Conclusions

In this chapter, I have argued with reference to data from the dative and benefactive constructions in comparison with other adjunctive constructions that there is a three-way distinction in the syntactic and semantic function which a prepositional phrase can play in a sentence.

• complement PPs: syntactically obligatory PPs which introduce a semantically entailed component of a verbal relation. The contribution of the PP can either be idiosyncratic, or can be the standard semantic contribution of that PP on a pseudo-complement use.
• pseudo-complement PPs: syntactically optional PPs which either mark an entailed participant in a verbal relation or extend a verbal relation in a way licensed by the verbal relation hierarchy.
• adjunct PPs: syntactically optional PPs which provide situational information but do not interact with the verbal relation expressed in the main clause.

Making this distinction allowed development of a lexical rule-based treatment of PP (more generally, adjunct) integration driven by the semantic properties of the prepositions and the verbs, as lexically defined. Combinations of VPs and PPs are constrained through the semantic relation hierarchy and the types of individual prepositions, in conjunction with lexical rules which control in very general terms the composition of semantics from the combined phrases.

The approach eliminates lexical representation of syntactically optional verbal ``complement'' PPs, instead treating these PPs as pseudo-complements and allowing them to be productively licensed on the basis of semantic properties of the verbs. This captures a generalisation about a verb's (potential) syntactic argument structure on the basis of lexical semantics.

The approach explicitly acknowledges the semantic content of prepositional phrases, and shows how this content may interact in various ways with the semantic content expressed by other elements in a sentence. Furthermore, syntactic ordering is taken into consideration in determining how this interaction might occur, thereby ruling out infelicitous readings (e.g. a PP which appears after other adjuncts cannot have a pseudo-complement interpretation). Finally, the lexical semantically derived interpretation of certain constructions was briefly discussed and shown to require support from pragmatic reasoning for complete felicity.

The proposals made in this chapter concerning the treatment of adjuncts go a long way towards appropriately handling the characteristics of adjuncts:

• Consistent semantic contribution: There is only one lexical entry required in this approach for each meaning associated with an adjunct, even if the adjunct is involved in different types of adjunction.
• Restrictive and operator adjuncts: Both of these types of adjuncts are accounted for and treated in a way which reflects precisely the type of modification which must be associated with them; namely that restrictive adjuncts directly modify the situation expressed by the verb and that operator adjuncts take a full situation as an argument. A third type of adjunct, thematic adjuncts, has also been identified as a type of adjunct which adds information about a situation as a whole.
• Surface order vs. semantic obliqueness: The interaction between surface order and semantic obliqueness for operator adjuncts is accounted for by maintaining lists which reflect both of these types of information -- surface order in the subcat list and semantic obliqueness in the op-adjs list.
• Redundancy constraints: Redundant PPs are avoided through use of a type system which keeps track of modifying information associated with a sentence. The lexical rules would then simply need to include a subsort check to prevent two modifiers of the same type in a sentence.
• Mittelfeld phenomena: Complements and adjuncts both appear on the subcat list of a head. There is thus nothing structural which prevents these elements from being interspersed. The linear precedence rules must be responsible for determining their allowed relative order.

The advantages of the approach presented in this chapter over the previous approaches from which it is derived can be summarised as the following:

• The division between external and internal semantics allows various types of modification, including types not handled in the previous approaches (thematic adjuncts and pseudo-complements), to be accommodated within the same framework. In particular, the use of a semantic object of type thematic common to external and internal semantics provides for a general treatment of prepositions which can behave both as a thematic adjunct and as a pseudo-complement. This treatment can even account for the ambiguity of interpretation found in sentences involving such prepositions.
• Dative alternation can be easily accounted for by defining variants of the basic pseudo-complementation lexical rule. It follows only from specific underlying lexical semantic properties of a verb.
• The interspersal of operator adjuncts with other types of adjuncts does not lead to interpretation errors.
• There is a more straightforward framework in which to account for the interaction between surface order and semantic precedence. The use of delayed evaluation and linear precedence rules which follow surface order allows the context to drive adjunct interpretation.