The authors prospectively investigated associations between potentially stressful work characteristics and type 2 diabetes incidence in 62,574 young and middle-aged women, aged 29-46 years at baseline in 1993, from the Nurses' Health Study II; 365 cases of type 2 diabetes accrued over 6 years of follow-up. Cox proportional hazards regression was used to simultaneously evaluate associations of hours per week in paid employment, years of rotating night-shift work, and job strain with incidence of type 2 diabetes. In multivariate-adjusted analyses, women working less than 20 hours per week had a lower risk of diabetes (relative risk = 0.80, 95% confidence interval: 0.50, 1.30), and those working overtime (> or =41 hours/week) had an elevated risk of diabetes (relative risk = 1.23, 95% confidence interval: 0.98, 1.55) compared with women working 21-40 hours/week (referent) in paid employment (p(trend) = 0.03). In subsequent analysis, the elevated association appeared stronger in unmarried women (p(interaction) = 0.02). A positive association between years in rotating night-shift work and diabetes was mediated entirely by body weight. Job strain was unrelated to risk of type 2 diabetes. In conclusion, working overtime predicted a slightly elevated risk of type 2 diabetes in young and middle-aged female nurses.