In their article in this issue of 'Cognitive Linguistics', Peirsman and Geeraerts (henceforth P and G), consider a very wide range of semantic shifts to be instances of metonymy. P and G propose a single conceptual process to account for all of these supposed metonymic semantic shifts, namely contiguity. P and G provide a prototype definition of contiguity. In great detail, P and G classify the range of examples of metonymy in their article according to three 'dimensions' of prototypicality: strength of contact, degree of boundedness and domain (from space to time to spatio-temporal and 'categorical' domain). In this comment, I argue that contiguity is neither a necessary nor a sufficient condition for metonymy, even granting the prototype definition of contiguity provided by P and G; that association, analyzed in terms of domain highlighting (Croft 1993, 2002) provides a valid necessary condition; and that many examples taken by P and G (and other cognitive linguists) to be metonymy are better explained in terms of other theoretical constructs developed in cognitive linguistics. (William Croft)