Joint work with Ilaria Frana, University of Goettingen

Heim (1979) famously observed that sentences like (1) are ambiguous:

(1) Miles knows the price that Clara knows.
    = Miles knows what the price of the iPhone is, and Clara does too. ("what value"/A reading.)
    = Miles knows which product Clara knows the price of. ("which concept"/B reading.)

Understanding this ambiguity and its consequences have been a central to understanding "concealed questions" and question-embedding verbs in general (Heim 1979, Romero 2005, Nathan 2006, Aloni 2008, Aloni and Roelofsen 2009, Frana 2010, Percus 2010, etc.). The most established proposal (Romero) is that concealed question DPs (such as the complement of "knows" in (1)) denote individual concepts or their intensions, and in fact aren't questions at all.

We present new data in the form of unconditional concealed questions, as in (2), and relational noun questions, as in (3), that lack parallel ambiguities in an unexpected direction. Furthermore, there is an unexpected interaction with verb type. With stative epistemic verbs such as "know", only a which-concept reading is available. For example, in (3a) one cannot answer with a price, only a product. With eventive communication verbs such as "tell" in (3b), however, either a price or a product might answer the question.

(2) a. Regardless of the price that Miles knows, I'll buy an iPhone.
    b. Regardless of the price that Miles tells us, I'll buy an iPhone.
(3) a. What price did Miles know?
    b. What price did Miles tell you?

The lack of "which-concept" readings in these examples present great difficulties for existing theories of concealed questions. We use such data to argue for the following three conclusions: (i) An understanding of the nature of concealed questions requires a fine-grained understanding of the properties of a range of particular predicates. (ii) What have been called concealed question DPs are heterogeneous: they are sometimes, but not always, question-denoting; a hybrid question/individual concept account is necessary. (iii) contra existing proposals, some apparent instances of Heim's ambiguity correlate with previously undetected ambiguities in certain verbs themselves, in particular whether the verb meaning licenses extensional/referential readings. From the analysis we propose, we can gain the beginnings of new insights into the lexical content of attitude predicates.