Dedicated de se and complement type: an argument against Uniformity

I argue for a semantic distinction between two classes of attitude predicates. One class is best modeled as quantifying over possible worlds compatible with what the attitude holder believes/says/expects/wants, in the tradition of Hintikka 1969. The other is best modeled as quantifying over individual-world-time triples such that the attitude holder believes/says/expects/wants that she might be the individual at the time in the world, in the tradition of Chierchia 1989 (in turn inspired by Lewis 1979). Much recent work has assumed that all attitude predicates are to be treated semantically in this second manner (Schlenker 1999, Anand 2006, Pearson 2015, Grønn and von Stechow 2010, i.a.). I refer to this hypothesis as Uniformity. Uniformity predicts that all attitude complements should be equally semantically able to host elements that must refer de se, such as shifted indexicals. Drawing on new evidence from Nez Perce, I demonstrate that this prediction is false, and argue that the best explanation for the distribution of dedicated de se elements comes from variation in whether or not attitude predicates quantify over worlds vs. centered triples.