A first step in understanding various positions in the theory of meaning is to distinguish several questions. In particular, it is important to distinguish the following two:
Answers to both have gone under the name ‘theory of meaning’; but it’s important to see the difference between the two questions. Intuitively, the first question asks what the facts about meaning are, whereas the second question asks for an analysis of those facts -- a statement of what it is in virtue of which those facts obtain, of the foundations of facts about meaning. One good way to avoid confusing these questions is to avoid the ambiguous theory of meaning in favor of semantic theory (for an answer to the first question) and foundational theory of meaning (for an answer to the second question). This course will primarily be concerned with answers to the second question: with the foundational theory of meaning.
The distinction can be made clearer by an analogy with two distinct questions in ethics: (i) What are the moral obligations of agents? and (ii) In virtue of what do those moral obligations obtain? Suppose that you think that we have a moral obligation not to lie; this partial answer to question (i) would still leave open the answer to question (ii). One might think that this moral obligation obtains because lying contravenes the categorical imperative, because God commands us not to lie, etc. Just so, an answer to question (ii) is not by itself an answer to question (i). You can answer question (ii) by being a divine command theorist; but, absent a list of what the divine commands are, this does not tell us what moral obligations agents have.
Matters are similar with respect to our two questions about meaning. If a theorist says what he thinks the meaning of a given expression is, this will usually not tell us much about the foundations of meaning; and a view about the foundations of meaning will usually not have any immediate implications for semantic theory. However, there are some connections between the two sorts of theories, so below I’ve listed some of the main positions with respect to both semantic theories and foundational theories of meaning. In the third section, I mention some other questions about language and intentionality which will be relevant to the course.
On their face, claims that a certain linguistic expression means such-and-such are claims that a certain relation -- the meaning relation -- holds between expressions and a certain kind of entity, the meaning of that expression. (This view is called ‘Platonist’ because adherents typically identify the meaning of a sentence with an abstract object.) On this view of ‘meanings as entities’, which is arguably the orthodoxy in contemporary philosophy, the first question a semantic theory should answer is: what kinds of things are meanings? Here we can distinguish three leading answers.
On one view, the meanings of expressions are the same sorts of things that are the constituents of facts: objects and properties in the world. On some versions of this view, the meaning of a simple name is the object for which it stands, and the meaning of a simple predicate is a property. (Properties are usually thought of as being similar to universals, or Plato’s forms.)
One advantage of this view is that it makes meanings relatively unmysterious. We might already have metaphysical reasons for believing in universals; so Russellian theories of meaning don’t force us into any extra metaphysical commitments.
A disadvantage of this view is that it makes meanings relatively ‘coarse-grained.’ Suppose, for example, that a Russellian says that the meaning of a simple name is its referent. Then the Russellian must say that all names that have the same referent have the same meaning. But it is clearly possible for me to understand both ‘Superman can fly’ and ‘Clark Kent can fly’ while thinking that the former is true and the latter false; and this has led many people to think that co-referential names like ‘Superman’ and ‘Clark Kent’ must, contra this kind of Russellian, differ in meaning.
Fregeans offer a more fine-grained approach to meanings. On one version of the view (which is not exactly Frege’s) the meaning of a name is not an object, but a way of thinking about an object, and the meaning of a predicate is not a property, but a way of thinking about a property. Frege called these ‘ways of thinking about’ objects and properties senses.
Associated with this view is also a criterion for determining when two expressions have different meanings, or senses, often called ‘Frege’s criterion.’ According to this criterion, two sentences differ in meaning if and only if it is possible for a rational person who understands both to think that one of the sentences is true, and the other one false. (Two names differ in meaning if and only if it is possible for two sentences which differ only with respect to those names can differ in meaning; and so on for other kinds of linguistic expressions.)
An advantage of this view is that it respects the intuition that, e.g., coreferential proper names often differ in meaning.
