My talk will be about moods in the evidential family: (arguably) the indicative and (definitively) the German reportative subjunctive (Konjunktiv I). Evidential moods, and moods more generally, are usually seen as dependent or ‘selected’ by certain verbs that embed sentential complements. This has all been a mistake, I think. I will argue that evidential moods (and moods more generally) are not SELECTED by attitude verbs, they are responsible for CREATING the modal semantics of attitude verbs.
More generally, the talk will suggest a different way of thinking about sentential complements of attitude verbs and verbs of speech. Attitude ascriptions and speech reports have a special place in Cognitive Science. ‘Recursion hunters’ (The New York Times, March 21, 2012) try to track them down in every human language, looking for evidence for complex syntax. Developmental psychologists consider them milestones in the cognitive development of children. Philosophers have linked them to ambiguities that don’t seem to exist anywhere else. What is it that makes attitude ascriptions and speech reports stand out? Why are they so hard to acquire? Where do those curious ambiguities come from? How are speech and attitude reports expressed in the languages of the world? What can vary? What explains the variability we see?
If there is anything that makes sentences like (1) or (2) special, it’s probably not the matrix verbs or the embedded clauses all by themselves.
(1) Mo thinks these are turnips.
(2) But Bo says they are parsnips.
A better bet is that what’s special about (1) or (2) is the way matrix verbs and embedded clauses connect up with each other. Current linguistic theory has trivialized this connection: embedded clauses like those in (1) and (2) are typically taken to denote propositions that, directly or indirectly, provide propositional direct objects for the matrix verb. Questioning and moving away from this assumption will pave the way towards explaining some puzzling facts about apparent embedded ‘complements’ in natural languages. More importantly, this step will uncover a surprisingly rich and varied toolkit of techniques and building blocks that natural languages use to construct speech and attitude reports.