S or VP complements?

Two comparable grammatical formalisms, Generalized Phrase Structure Grammar (GPSG) [#!gazdar85!#] and Head-driven Phrase Structure Grammar (HPSG) [#!PollardSag94:BK!#], have rather different treatments of sentential complements (S-comps). They both treat embedded sentences as VP's with subjects, which generates the correct structures but misses the generalization that S's behave similarly in both matrix and embedded environments, and VP's behave quite differently. Neither account has PRO  subjects of infinitival clauses- they have subjectless VP's instead. GPSG has a complete complementizer system, which appears to cover the same range of data as our analysis. It is not clear what sort of complementizer analysis could be implemented in HPSG. Following standard GB approach, the English XTAG grammar does not allow VP complements but treats verb-anchored structures without overt subjects as having PRO subjects. Thus, indicative clauses, infinitives and gerunds all have a uniform treatment as embedded clauses using the same trees under this approach. Furthermore, our analysis is able to preserve the selectional and distributional distinction between S's and VP's, in the spirit of GB theories, without having to posit `extra' empty categories.^8.1 Consider the alternation between that and the null complementizer^8.2, shown in sentences ((12)) and ((13)).

(11)0(11

(12)

He hopes $\emptyset$ Muriel wins.  (12)0(12

(13)

He hopes that Muriel wins. 

In GB both Muriel wins in ((12)) and that Muriel wins in ((13)) are CPs even though there is no overt complementizer to head the phrase in ((12)). Our grammar does not distinguish by category label between the phrases that would be labeled in GB as IP and CP. We label both of these phrases S. The difference between these two levels is the presence or absence of the complementizer (or extracted WH constituent), and is represented in our system as a difference in feature values (here, of the <comp> feature), and the presence of the additional structure contributed by the complementizer or extracted constituent. This illustrates an important distinction in XTAG, that between features and node labels. Because we have a sophisticated feature system, we are able to make fine-grained distinctions between nodes with the same label which in another system might have to be realized by using distinguishing node labels.