A disadvantage of this view is the mysterious nature of the entities it posits as meanings. Frege thought of senses as part of a metaphysical ‘third realm’, apart from both concrete objects existing in space and time and mental impressions. One might be skeptical about the existence of such a third realm; and, even if one is willing to admit abstract objects in general, one might be skeptical about admitting a large class of abstract objects just to serve as the contents of linguistic expressions.
Yet a third ‘Platonist’ approach to semantics identifies the meaning of a sentence with the set of circumstances which would make it true. On one version of this view, a circumstance is a way the world might have been - or, for short, a possible world. Then we can think of the meaning of a sentence as the set of possible worlds which would make it true.
An advantage of this view is that it respects the intuitive link between the meaning of a sentence and its truth-conditions.
An important disadvantage of this view is it implies that if two sentences are true in the same possible worlds, then they have the same meaning. But intuitively it seems that two sentences can be necessarily true - i.e., true in every possible world - without having the same meaning. E.g., ‘2+2=4’ and ‘Arithmetic is incomplete.’
Donald Davidson, beginning with “Truth and Meaning” (1967), rejected the Platonist idea that facts about meaning are facts about the relations between sentences and a special sort of entity. However, he wanted to preserve the link between the meanings of sentences and the conditions under which they are true. He combined these two ideas in proposing the idea that a theory of meaning is a special sort of theory of truth. A theory of truth in the relevant (Tarskian) sense is a theory that entails, for each sentence of the language, a ‘T-sentence’ of the form
S is true iff T
where ‘S is replaced by the name of a sentence, and ‘T’ is replaced by a sentence which is true if and only if the sentence named by S is.
An advantage of this approach is that it has fewer metaphysical commitments than any of the above Platonist theories.
A disadvantage of this approach is that it is puzzling why a theory which entails only T-sentences, which seem not to say anything about meaning, should be counted as a theory of meaning. Davidsonians typically respond by saying that it should count as a theory of meaning because knowledge of such a theory would be sufficient for understanding the language. But it’s not clear that this is true; I can know that a sentence is true if and only if such-and-such without knowing what the sentence means.
A quite different alternative is due to Chomsky. We saw that Platonist semantic theories think of the meaning of a sentence as an object, external to the sentence, to which the sentence bears a certain relation. (Different theorists might, of course, disagree about the nature of this object.) The things which serve as meanings then determine a set of truth-conditions for the sentence. Davidsonian semantic theories try to preserve the link to truth-conditions while giving up the Platonist understanding of meanings as entities. Chomskians, or internalists about meaning, give up both ideas: meanings are not entities distinct from linguistic expressions, and do not determine a set of truth-conditions. Rather (on some such views) the meaning of an expression is a property of that expression determined by its role in the language faculty of the user of the expression.
An advantage of this view is that it integrates the study of semantics with Chomskian approaches to syntax and language acquisition.
One disadvantage of this view is that it takes on an extra explanatory burden. Even if internalists claim that the meaning of a sentence does not determine truth-conditions, they typically concede that we can use sentences to say things about the world which can be true or false. Chomskians then distinguish (plausibly) a semantic theory from a theory of what we can use sentences to say, and claim that the semantic properties of a sentence constrain but do not determine what a sentence can be used to say. But they owe an account of how purely internalist properties of sentences can constrain facts about what sentences are used to say.
Opposed to all three of these approaches are theorists who say that, in some sense to be explicated, there are no facts about meaning. The skeptic Kripke discussed in his interpretation of Wittgenstein has such a view; on some interpretations, Wittgenstein himself held this sort of view.
There are at least five major approaches to the foundational theory of meaning: five answers to the question, What is it for an expression to have such-and-such meaning?
(There is also an important strand of thought which holds that facts about meaning are primitive, in the sense that there is no story about what it is for an expression to have a given meaning. This is less a foundational theory of meaning than a rejection of the need for such a theory.)
