This article investigates whether evolutionary accounts can offer new insights into the paradox of language change. Specifically, it examines three recent influential accounts (Haspelmath 1999, Keller 1994, and Croft 2000). As they contain a broad spectrum of positions on the relations between language and biology, they can be divided into metaphorical, biologistic and generalized views. Cross-cutting these, two types of evolutionary accounts are distinguished, which are labelled adaptive and two-level views, respectively. The paper critically evaluates their potential to provide satisfactory explanations for various types of change, drawing on examples from Romance and Germanic. Finally, it proposes a revised explanation scheme which brings together the two-level approaches with theoretical distinctions and explanatory factors that have been suggested in earlier non-evolutionary frameworks, so that a more comprehensive view of language change can be obtained.