In this presentation, I give a semantics of demonstratives and definite descriptions that captures the formers’ perspective-sensitivity. The treatment refines the framework from Roberts (2002,2015), which I implement in a dynamic semantics drawing on Dekker (1994), Stone (1999), and Bittner (2011). The linguistic phenomena that illustrate the perspective-sensitivity of demonstratives are as follows. First, two of the same demonstratives can be used in a single sentence to pick out different things (Maclaran 1982). Second, demonstratives sometimes signal that there is more than one relevant object satisfying their overt NP component (Maclaran 1982), and also sometimes have an emotive signal (Lakoff 1974). These two features of demonstratives result from (i) their conventional deictic features and (ii) the way that they interact with demonstrations. Here is roughly how (i) and (ii) are modeled in my semantics as aspects of the perspective-sensitivity of demonstratives. I add to the dynamic context a ranking of discourse referents that represents the perspectival dimension. Demonstrations are modelled as updates that bump an existing discourse referent to the top of the perspectival dimension and require that it bear a certain deictic relation to the speaker. Demonstratives are interpreted as variables that pick up the highest ranked member of the perspectival dimension that bears a deictic relation to the speaker corresponding to the demonstrative’s conventional deictic features. Finally, in certain cases, when a demonstrative is used without a demonstration, the accommodation of the deictic features leads to an emotive signal.