In this talk I will compare the iconic representation of other people's attitudes via written language quotation and sign language role shift. In writing, quotation is traditionally taken to be atomic verbatim representations of another's attitude, while sign language role shift has been argued to be something different, more integrated into the rest of a sentence. I argue that the missing link is spoken language "quotation"/direct speech reports, which are much closer to sign language role shift in providing a demonstration (Clark and Gerrig 1990, Lillo-Martin 1995) and not a verbatim representation. However, demonstrations aren't restricted just to attitude reports: one can quite easily demonstrate an action, either via co-speech gesture or, as has been argued by Zucchi et al. (2012), in sign language classifier predicates, and, I will argue, in Action Role shift (Schlenker 2014). In particular, I suggest that the same semantics of event modification should be used for the iconic demonstrations in quotation, classifiers, and role shift, and provide evidence coming from various domains including the morphological structure of classifiers and the language of bimodal ASL/English bilingual children.