Since its beginnings in the Chicago school, migration research has assumed that distinguishing between various immigrant communities and autochthons is the obvious starting point for understanding ethnicity. I show that this implies a Herderian perspective on the world which naturalizes its division into a series of distinct "peoples". Three major analytical and empirical problems of this approach are discussed, on the basis of comparative anthropological research. A more promising approach is the boundary-making perspective that looks at the dynamics of the emergence and transformation of ethnic groups. Seen from this perspective, "assimilation" and "integration" appear as reversible, power-driven processes of boundary shifting, rather than the result of overcoming cultural difference and social distance. The last section discusses four research designs that are most adequate for future work along these lines. They take territories, individuals, social classes, institutional fields or event chains instead of ethnic communities as units of analysis and observation.