This paper uses decennial Census data to examine trends in immigrant segregation in the United States between 1910 and 2000. Immigrant segregation declined in the first half of the century, but has been rising steadily over the past three decades. Analysis of restricted access 1990 Census microdata suggests that this rise would be even more striking if the native-born children of immigrants could be consistently excluded from the analysis. We analyze panel and cross-sectional variation in immigrant segregation, as well as housing price patterns across metropolitan areas, to test four hypotheses of immigrant segregation. Immigration itself has surged in recent decades, but the tendency for newly arrived immigrants to be younger and of lower socioeconomic status explains very little of the recent rise in immigrant segregation. We also find no evidence of increased nativism in the housing market. Evidence instead points to changes in urban form, particularly the tendency for ethnic enclaves to form as suburbanizing households leave older neighborhoods, as a central explanation for the new immigrant segregation.