically significant parts according to rules of sentence "composition". It seems just as clear that this abstract "meaning" of sentences is only one component of linguistic meaning, and that the results of such analysis will often seem incomplete or at odds with respect to the full-blooded meaning of utterances in context.

The abstract significance of the sentence thus isolated is nonetheless an important component of linguistic meaning, essential to natural language as we know it.