We investigated the risk of stroke related to long-term ambient air pollution exposure, in particular the role of various exposure time windows, using four cohorts from Stockholm County, Sweden. In total, 22,587 individuals were recruited from 1992 to 2004 and followed until 2011. Yearly air pollution levels resulting from local road traffic emissions were assessed at participant residences using dispersion models for particulate matter (PM10) and nitrogen oxides (NOX). Cohort-specific hazard ratios were estimated for time-weighted air pollution exposure during different time windows and the incidence of stroke, adjusted for common risk factors, and then meta-analysed. Overall, 868 subjects suffered a non-fatal or fatal stroke during 238,731 person-years of follow-up. An increment of 20 μg/m(3) in estimated annual mean of road-traffic related NOX exposure at recruitment was associated with a hazard ratio of 1.16 (95% CI 0.83-1.61), with evidence of heterogeneity between the cohorts. For PM10, an increment of 10 μg/m(3) corresponded to a hazard ratio of 1.14 (95% CI 0.68-1.90). Time-window analyses did not reveal any clear induction-latency pattern. In conclusion, we found suggestive evidence of an association between long-term exposure to NOX and PM10 from local traffic and stroke at comparatively low levels of air pollution.