I test these hypotheses using a national sample of children who live in metropolitan areas drawn from the 1990 Census. I find that levels of Spanish maintenance are extremely high among children of Latina/o immigrants, and that a large majority of children who are born in the U.S. speak English fluently. Multivariate analysis demonstrates that several dimensions of a metropolitan area's language context-in particular the saturation and segregation of Spanish speakers-have a strong effect on second-generation children's likelihood of speaking Spanish that persists even after controlling for household- and individual-level variables. Contrary to my original hypothesis, I also find that the language characteristics of the metropolitan area have a significant effect on children's English proficiency. This effect, however, is smaller than the effect of metropolitan context on Spanish use.
This analysis produces a better understanding of the specific elements of household and community context that affect language use. The results imply that children of immigrants are following multiple paths to language adaptation, and that metropolitan context is an important influence on this process of adaptation.