• Keine Ergebnisse gefunden

5.5 Wh-phrases

5.5.2 Scope marking

5.5 Wh-phrases 127 Whether we delete the reset now or later does not play a role because in either case it will not interfere with the scope taking of the wh-operator. The deriva-tion proceeds constructing the matrix CP. We skip these steps (and again ignore do-support). The resulting expression corresponding to the whole sentence is

which fightwh, did someone think every brave man feared•wh::CP, . . . with the following semantic expression of type (ttt)wh:qwh:t, where the denotation ofsomeoneis abbreviated asξk.Eperson:

hh((thinkh ∀x.(manx)∧(bravex)⇒((fear ξwhk.Efight)x)i ξk.Eperson)iiwh

On the syntactic side, the configuration triggers remerge, which checks the features, concatenates the form at the edge with the form of the nucleus and eventually yields the final stringwhich fight did someone think every brave man feared. On the semantic side, we still have two control transfers to execute.

First observe that the order in which we do this is not free. Suppose we first applied the reduction rule to the ξwh-expression corresponding to which fight.

It captures a context of type t and changes its type to q. If we then wanted to apply the reduction rule to the ξ-expression corresponding tosomeone, we would fail, because it requires to capture a context of typet, however finds only one of typeq. Therefore, the reduction rule first has to apply to the denotation of someone(which does not change the result type but returns an expression of type t) and only after that to the denotation of which fight. The result of applying the reduction rule first to the denotation ofsomeoneis the following (where I abbreviate (bravex)∧(manx) as (braveMan x)):

hh ∃y.(persony)∧((thinkh ∀x.(braveManx)⇒(fear ξwhk.Efight)x)i)y)iiwh

The remaining control transfer then yields the final result:

Wz.(fightz)∧ ∃y.(persony)∧((think (∀x.(braveMan x)⇒((fearx)z))) y) In prose it says: For which fight is it the case that there is someone who thought that every brave man feared it? (Ignoring tense, that is.)

This concludes the general mechanism for scope construal of displaced wh-phrases. Note that it works independently of whether the wh-phrase is dis-placed overtly or covertly. As long as it checks its wh-feature with a head carrying a corresponding probe feature, its scope will be construed in that do-main. Let us now look at how this mechanism naturally extends to the scope of wh-phrases in scope marking languages, although no displacement is involved.

128 A semantic procedure for scope construal 5 (5.63) Japanese (Tsai [117])

John-wa [[dare-o aisiteiru] onna-o] nagutta no John-topwho-accloves woman-acchit q

‘Who is the personxsuch that John hit the woman who lovesx?’

From the island insensitivity we concluded that those wh-expressions are not displaced, not even covertly. That is, they do not check features with a clausal head and, as a consequence, have to be assumed to not carry a wh-feature.

(If they did carry a wh-feature, they would automatically be split and carried along until the feature can be checked.)

Recall the observation from Chapter 2 that the scope of a Japanese in situ wh-phrase is determined by the occurrence of a question particle, here ka. The examples we saw were the following: In (5.64a), the particle occurs in the embedded clause, the wh-phrase thus takes scope over the embedded clause only. In (5.64b), on the other hand, the particle occurs in the matrix clause, the wh-phrase thus takes scope over the whole sentence. Moreover, in the presence of two question particles (in the embedded as well as the matrix clause) as in (5.64c), each of them can be taken to determine the scope, the sentence is thus ambiguous between a narrow and a wide scope reading of the wh-phrase.

(5.64) Japanese (Boˇskovi´c [122], Cresti [29])

a. Peter-wa [anata-ga dare-o mita-ka]tazuneta.

Peter-topyou-nom who-accsaw-q asked

‘Peter asked whom you saw.’

b. Kimi-wa[dare-ga kai-ta hon-o yomi-masi-ta]-ka?

you-top who-nomwrote book-accread q

‘Which personxis such that you read a book that xwrote?’

c. Hikaru-wa [Akira-ga dare-o hometa-ka]siri-tagatte-imasu-ka?

Hikaru-top Akira-nomwho-accpraised-q know-want-be-q

‘Does Hikaru want to know whom Akira praised?’

‘Which personxis such that Hikaru wants to know whether Akira praisedx?’

Let us look at how to specify lexical entries for the wh-expressions and the question particle in order to derive the correct scope effects.

