The hypothesized association between breast cancer and circadian disruption was evaluated in the Electromagnetic Fields and Breast Cancer on Long Island Study. Participants included 576 women with breast cancer diagnosed from August 1996 to June 1997 and 585 population-based controls (87% and 83% participation rates, respectively) aged < 75 years and living in the same Long Island, New York, home for > or = 15 years. An in-person interview ascertained light-at-night exposure histories through shift work (previous 15 years) and at home (previous 5 years). Odds ratios and 95% confidence intervals were estimated by unconditional multivariate logistic regression. Breast cancer was not associated with overall shift work (odds ratio (OR) = 1.04, 95% confidence interval (CI): 0.79, 1.38) or evening shift work (OR = 1.08, 95% CI: 0.81, 1.44). However, overnight shift workers were at lower risk than women never working shifts (OR = 0.55, 95% CI: 0.32, 0.94). Women who frequently turned on lights at home during sleep hours (> or = twice/week and > or = twice/night) had increased risks (OR = 1.65, 95% CI: 1.02, 2.69). The latter results suggest positive associations with residential light-at-night exposure, or they could reflect response biases. Furthermore, overall and evening shift work were not significant factors, and analyses of overnight shift workers yielded reduced risk estimates. The study thus provides mixed evidence for the light-at-night hypothesis.