Andrew Radford - Linguistics An Introduction [Second Edition]

subject and an intransitive verb, is not appropriate (we are simplifying our syntactic assumptions by ignoring T and its projections – taking account of these would not affect the point under discussion). What we appear to need is a representation which enables us to make it explicit that in determining the truth conditions of

(420a), we have to consider a number of individuals in turn, checking whether each of them has the relevant property. We can achieve this by introducing into the representation an expression which, unlike a proper name, does not pick out a unique individual but instead can vary in the individuals it picks out. Such an expression is a variable, and the sort of representation we need for semantic purposes appears in (421):

(421)

(every sheep x)(x snores)

Here x is a variable, and (421) is read as ‘for every individual x which is a sheep, x snores’. If we now suppose that (420a) can be somehow linked to (421), the truth conditions of (420a) can be formulated as in (422): (422)

Every sheep snores [or (every sheep x)(x snores)] is true just in case every x which is a sheep has the property of snoring

In (421), every sheep x looks like some kind of DP and x snores is a clause with the variable x functioning as its subject and snores as its verbal predicate.

However, at this stage, we are not concerned with the details of this structure –

what is important is that, taking account of the Principle of Compositionality in

(402), it is appropriate for determining the truth conditions of (420a). By contrast, the superficial syntax of (420a) is inappropriate for this purpose. As we have noted, (420a) contains nothing beyond a subject DP and an intransitive verb; specifically, nothing corresponding to a variable appears in this structure, and thus it is not possible to see this structure as providing the appropriate basis for the operation of the Principle of Compositionality.

Sentence meanings and Logical Form

339

Now, (420b, c) will yield to this sort of informal treatment readily enough, to give us the representations in (423) and (424), respectively: (423)

(most sheep x)(x snores)

(424)

(no sheep x)(x snores)

And, each of (423) and (424) can be integrated into the statements of truth conditions in (425) and (426) – remember that the apparent triviality that we perceive in such examples is due to the fact that English is serving as both object language and metalanguage:

(425)

Most sheep snore or [(most sheep x)(x snores)] is true just in case most x which are sheep have the property of snoring

(426)

No sheep snores or [(no sheep x)(x snores)] is true just in case every x which is a sheep does not have the property of snoring

To summarise up to this point, and putting (420d) to one side for the moment, we are suggesting that the superficial syntactic form of the other sentences in (420)

is not appropriate for revealing their semantic properties.

For (414) Shirley snores and similar sentences, there is a straightforward relationship between syntactic form and the computation of truth conditions; in the syntax, there are two entities, the subject DP and the intransitive verb, which are matched directly by the individual picked out by the name and a property in the semantics. In (420a, b, c), however, we find a different situation: again, there is a subject DP – in each case a quantificational DP – and an intransitive verb, but, in these cases, there is no individual picked out by the former; instead, the semantics requires some means of considering a range of objects, and this is achieved by introducing a variable into the representation. Accordingly, for these sentences, the representations we need for semantic interpretation do not appear to be directly reflected in their syntactic structures. For Frege, who was concerned to develop a semantic account of quantification for use in logical inference, this was of no concern, and he was at liberty to stipulate whatever representations were necessary and to rely on his own skill in ensuring that the representations he stipulated were appropriate. For a linguist, however, confronted with trying to produce an explicit theory of human linguistic competence, such stipulation and reliance on the skill of a nineteenth-century German logician is not comfortable; at this point, we must acknowledge an unacceptable gap between what syntax

provides and what semantics needs in the case of sentences containing quantified noun phrases.

Covert movement and Logical Form

A resolution to the dilemma posed above is approached via (420d). An obvious point is that, as (420d) is an interrogative, it doesn’t make much sense to

340

senten ces

talk about its truth conditions, but putting this to one side, we can ask whether the representation we need for its semantic interpretation is more appropriately construed along the lines of (414) or similarly to (420a, b, c). Specifically, we can ask whether its interpretation involves reference to an individual picked out by the phrase which sheep or whether we need representations like those in (421), (423) and (424) to make this interpretation more transparent. The answer is obvious.

There is no which sheep being referred to in (420d), just as there is no no sheep being referred to in (420c), in contrast to the individual picked out by Shirley in

(414). We are therefore led to the representation for (420d) in (427): (427)

(which sheep x)(x snores)

We can immediately note that (427) captures something important about the interpretation of (420d); someone who understands (420d) knows that it asks for a search through a set of sheep looking for one (or more) which has the property of snoring. The fact that we have a variable in (427) taking as values individual sheep provides us with a device for conducting such a search.

But now, for this example at least, we can note an interesting and important correspondence between what semantics requires and what syntax supplies.

Recall our discussion of movement in section 21 and the observation made there that members of a class of operator expressions, including wh-phrases, move from the position they occupy as a result of merger operations to a clause-initial position, namely spec-CP. In the earlier discussion, the moved wh-phrases originated in complement position as in (341a) repeated as (428a) with the derivation in (428b):

(428)

a.

What languages can you speak?

b.

CP

C'

TP

T'

VP

DP C

D

T

V

DP

What languages

can

you

t

speak

t

(I)

(II)

Furthermore, it was argued in section 21 that this movement leaves behind a trace copy (t in 428b) of the moved operator expression in the argument position,

Sentence meanings and Logical Form

341

and that the appearance of a wh-operator in spec-CP is necessary if this clause is to be interpreted as interrogative. Now, (420d) is interrogative, and we can propose that which sheep moves in this example from its original position in spec-TP to the spec-CP position as indicated in (429):

(429)

CP

C'

TP

T'

DP DP

C

T

V

which sheep

ϕ

t

ϕ

snores

Obviously, in (429) the final position occupied by which sheep and its initial argument position are linked, and we can make this linking explicit by the notational device of co-indexing the moved item and its trace copy (up to now, we have used either bold or italic type to perform this function). If we do this, (429) can be roughly represented as in (430) (ignoring the covert C and T positions): (430)

(Which sheepi) (ti snores)