This paper investigates the use of the that-complementizer with high-frequency matrix predicates such as I think, I suppose, which can also occur as medial and clause-final comment clauses. In clause-initial position they allow for the choice between that-complementizer and zero, as in (1), which raises the question whether these two constructional variants are functionally similar or different.

(1) I think (that) John lives in London

A crucial question in this connection is that of the syntactic status of the intial clause: is it a matrix clause, which syntactically governs a complement clause (e.g. Peterson 1999, Stenström 1995, Svensson 1976), or a parenthetical comment clause, which is in a syntactically supplementary relationship to the following clause (e.g. Kärkkäinen 2003, Thompson 2002, Thompson & Mulac 1991)?

The proposed paper tries to shed some light on this question by examining 200 occurrences of initial I think (52 of which with a that-complementizer) and other lexical predicates in a corpus of spoken English. More precisely, it investigates to what extent the corpus data provide evidence for the relative prominence of the initial clause, which in a cognitive-functional perspective is the underlying principle for distinguishing between main and subordinate clause status (e.g. Langacker 1991, Thompson 2002). In spoken language there are two formal cues for signalling prominence of the initial clause and hence a possible hierarchical difference between the two clauses: (i) the presence or absence of the that-complementizer as an explicit marker of syntactic subordination and (ii) prosodic prominence (cf. also Kaltenböck 2006, 2008). The present study takes a close look at the this syntax-prosody interface for different lexical predicates and investigates to what extent these parameters together with an assessment of the typical informational status of these predicates in context can provide evidence for a particular syntactic analysis of such initial clauses.

Analysis of the corpus data shows that a difference on the structural level, i.e. that vs. zero, does not correspond with a difference in prosodic realisation. Both constructional types exhibit a similar distribution of the three prosodic patterns identified: they are both most frequently realised as heads, less frequently as pre-heads, and only rarely with a separate nuclear accent. This means that the two formal signals available for indicating relative prominence of the initial clause, prosody and marker of subordination, do not correlate. The equivalence in actual use of the two syntactic variants, can be explained, however, by reassessing the function of the that-complementizer in spoken language. The corpus data show that it is not so much used as a marker of syntactic subordination but mainly as a filler, which is used to give weight to the initial clause or for rhythmic purposes.