In a case-control study, we investigated the possible etiological relevance to dementia of psychosocial network factors, such as marital status, confidants and close relatives, sports activities, cultural activities, club membership; and education. In 23 general practices we recruited 195 patients with dementia. Of these, 108 were suffering from possible Alzheimer's disease, 59 from possible vascular dementia and 28 from secondary or unclassified dementia. A total of 229 control subjects was recruited: 122 population controls and 107 dementia-free ambulatory patients. Data were gathered in a structured personal interview and analyzed using logistic regression, to control for age, region, sex, dementia in parents, education and smoking. There were significantly decreased odds ratios for the number of confidants, sports activities, and cultural activities at age 30, at age 50 and at 10 years before data collection. When all psychosocial network factors were included simultaneously in the logistic regression model, these factors remained statistically significant, indicating independent effects. Restriction of the analysis to cases with possible Alzheimer's disease or to cases with possible vascular dementia led to similar results. Adjustment for the psychosocial network neutralized the otherwise protective effect of education for dementia of any type and for possible vascular dementia. In keeping with the results from recently published studies, these results support a protective role for the psychosocial network-especially for the number of confidants and for sports and cultural activities-in the etiology of dementia.