In 1994-1998, the authors followed 946 breastfeeding women from Michigan and Nebraska for the first 3 months postpartum or until they stopped breastfeeding to describe mastitis incidence, mastitis treatment, and any associations between mastitis occurrence and hypothesized host characteristics and behaviors. Participants were interviewed by telephone at 3, 6, 9, and 12 weeks postpartum or until they ceased breastfeeding. A total of 9.5% reported provider-diagnosed lactation mastitis at least once during the 12-week period, with 64% diagnosed via telephone. After adjustment in a logistic regression model, history of mastitis with a previous child (odds ratio (OR) = 4.0, 95% confidence interval (CI): 2.64, 6.11), cracks and nipple sores in the same week as mastitis (OR = 3.4, 95% CI: 2.04, 5.51), using an antifungal nipple cream (presumably for nipple thrush) in the same 3-week interval as mastitis (OR = 3.4, 95% CI: 1.37, 8.54), and (for women with no prior mastitis history) using a manual breast pump (OR = 3.3, 95% CI: 1.92, 5.62) strongly predicted mastitis. Feeding fewer than 10 times per day was protective regardless of whether or not feeding frequency in the same week or the week before mastitis was included in the model (for the same week: 7-9 times: OR = 0.6, 95% CI: 0.41, 1.01; < or =6 times: OR = 0.4, 95% CI: 0.19, 0.82). Duration of feeding was not associated with mastitis risk.