Rafts have been conceptualized as lateral heterogeneities in the organization of cholesterol and sphingolipids, endowed with sorting and signaling functions. In this review we critically examine evidence for the main tenet of the 'raft hypothesis', namely lipid-dependent segregation of specific membrane components in the plasma membrane. We suggest that conventional approaches to studying raft organization wherein membranes are treated as passive, thermally equilibrated systems are unlikely to provide an adequate framework to understand the mechanisms of raft-organization in vivo. An emerging view of raft organization is that it is spatio-temporally regulated at different scales by the cell. This argues that rafts must be defined by simultaneous observation of components involved in particular functions. Recent evidence from the study of glycosylphosphatidyl inositol-anchored proteins, a common raft-marker, supports this picture in which larger scale, more stable rafts are induced from preexisting small-scale lipid-dependent structures actively maintained by cellular processes.