One intuitively plausible idea about the foundations of meaning is that the meanings of sentences are inherited from the thoughts of users of those sentences. On this view, mental representation is more fundamental than linguistic representation.
There are different ways of spelling out this idea. One, associated with Grice, claims that the meanings of sentences are determined by the contents of the communicative intentions of users of those sentences. Very roughly, the idea is that the meaning of a sentence is determined by what agents typically intend their audience to believe when they use the sentence. Another idea, associated with (among others) David Lewis, is that the meaning of a sentence is fixed by the beliefs with which the sentence is correlated. Perhaps, for example, the meaning of a sentence is the same as the content of the belief which most people would acquire upon hearing the sentence. Both of these approaches, in effect, treat linguistic meaning as a kind of derived intentionality, and treat propositional thought as a more basic form of intentionality. This is probably the most widely held view in contemporary philosophy.
A different sort of approach, identified with Davidson, has it that the meanings of sentences are determined by what an ideal interpreter would assign as the meanings of those sentences.
Different versions of this view are possible depending on what rules one takes to govern the ideal interpreter. On Davidson’s version of this view, the ideal interpreter (roughly) interprets agents so as to maximize the truth of their beliefs and utterances. In this sense, a Davidsonian interpretational theory takes thought and language to be interdefined. This is a point of contrast with mentalist theories, which take linguistic meaning to be defined in terms of the properties of various kinds of thoughts.
Causal theories of meaning hold that the meaning of an expression is fixed neither by the thoughts of agents nor by what an ideal interpreter would say, but rather by the causes of utterances. For example, one might hold that the meaning of a predicate is whatever property is such that instances of that property typically cause uses of the predicate.
There’s a sense in which causal theories of meaning go most naturally with Russellian semantic theories, since it is hard to see how Fregean senses or sets of possible worlds could cause utterances of expressions.
A different sort of view, often associated with Wittgenstein, is that the meaning of an expression is determined by its use. So put, this view is massively under-specified. Is it just actual uses of sentences which are relevant, or also dispositions to use sentences in various ways? Community-wide uses or individual uses? And what properties of uses are relevant? We’ll be discussing a number of different versions of this kind of view.
On internalist theories, linguistic expressions are, first and foremost, representations in the brain. Internalists say that the meanings of these representations are fixed by their relations to other parts of the language faculty. This view is more closely connected with semantic theory than the other sorts of foundational theory listed above. Recall that internalists about semantics hold that the meaning of a word is a property determined by some aspect of its role in the language faculty. This view obviously fits very neatly with an internalist response to foundational questions about meaning. (Conversely, non-internalist semantic theories seem to present problems for internalist foundational theories.)
Internalists do not have to deny that spoken words, for example, are not linguistic expressions, or lack meaning. Rather, they can say that these words acquire their semantic properties from the internal representations with which they are associated, which in turn get their semantic properties from their roles within the language faculty.
One important question about the foundations of language is whether languages are primarily social things shared among groups, or primarily ‘idiolects’ particular to individuals. Some of the foundational theories above carry commitments on this point. For example, mentalists who claim that the meaning of a sentence is fixed by what beliefs are correlated with the sentence in a community are clearly thinking of public languages as primary. Internalists who claim that expressions are fundamentally representations in the brain are clearly thinking of idiolects as primary. Interpretational, causal, and use theories can be stated to be compatible with either view of language; but they might be more or less plausible depending on which view is adopted.
We apparently have the ability to represent the world in thought and in language. Suppose for the moment that we regard languages as social things shared among groups of people. Then there are social facts about representation, and individual facts. A fundamental question is then how social facts about linguistic representation relate to individual facts about mental representation. Are social facts determined by more fundamental facts about the thoughts of individuals? Or do individuals acquire their ability to have certain kinds of thought from social systems of representation?
We can see mentalists as opting for the first course, some use theorists (including, perhaps, Wittgenstein) as opting for the second course, and internalists as rejecting the presupposition that linguistic representation is social.