• Keine Ergebnisse gefunden

Interpretation of the phrase (ad 4)

Im Dokument NP-Arguments in NPs (Seite 103-108)

Phrase Structure Grammar

2.5 Rules, principles, constraints

3.1.4 Interpretation of the phrase (ad 4)

The interpretation of the phrase, that is its semantic information, is driven by a principle which has not been presented yet, but which was mentioned in the discussion in Section 3.1.2. This principle – called Semantics Principle (SemP) – is similar to the already known HFP, despite that the SemP applies to thesemantics part of the sign, i.e. the features and values under cont.

The semantics of a whole phrase is determined in particular7 by the semantics of the structural head daughter. Considering for instance the examples (1a) and (1b) above. Although, they are the combination of different linguistic signs, the main semantic contribution is in both situations made by Gewinn ‘win’. In both cases, we are talking about a ‘win’ which is specified in one case by an argument (i.e. what has been won in example (1a)) and in the other case by an adjunct (i.e. how it has been won in example (1b)).8 Since the information about the meaning of the phrase is not syntactic but semantic in nature, it is encoded under the cont attribute and its projection to the phrasal level is hence not covered by the HFP. Therefore, a counterpart of HFP in semantics is needed to ensure the projection of semantic information from the head of a structure to its mother node, or to the phrasal level. This task is accomplished by the SemP which is a constraint applying only to phrases with a head, i.e. only to phrases of type headed-structure.

For reasons to be illustrated in detail in Sections 3.2.3 and 3.3, the SemP does not apply in the same way to all of the subtypes of headed-structure; to be more specific: Phrases of type head-adjunct-structure and head-specifier-structure show a different behaviour with respect to the SemP. According to this restriction, the preliminary version of the SemP applies to the types: head-subject-structure and head-complement-structure. Since one constraint applies to both subtypes, they can be grouped under a supertype calledhead-argument-structure. This reflects the idea proposed here that subjects and complements are arguments, but specifiers are not. On the other hand, specifiers are not adjuncts either and share some

7See the discussion of the SemP with respect to adjunct and specifiers in the following sections.

8The details on the argument-adjunct distinction will be explained in Section 3.2.

properties with subjects and complements, for instance, they all are affected by the ValP. Thus, a reorganisation of the type hierarchy with further (negative) types (i.e.non-X-structure) as was proposed in Müller (2013a: 69–70) is been used here as well.

The typeheaded-structureis divided into two subtypeshead-non-adjunct-structure and head-non-argument-structure. Since adjuncts are clearly not arguments, the head-adjunct-structure is a subtype of the latter, and since specifiers are neither arguments nor adjuncts the head-specifier-structure is a subtype of both types head-non-adjunct-structure and head-non-argument-structure, i.e. a case of multi-ple inheritance. Due to this reorganisation of the type hierarchy, constraints can apply to subjects, complements, and specifiers, or to specifiers and adjuncts.9 With this modification, the type hierarchy is modelled as Figure 3.8 shows.

The SemP can thus be formulated by means of the typehead-argument-structure as in (3).10

(3) Semantic Principle (SemP) (1st preliminary version)

The cont value of a headed phrase of type head-argument-structure is structure-shared with the contvalue of the head daughter.

But there are some problems with the formulation of the SemP as given in (3).

Here, I will address only one difficulty of (3) with respect to agreement across sentences (cf. Müller, 1999: 31–34) and exemplify a second preliminary version of the SemP. Example (4) shows that if two sentences are combined and one referent – in this caseMaria – is picked up in the second sentence by a pronoun, then some kind of agreement must take place.11

9See Müller (2013a: 195) for a further development of the negative-types non-X-structure, and Van Eynde (2006: 158–160) for further similarities between modifiers and specifiers.

10For distinct formulations of the SemP according to different grades of accuracy in the analysis of phenomena, see Pollard and Sag (1987: 99, 104, 109, 110); Pollard and Sag (1994: 48, 56, 322-323); Richter (2000: 368); and Müller (2013a: 69–70, 78).

