Please join us this Friday as Sophia Sklaviadis from the Philosophy Department presents work on linguistic subjectivity and faultless disagreement.

**Date and time: **Friday, Apri 1, 10:30 a.m. – 12:20 p.m.

**Location:** Rosenwald 208 (Linguistics seminar room)

**Title:** Exemplars of honesty (i.e. felicity): Can subjective ‘find’ discover?

or: Can naturalness of embedding under (subjective) ‘find’ predict scalar faultlessness?

**Abstract:**

Knobe & Khoo (2015) define an interesting relation of exclusionary content: “Two claims are *exclusionary* (or have exclusionary content) iff it has to be the case that at least one of them has to be false.” Equivalently, two claims are exclusionary iff they are contradictory but at most one is correct (or true). We can define the inverse of Knobe & Khoo’s (2015) exclusionary content as follows: Two claims are *inclusionary* (or have inclusionary content) iff it has to be the case that *neither* of them has to be *false*. Inclusion seems like a natural charterization of the relation obtaining between the contradictory assertions (as well as their expressed contents[1]) that give rise to *faultless disagreement* (FD), which is prominently licensed by *predicates of personal taste* (PPTs), e.g. ‘tasty’, ‘fun’, in their positive and comparative forms. [2] Far from being exhausted by PPTs, faultlessness is a pervasive phenomenon that can mediate communication: *relative* *gradable* predicates license FD in their positive forms (but block it in comparative constructions[3]); *absolute* *gradable* predicates are interesting intermediary cases. Besides gradable adjectives, a well-known class of examples of FD is licensed by epistemic modality effects.[4]

Studies on the cognitive psychology of intuitions about disagreements, suggest that inclusion (or faultlessness) is itself a gradable or scalar property.[5] At the same time, empirical work on the linguistic behavior of vague predicates (specifically relative gradable adjectives, including ‘tall’, ‘delicious’, ‘salty’, and ‘green’) suggests that (the class of) relative faultlessness of disagreements about vague predicates, depends on the (semantic) meaning of the particular disputed vague predicate.[6] I try to characterize two empirical measures, whose correlations, are (hopefully) meant to help understand the relation between (degree of) inclusion or faultlessness, and vagueness (in the sense of *tolerance* defined as: (i) giving rise to Sorites series, (ii) licensing *borderline *cases[7]): One measure will be the mean score of inclusion of disagreements about each of 41 predicates (modifying Knobe & Khoo’s (2015) empirical model). The second measure will be the relative difference in felicity (or naturalness or acceptability) of each of these predicates, embedded, respectively, in the nonfinite complement clauses of the two matrix verbs, ‘find’ and ‘consider’.[8] My basic hypothesis is that perceived subjectivity influences the degree to which people find it acceptable to embed an adjective under ‘find’. A first model will be a multiple regression of the judgments about felicity of each predicate under ‘find’, on both judgments about each predicate under ‘consider’, and judgments about FD for each predicate.[9] I also want to use (i) the ranked differences of the ‘find’ and ‘consider’ mean scores, and (ii) an ordering by fitted residuals of a fit between observed-‘find’ score and predicted-‘find’ score (predicted-‘find’ is estimated by linear regression on ‘find’ scores from ‘consider’ scores),[10] in order to see how well these orderings can predict the mean faultlessness scores.

[1] I want to follow Knobe & Khoo (2015) in remaining neutral on the usefulness of this kind of a distinction.

[2] Kölbel (2003); Lasersohn (2005); Kennedy (2013).

[3] Fleischer (2013) who develops a different explanation of this observation.

[4] vonFintel & Gilles (?); Willer (2013).

[5] Cohen & Nichols (2010); Knobe & Khoo (2015); Goodman, et al. (2016); Sarkissian, et al. (2011).

[6] Cohen & Nichols (2010); more recently (and somewhat more indirectly) Lassiter & Goodman (2014).

[7] Cf. Cobreros, Egre, vanRooij, Ripley (2012), (2015).

[8] The construction ‘find x pred’, contrasted with ‘consider’, and especially in English, has strong selection restrictions on the distribution of its (nonfinite) complement clauses: cf. Szaebo (2009); Kennedy (2013).

[9] The intuitive idea motivating this model is the following: The *variance* in felicity of ‘find’ judgments that is unrelated to subjectivity will be accounted for by the comparatively neutral embedding matrix ‘consider’; if this model works, the ‘find’ variance that is left over will be predicted by faultlessness judgments.

[10] This ordering represents ‘find’ being increasingly preferred to ‘consider’ when the same predicate is embedded in the respective complements. (This fitting is motivated by the assumption that ‘consider’ is neutral in the sense of not imposing (semantic) restrictions on its nonfinite complement clause.)