When referring to an object using a description, speakers need to select properties which jointly distinguish it from any potential distractors. Previous empirical and computational work addressing this content selection process has highlighted the role of both (i) the discriminatory power of properties of a referent, i.e. how many of the distractors in a domain each property excludes; (ii) how inherently salient or preferred a property is. To date, there has been no attempt to systematically investigate the trade-o between these two potentially competing motivations. This paper investigates experimentally the extent to which speakers take discriminatory power versus preference into account during content selection for reference production. Our results suggest that discriminatory power in fact plays a relatively unimportant role. We discuss the implications of this for computational models of reference production.