This study aimed to investigate the association between shift work and incident dementia in two population-based cohorts from the Swedish Twin Registry (STR). The STR-1973 sample included 13,283 participants born 1926-1943 who received a mailed questionnaire in 1973 that asked about status (ever/never) and duration (years) of shift work employment. The Screening Across the Lifespan Twin (SALT) sample included 41,199 participants born 1900-1958 who participated in a telephone interview in 1998-2002 that asked about night work status and duration. Dementia diagnoses came from Swedish patient registers. Cox proportional-hazards regression was used to estimate hazard ratios (HR) with 95% confidence intervals (CI). Potential confounders such as age, sex, education, diabetes, cardiovascular disease and stroke were included in adjusted models. In genotyped subsamples (n = 2977 in STR-1973; n = 10,366 in SALT), APOE epsilon 4 status was considered in models. A total of 983 (7.4%) and 1979 (4.8%) dementia cases were identified after a median of 41.2 and 14.1 years follow-up in the STR-1973 and SALT sample, respectively. Ever shift work (HR 1.36, 95% CI 1.15-1.60) and night work (HR 1.12, 95% CI 1.01-1.23) were associated with higher dementia incidence. Modest dose-response associations were observed, where longer duration shift work and night work predicted increased dementia risk. Among APOE epsilon 4 carriers, individuals exposed to 20 years of shift work and night work had increased dementia risk compared to day workers. Findings indicate that shift work, including night shift work, compared to non-shift jobs is associated with increased dementia incidence. Confirmation of findings is needed.