BACKGROUND: Job characteristics may influence dementia risk, but some types of job complexity remain to be examined. Twin studies provide a useful methodology to examine job differences between pairs who share many environmental and genetic influences. METHODS: Members of the NAS-NRC Twins Registry of World War II Veterans received a clinical evaluation for dementia and had job ratings from the Dictionary of Occupational Titles. RESULTS: Cotwin-control models (n = 220 pairs) indicated lower dementia risk with greater job demands of reasoning, mathematics, language, and vocational training, with comparable results in case-control models (n=425 cases). These effects were significant among twin pairs discordant for 6 or more years, but not among those discordant between 3-5 years. Results were similar for Alzheimer's disease, and main effects were not further explained by zygosity or apolipoprotein E genotype. CONCLUSIONS: Jobs that utilize data, academic skills, and extensive vocational training may protect against dementia; however, in twin pairs these effects only emerged among individuals who remained free of dementia several years after onset in their sibling.