Assessor-Relativizable Predicates Phil Crone — Stanford // Deniz Rudin — UC Santa Cruz [email protected], [email protected] A large number of predicates optionally allow for modification by a phrase of the form “to x,” where x is some assessor. This class includes predicates of personal taste (PPTs), eg. This chili is tasty (to me/you/Donald), predicates of clarity, It is clear (to him/her/Hillary) that Mary is a doctor, and a wide variety of other predicates (e.g. frightening, shocking, acceptable, and so on). We refer to these predicates as “assessor-relativizable predicates” (ARPs). We discuss empirical generalizations about how BARE ARPs (without “to x”-modification) relate to PERSONAL ARPs (with “to x”-modification). We explain these generalizations via a semantics containing a variable whose default interpretation is an “archetypically qualified assessor,” but which can be bound and saturated by explicit “to x”-modification. Bare ARPs typically commit the speaker to the same predication relativized to the speaker (1). Conversely, (non-speaker oriented) personal ARPs carry no implications about whether the ARP holds relativized to the speaker (2). (1)

This chili is tasty. ↝ This chili is tasty to me.

(2)

This chili is tasty to Donald Trump ↝ / This chili is tasty to me.

However, there are cases in which the implication in (1) does not go through. These include deep past tense uses (3) and uses in which the speaker’s assessment is irrelevant to the applicability of the ARP (4). (3)

It was clear in Ptolemy’s time that the Earth was at the center of the universe.

(4)

This cat food is tasty. Whenever I feed it to Spot, she eats the whole bowl.

In addition, Bare ARPs typically have a normative flavor: asserting them encourages addressees to adopt the same opinion. Personal ARPs do not have this normative flavor. An adequate semantics for ARPs should explain these empirical generalizations while also explaining why only predicates in this class accept “to x”-modification (via a compositional account of such modifiers). We account for these properties by taking all ARPs to have in their semantics an assessor variable α. Explicit “to x”-modification binds α via lambda abstraction and saturates it with the value of x. However, if no explicit assessor is introduced, α is saturated with a default value QUAL, which denotes an archetypically qualified assessor. On this account we expect the inference in (1) to go through in a given context iff the speaker is taken to be a qualified assessor of the tastiness of the chili in that context. In deep past contexts, the speaker fails to be a qualified assessor in virtue of not being alive at the relevant time (3). In other contexts, the speaker fails to be the relevant kind of assessor, e.g. the speaker of (4) is not a cat. The normativity of bare ARPs follows from the fact that assertions of bare ARPs ask interlocutors to accept either that they find the chili tasty or that they are not qualified to judge it. Though to the best of our knowledge there are no previous accounts of the behavior of ARPs as a class, the literature contains several prominent analyses of PPTs and predicates of clarity. We contrast our account with existing accounts of PPTs (Lasersohn 2005, Stephenson 2007, Pearson 2013), as well as existing accounts of assertions of clarity (Barker & Taranto 2003, Barker 2009, Wolf & Cohen 2011). We focus on accounts that introduce quantification and that are judge dependent, arguing that our account allows for a better understanding of ARPs as a class, while preserving the key empirical insights of prior work.

1