Speech reports exhibit various types of mismatches between the complement of the speech predicate and the original utterance in the communicative event being reported. This paper identifies a new case of mismatch in fill-in-the-blank (FitB) contexts, where the agent of the reported speech does not produce a complete sentence but only a part of the full clause being attributed to them. In addition, this paper illustrates that speech reports in FitB contexts exhibit focus sensitivity. This paper proposes that speech reports in FitB contexts are governed by two requirements: one in terms of contextual commitment and one in terms of a relation between the focus structure of the embedded clause and the Question Under Discussion (QUD) in the reported communicative event. In particular, extending Anand & Hacquard's (2014) programmatic semantics of speech reports, this paper proposes that, in addition to the common ground, speech predicates can also access the QUD in the reported communicative event.