Complement clause selection as a (non-)local relation - a talk at the Cambridge Linguistics Forum by Professor Anna Roussou, University of Patras

Professor Anna Roussou, Professor of Philology at the University of Patras and Lewis-Gibson Visiting Fellow at the Cambridge Centre for Greek Studies, discusses the grammar of complement causes in a seminar hosted by the Cambridge Linguistics Forum in partnership with the Cambridge Centre for Greek Studies.

Abstract

1. Complement clause selection is treated as a local relation. Thus, a verb like ‘believe’ takes a that-complement as in (1), while a verb like ‘wonder’ takes an if/whether-complement or a wh-clause as in (2):

(1) a. I believe that/*if Mary left b. *I believe who left

(2) a. I wonder if/*that Mary left b. I wonder who left.

The ungrammatical versions in (1) and (2) are consistent with the view that while c-selection is the same (a clausal complement), the predicates differ in terms of s-selection (an embedded declarative vs an embedded interrogative) (Grimshaw 1979, Pesetsky 1982, a.o.). On the other hand, (some) propositional attitude predicates may allow for an embedded interrogative under certain conditions, as in (3):

(3) a. I don’t know if John left b. I know who left.

b. *I don’t believe if John left b. I can’t believe who left.

Adger & Quer (2001) argue that the if-clause in (3a) is an Unselected Embedded Question (UEQ) which is subject to the same licensing conditions as polarity items. This renders the complement clause an indefinite-like element, attributed to the properties of if (Roussou 2010). Still the problem remains for the examples in (3) and (4).

2. The patterns in (3) and (4), replicated for other languages including Greek, are taken as evidence that complement clause selection is not necessarily a local relation. It is argued that selection in this case is not computed at the vP phase but extends to the next phase up, namely C; this is what gives rise to the ‘non-local’ flavor. Assuming that the distinction between c- and s-selection is valid, the picture that arises shows that s-selection is the one computed at the C phase. Within this perspective, the look-ahead problem that arises in (3) and (4) is resolved, as far as a bottom-up derivational model is concerned. Further data from Greek support this line of theorizing with respect to other phenomena such as control, triggered by properties of the matrix clause in connection with the selecting predicate.

The Department of Theoretical and Applied Linguistics (DTAL,) part of the Faculty of Modern and Medieval Languages and Linguistics (MMLL,) hosts the the Cambridge Linguistics Forum to promote and support linguistics research at Cambridge University.

Photo by Amador Loureiro on Unsplash