Philosophers of language and linguists have attempted to classify the aspects of meaning that an utterance may have, and have distinguished between the semantics of the sentence uttered and the pragmatics of the utterance. Conventionally, the semantics are to do with the truth or satisfaction conditions of the sentence, ignoring those aspects of the utterance meaning that are influenced by the context of its use or its intended purpose. The pragmatics are everything to do with the context and the intentions of the speaker and addressee, and to a large extent this category is used to sweep up all the difficult aspects of meaning that researchers wish to postpone considering. This chapter deals mostly with those aspects of meaning that are regarded as semantics, whereas in Chapter 10 we will consider more pragmatic aspects of meaning.
Those people who argue for the semantics-pragmatics distinction tend to claim that the truth-conditional, context-independent component of the meaning of an utterance is an important component of the utterance's full meaning, even if it is not the whole story. Thus, if I am sitting in the same room as you and you are next to an open window, I might say:
It is cold in here.
as a kind of polite request for you to close the window. Of course, the truth conditions of the sentence are something quite different - they involve a claim about the world. On the other hand, it could be argued that, to respond appropriately to this utterance, one strategy would be to grasp the truth conditions and then subsequently work out for what purpose I might wish to call your attention to this particular fact about the world. This is the standard model of the semantics-pragmatics relation at the present time, and the one that we shall assume.
How can we sensibly organize the process of semantic interpretation; that is, the derivation of the meaning of a sentence? The answers that people have found to this question are various, but one principle attributed to the philosopher Frege stands out in just about every approach that has been made. It can be informally stated as:
The meaning of the whole is a function of the meanings of the parts.
According to this principle, known as the principle of compositionality, the meaning of a sentence can be expressed in terms of the meanings of the phrases within it. The meanings of these in turn depend on the meanings of the subphrases within them. And so on, until we are down to the meanings of individual words or even the meanings of the morphemes that make up the words.
Adopting Frege's principle does not make the extraction of the meaning a trivial process. It merely gives us a framework for organizing our ideas and decomposing the problem. In particular, we must resolve the following questions:
What are the appropriate subphrases to consider when we want to
obtain the meaning of a phrase?
Just how does the meaning of a particular phrase depend on that of
its subphrases?
What are the meanings of the minimal units (the phrases, words or
morphemes that do not themselves subdivide into smaller meaningful units)?
What kinds of things should meanings be anyway?
Are they symbols in the machine?
Are they things in the world?
Are they some kind of relationship between the two?
As for what the appropriate subphrases of a phrase are, we might hope that these would be precisely those that are given by the syntactic structure. So, for instance, when we have a rule in the grammar that says:
Rule S -> NP VP.
then, as well as saying something about what forms a sentence can take - a sentence can be a noun phrase followed by a verb phrase - it might say something about how the meaning is to be extracted - the meaning of the sentence is some function of the meaning of the noun phrase and the meaning of the verb phrase. Of course, it might turn out to be asking too much for the same break-down of a phrase into subphrases to yield both syntactic and semantic generalizations.
We can, nevertheless, hope that there is some consistent relation between syntax and semantics in natural languages, and that a semantically appropriate grammar also captures syntactic generalizations.
Assuming that a syntactic analysis divides up phrases into semantically appropriate subphrases, it remains for us to specify just how the meaning of a complex expression (for example, the truth conditions of a sentence) depends on the meanings of the parts (for example, the noun phrase meaning and the verb phrase meaning). It is desirable to be able to cope with every possible way that complex phrases can be made up of immediate subphrases. This leads to an approach where syntactic and semantic rules are paired: corresponding to each syntactic rule that says how a phrase can be constituted, there is a semantic rule that says how the meaning of such a phrase is composed out of the meanings of the parts. Many approaches to semantic interpretation adopt this strategy, which is known as the rule-to-rule hypothesis.
There are various ways we can implement semantic rules computationally. In one method, direct use is made of the phrase structure tree of the natural language utterance in obtaining the meaning. Imagine we have the following rule in a phrase structure grammar:
Rule A -> B C D.
A semantic rule attached to this rule specifies how the meanings of phrases of types B, C and D, when these phrases appear in this order, can be combined to give the meaning of a phrase of type A. This is exactly equivalent to specifying how the meaning of a phrase of type A can be constructed, given that its parse tree is of the form shown in Figure 8.1.
Thus, on this view, semantic rules can be seen as mappings from local phrase structure trees to meanings. We pursue this idea further in the sections that follow.
Send us a comment.