The purpose of this study was to assess the validity and reproducibility of an interviewer-administered, semi-quantitative food frequency questionnaire (FFQ) among 132 volunteer New Mexico Hispanic (H) and non-Hispanic white (NHW) women, aged 35-74 years, with (n = 47) and without (n = 85) a breast cancer history, and to add to the limited data presently available on the performance of FFQs among different ethnic groups. Validity was measured at one month and six months from baseline against four-day food records, and reproducibility was tested by comparing FFQs. Unadjusted validity correlation coefficients were highest at one month, ranging from 0.38 (polyunsaturated and monounsaturated fat) to 0.57 (calcium); energy-adjusted correlation coefficients were highest at six months, ranging from 0.15 (polyunsaturated fat) to 0.68 (calcium). Energy-adjusted correlation coefficients were statistically significant by ethnicity for vitamins A and C, protein, carotene and calcium, and by case status for saturated fat, folate, fiber, and vitamins A and E. Reproducibility correlation coefficients (unadjusted) ranged from 0.40 (polyunsaturated fat) to 0.71 (carbohydrate, retinol); energy-adjusted correlation coefficients ranged from 0.42 (vitamin E) to 0.78 (fiber), and differed significantly by ethnicity for saturated fat and retinol, and by case status for carbohydrate. Overall, our FFQ has comparable characteristics to other FFQs and is suitable for use with New Mexico's H and NHW women.