Tiwa, a Tibeto-Burman language of northeast India, has a disjunction particle ba which must take narrow scope with respect to any operator higher in the structure. In this talk, I present a novel argument from phrasal comparatives against a Boolean approach to Tiwa's ba disjunction, and develop an analysis that treats disjunction as alternative-denoting (following Simons 2005, Alonso-Ovalle 2006, Aloni 2007, a.o.). I consider how ba's obligatory narrow scope can be derived within this framework, and the larger question of where the locus of variation in scope-taking lies (i.e. how to rule out a scope mechanism that seems to apply in one language, but not another).