A study of breast cancer survival was conducted among New Mexico Hispanic and non-Hispanic white women and New Mexico and Arizona American Indian women diagnosed between 1973 and 1992. The goals were to determine whether, after adjusting for first treatment and the extent of disease at diagnosis, American Indian and Hispanic women had poorer survival than non-Hispanic whites and, if survival had improved over time, whether comparable improvements had been made for the three racial/ethnic groups. Five-year relative survival rates were calculated, and a Cox proportional hazards model was constructed to compare survival between races/ethnicities, adjusting for first treatment and the extent of disease at diagnosis. Findings indicate that during 1983-1992, breast cancer was more commonly detected at a local stage for all three groups compared to 1973-1982. Five-year relative survival improved for non-Hispanic white and American Indian women with local or regional disease, but the improvement was statistically significant only for non-Hispanic white women and for American Indian women with local disease. Despite earlier stages at diagnosis, Hispanic females showed less improvement in overall or stage-specific survival than non-Hispanic whites. The Cox model indicated that American Indian women experienced poorer survival than non-Hispanic whites during both time periods. Survival of Hispanic women with breast cancer was comparable to non-Hispanic whites during 1973-1982 but was significantly worse during 1983-1992. The significance of this lower survival is amplified by increasing breast cancer incidence among New Mexico Hispanics and American Indians.