Shifty Operators in Changing Contexts

It 's fair to say that the study of indexicality has been primarily occupied with an­ swering one question: Why are certain elements I, you, yesterday opaque to modal quantification? Indeed, Kaplan ( 1977), the most celebrated theory of index­ icality, is designed precisely to answer this question: indexicals are rigidly speci­ fied once the character of a sentence is applied to the utterance context, before the content is derived. And yet, recent work in a variety of languages (e.g . , Aghem (Hyman 1 979), Amharic (Schlenker 2003) , and Navajo (Speas 1 999)) has recently converged on the general conclusion that this central empirical claim of indexical research was too hasty. That is , in some cases in these languages, sentences with the form John said that I am hungry may report John's self-report of hunger. Based on data from two additional "indexical-shifting" languages, Zazaki1 and Slave,2 we argue that the interpretive possibilities of shifting indexicals are highly constrained. Our data come from three environments : cases with more than one embedded indexical , cases with diff erent types of attitude verbs, and cases with more than one embedded speech-report. These data give rise to two interesting restrictions on indexical interpretation: