Andrew Radford - Linguistics An Introduction [Second Edition]

ϕ

DP

T'

the paediatrician

T

VP

–d

V

DP

examine

the baby

Here, we can see that the paediatrician occupies a specific position in the

structure, the specifier of T, and the baby occupies the position of complement of V. What we might suggest, then, is that the thematic roles Agent and Affected Object, which will be included in the structure as part of the lexical representation for the verb examine, are explicitly related to the two positions in question via what are often referred to as Linking Rules. Thus, part of what is involved in taking a syntactic representation like (413) and converting it to a semantic representation or Logical Form for a sentence involves the application of Linking Rules, which will explicitly assign thematic roles to the expressions occupying specific positions in the structure. Obviously, the example we have described here is extremely simple, and it is easy enough to begin to pose difficult questions for the approach, many of which have been pursued in the research of the last twenty years (exercises 1, 2 and 3).

We now turn to a rather different aspect of the semantic interpretation of sentences, and in order to pursue this matter, it will be necessary firstly to say something about one way in which philosophers have studied the meanings of sentences.

336

senten ces

A philosophical diversion

Consider the sentence in (414) and suppose, for the sake of argument, that the name Shirley is the name of a specific sheep:

(414)

Shirley snores

A view adopted by many philosophers and linguists is that at least part of what is involved in understanding a sentence in a language (i.e. grasping its interpretation) is knowing what the world would be like if the sentence were true; to know this is to know the truth conditions of the sentence. Note that knowing the truth

conditions of a sentence does not require that we know that the sentence is or is not true; to know this latter for every sentence you understand would be to

approach omniscience, and it would be absurd for linguistics to claim that knowledge of a language has (near-)omniscience as a consequence.

You can persuade yourself that the position outlined in the previous paragraph is plausible by considering a small experiment that you might undertake. Suppose you take a picture and construct some simple English sentences which are true or false of the picture; then you present the sentences and the picture to someone in whose linguistic competence you are interested, asking them to respond with

‘true’ or ‘false’ to each of the sentences. If their responses were incorrect in some cases, you would probably conclude that they did not understand those particular sentences; if their responses appeared to be random across the set of sentences, you would probably conclude that they did not understand English at all –

imagine the responses you would get from a monolingual French speaker who

is told (in French) to respond with vrai (‘true’) or faux (‘false’) to a set of English sentences.

At least part of what (414) means, then, can be identified with its truth conditions. What might these conditions look like? Well, Shirley is a particular type of DP (with a null determiner), a proper name, and, we might suppose for simplicity, that it names a unique individual, the sheep called Shirley. The verb snores names a property, the property of snoring. Then, we might state the truth conditions for

(414) as in (415):

(415)

The sentence Shirley snores is true just in case the individual named by Shirley has the property of snoring

At this point, you may feel that while (415) is itself true, it is pretty unhelpful, since what it says is so trivial. But this reaction, while understandable, is misplaced and is due to the fact that in (415) we are using English to talk about English – more technically, we are using English as a metalanguage to talk about English as an object language. Obviously, if we are going to present the truth conditions for a sentence, we are going to have to use some language or other to do this. Readers of this book understand English, so our metalanguage is English throughout, but now suppose that we want to consider the truth conditions for the French sentence in (416):

Sentence meanings and Logical Form

337

(416)

Delphine ronfle

‘Delphine snores’

And suppose, again for simplicity, that the DP Delphine names a unique individual. Using English as our metalanguage, the truth conditions for (416) appear in (417):

(417)

The sentence Delphine ronfle is true just in case the individual named by

Delphine has the property of snoring

Now, if you don’t know French, but you do understand English, (417) will tell you something about the interpretation of (416); the reason you feel that (415) tells you nothing about the interpretation of (414) is entirely due to the fact that (415) uses English to tell you something about English, a language you understand.

It is easy now to generalise on the basis of additional examples of sentences consisting of a proper name and an intransitive verb that we might care to

construct. Some such sentences appear in (418) and a generalisation is formulated in (419):

(418) a.

Smythe smokes

b.

Jones jogs

c.

Stevens stammers

(419)

For any sentence consisting of a DP α followed by an intransitive verb β, the sentence is true just in case the individual named by α has the property named by β.

Note how (419) begins to acknowledge the Principle of Compositionality in (402), by stating how the interpretation of a sentence (its truth conditions) is determined by the semantic properties of its component words (names refer to individuals and intransitive verbs to properties) and the sentence’s syntax (the DP

precedes the intransitive verb). Obviously, we have deliberately chosen a very simple type of sentence, and the only aspects of syntax to which we have referred are the categorial status of the constituents and their order. However, this is sufficient to enable us to contrast the sentences in (414) and (418) with those in (420):

(420) a.

Every sheep snores

b.

Most sheep snore

c.

No sheep snores

d.

Which sheep snores?

Take (420a); as every sheep is a DP, consisting of the D every and its complement N sheep, and snores is an intransitive verb, its syntactic structure fits the description in (419), but if we try to apply (419) to formulate the truth conditions of (420a), we run into a major problem. This problem concerns the DP in subject position, every sheep. The question (419) raises is that of what individual is named by every sheep? But it is not sensible to ask this question of this expression.

338

senten ces

Arguably, it is even less sensible to ask it of most sheep in (420b), and just plain nonsense to ask it of no sheep in (420c) and which sheep in (420d). These expressions, while evidently DPs, do not name individuals in the straightforward way that proper names do, and it appears that (419) is simply not applicable to sentences containing such phrases.

The problem we have arrived at here was already appreciated at the end of

the nineteenth century by the German philosopher Gottlob Frege and his British contemporary Bertrand Russell. The solution to it that they developed can be sketched by talking informally about the truth conditions for (420a). We have seen that we cannot formulate these truth conditions in terms of an individual named by every sheep which has the property of snoring. Instead, what we need to do is examine each individual sheep (none of which is every sheep) in turn, checking whether it has the property of snoring. If the answer is ‘Yes’ for every sheep, the sentence is true. But this seems to require that from a semantic perspective, the simple syntactic representation of (420a), whereby it contains just a DP