Concerning the wh-phrasedare(‘who’), we need to decide on the mode and the flavor of the involved shift operator. The flavor will be assumed to be wh, because the captured context depends not so much on the presence of a clause boundary but rather on its being question marked. As to the mode, Tanaka [116] proposes that a Japanese wh-phrase must take scope according to the closest question marker. (He assumes this to be an LF principle but that does not matter here.) We would therefore assume the mode to be strong or weak, but not free. Let us assume a strong mode. Here is a lexical entry that takes

5.5 Wh-phrases 129 this into account and is, in fact, completely parallel to the entries for English wh-expressions:

(dare:: NP, ξwhk.Wx.(personx)∧(k x)::eqt)

Now, what provides the delimiter for the scope of dare? In the previous subsection, it was assumed that the probe feature•whof the clausal head plays this role. Here we do not have such a feature. However, the clausal head is the question particleka(according to Takahashi [115]). Since it determines the scope of the wh-expression, it seems straightforward to connect the delimiter to its denotation. In specifying that denotation, I follow Takahashi in takingka to be ambiguous between a yes/no-question marker and a wh-question marker.

He argues for this ambiguity on the basis of different requirements they have:

both have to be governed by a tense feature, and the wh-particle ka has to be additionally governed by a politeness feature, whereas the yes/no-particle ka does not. From this point of view, the two readings of (5.64c) are due to the different nature of the two occurrences of ka. If the embedded ka is the wh-particle and the matrix one is the yes/no-particle, the first, direct question reading comes about. If the distribution is vice versa, i.e. the embedded ka is the yes/no-particle and the matrix one is the wh-particle, then the second, wh-question reading comes about.

The tense requirement of both can be captured by assuming an additional projection TP that constitutes the level at which tense features are checked.

The politeness requirement, on the other hand, I will omit. Then the following two lexical entries can be specified forka: (5.65a) for the wh-particle, introduc-ing awh-flavored delimiter, and (5.65b) for the yes/no-particle, introducing a question operator ? that turns a declarative clause into an interrogative one (the exact modeltheoretic interpretation is again of no concern here).

(5.65) a. (ka:: TP< →CP, λp.hpiwh::t→t) b. (ka:: TP< →CP, λp.?p::t→q)

I will not give any details of the derivation of (5.64c) because it proceeds pretty much like all derivations we saw in this chapter. But I will specify the lexical entries of the expressions involved (see the table below) and show the result they lead to. I ignore case, simplify the denotation of siri-tagatte-imasu, and abbreviate the typese→(e→t) andq→(e→t) ase(et) andq(et).

Form Meaning

Hikaru-wa :: NP hikaru ::e

Akira-ga :: NP akira ::e

dare-o :: NP ξwhfreek.Wx.(personx)∧(k x)::eqt hometa :: NP<→(NP<→VP)admire ::e(et) siri-tagatte-imasu:: CP<→(NP<→VP) wantToKnow ::q(et)

130 A semantic procedure for scope construal 5

Note that wantToKnow is assumed to expect a question denotation. This is mainly because it is applied to a question denotation in the considered example.

If we wanted to be more general, we would assume it to have a polymorphic typer→(e→t), wherer can be instantiated by any result type, in our case t or q. This is what we will in fact do for the predicate know in the next subsection.

The semantic expression that results from generating (5.64c) by using the wh-particle (5.65a) in the embedded clause and the yes/no-particle (5.65b) in the matrix clause is the following expression of typeqtq:

?((wantToKnowh((admire ξwhk.Wx.(personx)∧(k x))akira)iwh)hikaru) Reducing it yields an expression corresponding to the direct question reading of (5.64c):

?((wantToKnow Wx.(personx)∧((admire x)akira))hikaru) Generating (5.64c) by using the yes/no-particle in the embedded clause and the wh-particle in the matrix clause, on the other hand, results in the following semantic expression of typetqt:

h((wantToKnow?((admire ξwhk.Wx.(personx)∧(k x))akira))hikaru)iwh

Reducing it yields an expression corresponding to the wh-question reading of (5.64c):

Wx.(personx)∧((wantToKnow?((admirex)akira))hikaru)

So we can treat in situ wh-phrases in scope marking languages along the same lines as wh-phrases in languages invoking displacement, the only differ-ence being that the delimiter is not introduced by a probe feature but by a specific lexical item, usually a question particle.