11The coindexing of Maria and the pronouns just indicates that the intended reading is one in whichMariaand the pronouns have the same referent.

sign word

. . . .

phrase

non-headed-structure

. . . .

headed-structure

head-

non-adjunct-structure

head-

argument-structure

head- subject-structure

head-

complement-structure head- specifier-structure

head-

non-argument-structure

head- adjunct-structure . . .

Figure 3.8: Type hierarchy of sign (preliminary)

(4) ( dass ) that

Mariai

Maria.3.sg.fem joggt, jogs

obwohl even though

siei

she.3.sg.f / * eri

he.3.sg.m krank

ill

ist.

is

‘[. . . ] Mariai jogs even though shei is ill.’

The pronoun used in example (4) to pick up the referent must agree with the expression Maria in person, number, and gender. The pronoun sie ‘she’ is there-fore grammatical, while er ‘he’ is not. The relevant features are encoded in the cont|ind value of Maria. Since the relevant phrase Maria joggt in example (4) is a phrase of type head-subject-structure, the verb joggt ‘jogs’ is licensed as the head daughter andMaria as the non-head daughter. According to the SemP in (3) – because a head-subject-structure is a subtype of head-argument-structure – the cont value of thehd-dtr is identical to the contvalue of the whole structure.

Hence, the ind value of Maria would be really deep embedded in the structure.

One way to get rid of this problem is to divide the amounts of structure-shared information into two; that is, to distinguish the structure-sharing of the indvalue and of the rels value (cf. (5a) and (5b)) as proposed in (5).

(5) Semantic Principle (SemP) (2nd preliminary version) For a headed phrase of type head-argument-structure:

a. Its cont|indvalue is structure-shared with the cont|indvalue of the head daughter and,

b. itscont|rels value is the concatenation of the rels lists of the head daughter and the non-head daughter.

I am going to exemplify the result of this modification taking a substructure of Figure 3.7, namely Nii [Gewinn der WM] ‘win of the World Championship’, given as as Figure 3.9 below.

The combination of the nominal headGewinnwith the NPder WM is licensed by the head-complement-structure which is a subtype of the head-argument-structure. Hence, the ind value of the hd-dtr N0ii, i.e. 1 , is identical – viz. structure-shared – with the indvalue of the resulting structure Nii by means of the SemP.

This reflects the fact that the result of the concatenation Nii has, for instance, the gend value masc, and not fem as the NPi has. Moreover, the rels value of the structure is the concatenation of therelsvalues of both daughters. Since the noun Gewinn is the nominalisation of a verb gewinnen ‘to win’, it has two elementary

Nii

of the World Championship Figure 3.9: Illustration of the Semantics Principle

predications (cf. Section 2.4). The nominalisation is the result state (win result) of an event (win). The “winning-event” has a theta-role theme which is fulfilled by the element which is the complement ofGewinn (cf. 4 in example (8)).

As an implicational constraint, the 2nd preliminary version of the SemP can be formalised, as shown in (6).12 With the reformulation made in (5), not only the rels value of the head daughter is being projected to (i.e. structure-shared with) the phrase, but also the one of the non-head daughter. That is to say, by means of the SemP the head projects to its phrase what kind of semantic object (indvalue) the whole phrase is, e.g. entity vs. event, and by the concatenation of the rels values of the hd-dtr and the nh-dtr, the semantic relation between head and non-head gets projected to the phrase as well.

(6) Semantic Principle (SemP) (2nd preliminary version)

head-argument-structure

⎢⎢

⎢⎢

⎢⎢

⎢⎢

⎢⎢

⎢⎢

synsem|loc|cont

⎢⎢

⎢⎣

ind 1 rels 23

⎥⎥

⎥⎦

hd-dtr|synsem|loc|cont

⎢⎢

⎢⎣

ind 1 rels 2

⎥⎥

⎥⎦ nh-dtr|synsem|loc|cont[rels 3]

⎥⎥

⎥⎥

⎥⎥

⎥⎥

⎥⎥

⎥⎥

Im Dokument NP-Arguments in NPs (Seite 103-108)