OBJECTIVE: We aimed to investigate three reports of a possible role of early parental death in late onset dementia. We tested a multivariate model of risk factors for late onset dementia that included established (female sex, a family history of dementia, APOE epsilon4) and putative influences (vascular risk factors, years of full-time education, parental ages at death, and childhood IQ) on dementia risk. METHODS: We examined contributions of early life and late life risk factors for dementia by using childhood social and family data and blood samples obtained at interview at age about 78 years. In 1997-1999, we recruited 281 subjects without dementia from a 1932 Scottish IQ survey of children born in 1921 and followed them up to 2010 (at age 88). Binary logistic regression and Bayesian structural equation modelling were used to model dementia risk. RESULTS: Risk of dementia was associated with increasing age from 77 to 88 years, female sex, death of either parent before age 11 and APOE epsilon4 genotype. Family history of dementia, childhood IQ, years of education and vascular risk factors did not contribute to the model. CONCLUSIONS: Our multivariate models of the possible causes of late onset dementia confirm previous associations of dementia with female sex and APOE epsilon4 genotype and supports earlier reports of a role for early parental death.