This article examines ethnic segregation, defined as segregation among racial groups as well as between native-born and immigrant students, across elementary school classrooms in New York City. Specifically, the study compares patterns in within-school segregation across ethnic groups, grades, boroughs, and years. Current levels of within-school segregation are also compared to levels of across-school segregation and to levels of segregation that result from three simulations where students are assigned to their classrooms: (a) randomly, (b) to achieve complete ethnic segregation, and (c) according to their prior year test scores. Results indicate that racial segregation across schools is far greater than racial segregation within schools, however the segregation of immigrants within-schools is equal to the segregation of immigrants across schools. Within-school segregation cannot be entirely attributed to random processes or to the use of ability grouping practices, particularly in the case of black and Hispanic segregation. Finally, segregation within-schools varies considerably across the five boroughs and declined during the second half of the 1990s. (Contains 9 tables, 1 figure, and 19 notes.)