The molecular pathways leading to Alzheimer-type dementia are not well understood, but the amyloid beta-protein is believed to be centrally involved. The quantity of amyloid beta-protein containing plaques does not correlate well with clinical status, suggesting that if amyloid beta-protein is pathogenic it involves soluble non-plaque material. Using 43 brains from the Newcastle cohort of the population-representative Medical Research Council Cognitive Function and Ageing Study, we examined the relationship between biochemically distinct forms of amyloid beta-protein and the presence of Alzheimer-type dementia. Cortical samples were serially extracted with Tris-buffered saline, Tris-buffered saline containing 1% TX-100 and with 88% formic acid and extracts analysed for amyloid beta-protein by immunoprecipitation/western blotting. The cohort was divisible into those with dementia at death with (n = 14) or without (n = 10) significant Alzheimer-type pathology, and those who were not demented (n = 19). Amyloid beta-protein monomer in extracts produced using Tris-buffered saline and Tris-buffered saline containing 1% TX-100 were strongly associated with Alzheimer type dementia (P < 0.001) and sodium dodecyl sulphate-stable amyloid beta-protein dimer was detected specifically and sensitively in Tris-buffered saline, Tris-buffered saline containing 1% TX-100 and formic acid extracts of Alzheimer brain. Amyloid beta-protein monomer in the formic acid fraction closely correlated with diffuse and neuritic plaque burden, but was not specific for dementia. These findings support the hypothesis that soluble amyloid beta-protein is a major correlate of dementia associated with Alzheimer-type pathology and is likely to be intimately involved in the pathogenesis of cognitive failure.