New data from the IPUMS (Integrated Public Use Microdata Series) project permit an exploration of the demographic basis for ethnic survival across successive generations. I first explore the degree of ethnic blending among the grandchildren of early- to mid-19thcentury German immigrants; second, these descendants' own marital choices; and third, the likely composition of the fourth generation to which they would give birth. Fundamental questions include: How high is the rate of single versus mixed origins after so many generations in America? How large an absolute number of single-origin individuals remain (given the combined impact of out-marriage, on the one hand, and cumulative fertility, on the other)? How much less likely are single-origin individuals of the third generation to in-marry relative to those in the second generation? And how do all these patterns differ across 31,000 local geographic areas? I exploit the full-count 1880 Census dataset and the Linked Representative Sample, which captures males in 1880 as well as in one of the 1900-30 enumerations. Limiting attention to those who were adolescents in 1880, we have three generations' worth of ethnic information on each sample member traced across time (birthplace as well as parents' and grandparents' birthplaces, from their parents' responses) and ethnic information covering two generations for the women they eventually married.