It has been argued that each discourse aims to answer an explicit or implicit question, the Question Under Discussion (QUD) (see Klein/von Stutterheim 1992 and Simons et al. 2011). The QUD affects not only the interpretation of utterances, but it also affects the linguistic form of what is said. By several means, the grammar of a language marks content that aims to answer the actual QUD as at-issue content, whereas additional side-comments are marked as not-at-issue. In this way, grammar gives some hint about what kind of QUD an utterance may answer. Bartsch (1978), Hartmann (1984), Posner (1972), and von Stutterheim (1989, 168) assume that sentential subordination conventionally marks not-at-issueness. However, already Brandt (1989, 1996), Holler (2009), and Simons (2007) observe that embedded clauses can express at-issue content as shown in the following example from Simons (2007, 1035):
(1) Q: Where was Harriet yesterday?
B: Henry discovered that she had a job interview at Princeton.
Here, what is clearly relevant to the QUD is the content of the embedded clause. Thus, whereas the clause embedded by the semifactive verb discover can express at-issue content, the same is not possible for the complement of a factive predicate:
(2) Q: Where was Harriet yesterday?
B: #Henry ignores that she had a job interview at Princeton.
In this dissertation, I will investigate different types of dependent clauses, showing that they form two groups with respect to their discourse function. One group of dependent clauses is conventionally marked as not-at-issue and therefore cannot express an information that is relevant to the current QUD. Characteristic features of these dependent clauses are that they project under the scope of an entailment-cancelling operator and that they cannot be denied directly, what is going to be confirmed by the results of a questionary. In contrast, depending on the context, dependent clauses of the second group can express at-issue or not-at-issue content. Therefore, only a subset of the totality of dependent clauses can be used to answer
the QUD.
Interestingly, only dependent clauses that express at-issue content can undergo V2 movement in German, which is subsumed under the notion of Main Clause Phenomena. The licensing of V2 in German dependent clauses is typically explained in terms of presupposition and assertion. It is assumed that only asserted clauses can undergo V2 movement. In this dissertation, however, I will show that the Assertion Hypothesis leads to wrong predictions. Instead of relying on assertional force, I will argue that V2 in German dependent clauses is an optional marker of at-issueness. The contrast between dependent clauses that can express at-issue content and those that cannot is going to be investigated in more detail by comparing temporal and
causal clauses, whereas only the latter can be used to answer the current QUD.
Based on the results of Johnston (1994), I will show that this contrast can be traced back to a fundamental difference between the semantic contribution of temporal and causal conjunctions: Whereas causal weil expresses a relation between two events, temporal conjunctions such as bevor or nachdem are non-relational. As one-place
predicates they map their IP-complement to a time interval i which restricts the nuclear scope of an overt or covert quantifier. As quantifier restrictions, temporal clauses are subject to an existence presupposition and therefore cannot be used to express at-issue content.