• No results found

predicate-based vs lexical entry driven approaches: a summary

1.3 Mapping systems '

1.3.3 predicate-based vs lexical entry driven approaches: a summary

All approaches, whether predicate based or lexical entry driven, acknowledge the tight correlation between some aspects o f a verb's meaning and certain syntactic structures. The main debate between the two types of approaches is whether information concerning the linking o f arguments is listed in both the syntax or the lexicon, or only in the latter. I made a conceptual point in favour of the latter approach, arguing that since structure-meaning correlations in the form o f linking rules are needed anyway, any more specifications in the lexicon should be redundant. In the rest o f this chapter I will bring some empirical motivation in favour of a predicate-based approach, based on cases of verb alternations. I will start, in section 1.4, with some better-known alternations and will discuss the challenge they pose for the theory of the syntax-lexicon interface, as well as some o f the solutions suggested for them in the framework o f lexical-entry driven approaches. Then in section 1.5 I present data from

' ^ See Manzini and Roussou (1997) and Manzini and Savoia (1998) for a different formulation o f Borer's idea: thematic roles are formulaed in terms o f aspectual features, originator, measurer and delimiter. Clitic arguments are directly merged in their surface positions in a clitic shell above the VP, while the head attract the appropriate feature from the VP, thus checking it, and getting the interpretation o f the argument.

Pinker (1994) distinguishes between root meaning and fram e meaning, the first being what we intuitively think o f as the contents o f the verb (e.g., that boil is about a hot liquid releasing bubbles), while that latter includes those aspects o f meaning shared by all the roots appearing in a given syntactic frame (e.g., by I boiled the water, I opened the door, and I melted the ice). Other linguists distinguish between idiosyncratic and structural meaning or between semantic content and semantic structure (LRH

intransitive verbs which are unstable in their syntactic characterisation: unaccusative verbs which can appear as unergatives in certain syntactic contexts, and unergative verbs which can be unaccusativized in certain syntactic contexts. Such cases, I argue, are very hard to account for in a lexical-entry based approach, whereas a predicate-based approach can accommodate them with no difficulty.

1.4 Verb alternations and their importance for the interface

Verb alternations are cases in which a single verb appears in more than one syntactic structure. Such cases include the dative alternation, the transitive-inchoative alternation, the locative alternation, etc. I will briefly illustrate such alternations here and discuss their specific properties.

The transitive/inchoative alternation:

Many verbs in many languages undergo this alternation (in some languages, like Italian and Hebrew, the two alternants are associated with different verb morphology). The transitive alternant is the causative, including both an originator, which is lexicalized as the subject, and a measurer, which is lexicalized as the object, while the intransitive one includes only a measurer, which is lexicalized as a subject (for some recent references see Pinker 1989, LRH 1995, Reinhart 1996, van Hout 1996, HK 1997a):

(38) a. The army sank the ship (originator and measurer) b. The ship sank (measurer)

(39) a. John broke the vase (originator and measurer) b. The vase broke (only measurer)

This alternation is an instance of causativization. It can be noted that the change in the number o f syntactic arguments correlates with a change in the interpretation of the

event: the transitive verb denotes a causative event, while the intransitive one denotes an event o f change o f state (see discussion of causativization in chapter four).

The locative alternation:

The locative alternation is the alternation exhibited by verbs such as load, spray, smear, clear etc. It is found in English, Dutch, Hungarian, Japanese and Hebrew (see Rappaport and Levin 1988, Tenny 1992, Dowty 1991, Pinker 1989). Consider the following example:

(40) a. The farmer loaded the hay on the truck, b. The farmer loaded the truck with hay. (41) a. They cleared the table o f soiled dishes.

b. They cleared the soiled dishes off the table.

This alternation poses a problem to lexical-entry driven systems, because the same argument , theme or location, may be mapped into different positions in the two alternants. It thus violates any UTAH-based mapping system: both arguments can serve as direct objects, or alternatively be licensed by a preposition. Note that when the argument appears as a direct object is interpreted as a measurer o f the event by virtue o f this property: in (40a) it is the quantity o f hay which has been loaded which serves as a scale for the proceeding o f the event. The event terminates when all the hay has been loaded. In (40b), on the other hand, it is the truck which measures out the event: the event proceeds simultaneously with the truck getting loaded and terminates when it is completely full.

The dative alternation:

This is one of the best studied alternations in English. For some basic references see Oherle (1976), Larson (1988), Pinker (1989), Goldberg (1995), Pesetsky (1995):

(42) a. Paul gave Nina a book b. Paul gave a book to Nina.

The linking pattern of the two arguments, goal and theme, seems to be unstable, mapping them in one case into the direct object position and the other into the indirect object position. Many verbs undergo this alternation, including verbs o f transference

(send), verbs o f motion (throw) and verbs which take a beneficiary role:

(43) a. She sent a book to Mary, b. She sent Mary a book.

(44) a. She threw / passed / kicked the ball to Paul, b. She threw / passed / kicked Paul the ball.

(45) a. She bought / baked / got a cake for Nina, b. She bought / baked / got Nina a cake.

The same argument, the goal, may either appear as an object o f a preposition, or be an object o f the verb.

These three alternations are a representative sample of the set o f syntactic alternations in which verbs participate. Such alternation^ form a challenge to a lexical-

* ^ On the other hand, there are many verbs which are semantically almost identical to give-type verbs, but which do not allow this alternation:

(i) a. Mary donated a book to Oxfam. b. *Mary donated Oxfam a book.

entry driven system: if syntactic structures are trivially projected from the lexicon, then how to account for cases in which the same verb appears in two different structures?

Two main solutions have been employed to account for such alternations. One is a syntactic, derivational account: one alternant is basic, and the second alternant is derived from it. This solution was suggested for the dative alternation (Baker 1988, Larson 1988), as well as for the transitive-inchoative alternation. In both cases the alternants are identical at the level of D-structure. Syntactic considerations such as Case marking will motivate further movement at the level o f S-structure. ’ ^ Another solution is to assume that the difference between the alternants is captured in the lexicon. The two alternants have slightly different lexical representation which project different syntactic structures. This solution is proposed by Rappaport and Levin (1988) to account for the locative alternation: the same verb, load, is associated with two representations at the level of LCS, which further project two argument structure representations. The two LCS structures may be linked through lexical redundancy rules (cf. Bresnan 1980), thus avoiding listing /oaJ-type verbs twice in the lexicon.