To examine the association between incidence of measles and immunization coverage among preschool-age children. An ecological study in which measles incidence was compared with immunization coverage among census tracts. The independent effects of race and population density were controlled for. A recent measles outbreak in Milwaukee, Wis. Immunization coverage data were estimated from a retrospective, school-based survey of Milwaukee grade school students. One thousand eleven persons (less than or equal to 17 years) who had confirmed measles from September 1989 through June 1990. Confirmed measles cases grouped by census tract, corresponding census tract preoutbreak immunization coverage, racial breakdown, and population density. Census tracts stratified into four levels, with mean immunization rates of 50.4%, 60.2%, 69.9%, and 81.0%, had respective median attack rates of 11.6, 5.0, 1.7, and 0.0 cases per 1000 persons (P less than .01). The association between immunization coverage and measles attack rate remained significant even after controlling for race and population density. Modest improvements in low levels of immunization coverage among 2-year-olds confer substantial protection against measles outbreaks. Coverage of 80% or less may be sufficient to prevent sustained measles outbreaks in an urban community.