Again I’m quoting Paul Ziff’s Semantic Analysis:
Consequently, if my contention about meaning is correct, then the first ‘do’ in ‘Please do not do it!’, unlike the second ‘do’, does not have meaning. This is testified to by the fact that generally the first ‘do’ in ‘Please do not do it!’, unlike the second ‘do’, will not admit of being stressed. Thus ‘Please do not do it!’ unlike ‘Please do not do it!’ is somewhat odd. Again, notice that the same is true of ‘to’ and ‘through’ in ‘I want to go through Istanbul.’. There is nothing odd about ‘I want to go through Istanbul.’ but ‘I want to to go through Istanbul.’ is generally quite odd. And this should not be surprising: if an element does not have meaning in an utterance, stressing the element is not likely to be, and indeed can hardly be, significant.
(There is a case, however, in which the ‘to’ in question will bear a stress. If I say ‘I want to go through Istanbul.’ and someone says ‘You want not to through Intanbul?’, I may reply ‘I want to go through Istanbul.’. An explanation of this is not hard to find. If I say ‘I want him to go.’ and someone says ‘You want them to go?’, I may reply ‘I want him to go.’, stressing the word after the verb for that was the point at which the utterance was misunderstood. But if I say ‘I want to go through Istanbul.’ and someone says ‘You want not to go through Istanbul.’, the confusion is owning to the insertion of ‘not’ after the verb. Thus in reply one is likely to stress whatever occurs over the segment immediately after the verb. Thus not ‘to’ but the stress it bears is significant in ‘I want to go through Istanbul.’: the stress contrasts with ‘not’ in the previous sentence.)
Meaning and meaningful words in sentences
Why do I quote this passage? Because I sometimes suggest this thesis in discussions:
1. A sentence is meaningful ⇔ Every word in that sentence is meaningful.1
This might seem obvious to some and it seems interesting to me. There is an, perhaps, obvious type of possible counter-example too. Here are a couple:
2a. kjjd is not meaningful.
2b. The word “kjjd” is not meaningful.
(2a) appears to be a counter-example to (1) since there is a word2 in it that is not meaningful. However, (2a) is an unclear sentence and perhaps grammatically incorrect.3 A more refined version is (2b) where it is clear that the sentence is about some word. There are a couple of solutions or explanations that spring to my mind about this.
One, one could see “the word “kjjd”” as a noun phrase that refers to the word “kjjd”. This seems unproblematic to me.
Two, one could try to limit (1) to some particular subset of sentences. One idea is to exclude sentences that are about words or phrases (meta-language). Though this seems excessive to me.
Three, one could exclude words that start and end with quotation marks (“) or whatever character is used to mark words or phrases. (Some people, like Ziff above, use apostrophes (‘).)
I favor the noun phrase theory or some similar theory. If that theory is true, then sentences like (2a) are not a problem for my thesis, that is, (1).
Meaningless words in meaningful sentences without metalanguage
In the quoted paragraph Ziff argues that some words in some meaningful sentences are not meaningful. His two examples are:
3a. Please do not do it!
3b. I want to go through Istanbul.
I think that it is uncontroversial whether these sentences are meaningful, they clearly are.4
Ziff thinks that:
4. A word in a sentence does not admit of being stressed without it being odd ⇒ That word does not have meaning.
This seems somewhat plausible and it is a problem for my thesis, that is, (1). Since (1) implies that all words in (3a) and in (3b) are meaningful but (4) implies that there is at least one word in (3a) and in (3b) that is not meaningful. How might one resolve this? Obviously one can simply deny that Ziff’s claim is true though it does seem rather intuitive to me, and I guess to many other people too.
Tokens and types
One might try to fix the problem by introducing the token-type distinction.5 Is (1) about types or tokens?:
1a. A sentence is meaningful ⇔ Every word token in that sentence is meaningful.
1b. A sentence is meaningful ⇔ Every word type in that sentence is meaningful.
Is (4) about tokens or types?:
4a. A word in a sentence does not admit of being stressed without it being odd ⇒ That word token does not have meaning.
4b. A word in a sentence does not admit of being stressed without it being odd ⇒ That word type does not have meaning.
The relationships between (1)’s and (4)’s are less clear. Let’s examine them in turn.
One, (1a) and (4a)
This appears to be the same situation as before.
Two, (1b) and (4a)
(1b) seems true to me but it is rather unclear what it means to say that a word type is meaningful. They do not seem inconsistent; The word type “to” is meaningful in (3b) and there is according to (4a) both a meaningful and a meaningless word token of “to” in that sentence. It is curious that some type can be meaningful yet tokens can be both meaningful and meaningless. (In the same language of course.)
Three, (1a) and (4b)
(4b) is false. Consider examples similar to the (3)’s, (4b) materially implies that the word type “to” and the word type “do” is both meaningful and meaningful. Contradiction.
Four, (1b) and (4b)
This is even worse than the case above. (4b) is false for the same reason as above, and (1b) materially implies not-(4b).
The type-token distinction did not help much, even though it clarified some things. (4b) is to be avoided, and (1a) and (1b) are interesting and problematic.
An idea is to reject (1) but accept some similar thesis:
1c. A sentence is meaningful ⇔ Every word token in that sentence is meaningful or is part of a meaningful phrase token in that sentence.
This seems more plausible than the other (1)’s so far to me. It also seems consistent with Ziff’s examples since the meaningless “to” tokens are part of a meaningful phrase token “want to”.
It also avoids the question of what it means to say that a type is meaningful.
Stress and word-parts
Notice that in the above paragraph that it is not odd to stress parts of words. (I stressed “less” and “ful”. “Less” functions as a logical negation in this case and many others.) Is this an indication that word parts are sometimes meaningful too? It doesn’t follow from (4)’s but if we created general principle out of (4):
4c. A language part is able to be stressed with it being odd. ⇒ That language part is meaningful.
(4c) materially implies that word parts (morphemes) are sometimes meaningful too.
And yet, there are still other counter-examples (1c). Consider this famous sentence:
5. Colorless green ideas sleep furiously.6
(5) is clearly a counter-example to (1c) since all the words in this sentence are meaningful, and yet the sentence itself is meaningless. Perhaps another thesis similar to (1) is needed:
1d. A sentence is meaningful ⇔ Every word in that sentence is meaningful or is part of a meaningful phrase in that sentence, and all words are meaningful in the relation they are stand in or are part of a phrase that is in a meaningful relation.
This seems to effectively deal with sentences similar to (5).
1“⇔” means is logically with.
2I use the word “word “ here in a less strict sense. It is sometimes defined like “(linguistics) A distinct unit of language (sounds in speech or written letters) with a particular meaning, composed of one or more morphemes, and also of one or more phonemes that determine its sound pattern.” By “word” here I mean something like a string of latin characters (without spaces). The strict definition above is taken from Wiktionary. http://en.wiktionary.org/wiki/word#Noun
3It seems unusually hard to judge whether it is grammatically incorrect or not in this case.
4Though it is curious how to best establish that a specific sentence is meaningful or meaningless in a specific language. I suppose that if the vast majority of the native speakers of language L understands sentence S, then S is meaningful in L. But there are problems with this. I will not discuss them in this essay.
5See Wikipedia for an explanation. http://en.wikipedia.org/wiki/Type-token_distinction but see SEP’s article on it for a more thorough explanation http://plato.stanford.edu/entries/types-tokens/#WhaDis.