An adjectival construction is evaluative if and only if it conveys that the property associated with the adjective exceeds a relevant threshold. The questions of which adjectival constructions are evaluative and why have formed the foundation for semantic theories of these constructions and of adjectives themselves (Klein 1980, von Stechow 1984), although it has been alleged that these theories are based on an incomplete picture of the phenomenon of evaluativity (Bierwisch 1989, Rett 2008a). We present the first experimental tests of the scope and nature of evaluativity across adjectival constructions and adjective types. These studies confirm that evaluativity is conditioned by adjective type (relative or absolute, Kennedy & McNally 2005) and is not restricted to the positive construction. However, they also show several new and surprising aspects of evaluativity: that it is perhaps better characterized as a gradable property than a binary one; that the ways in which relative and absolute adjectives differ in their evaluativity vary across construction; and that, contrary to standard intuitions, subjects are willing to attribute evaluativity to the subject position of comparative constructions like Sue is taller than Bill. We show that this last particularly surprising result reveals a lot about how subjects interpret contextually sensitive constructions, and we discuss its consequences for experimental studies and semantic theories of adjectival evaluativity as well as context-sensitive phenomena more generally.