Samples of brain from 15 prospectively tested women over 75 years of age, living in the same institution and with a Blessed test score between 0 and 28 were studied to evaluate seven different staining techniques for counting senile plaques and neurofibrillary tangles. The methods included Bielschowsky (modified by Yamamoto & Hirano, 1986), Cross, Gallyas, Naoumenko-Feigin, silver methenamine, Bodian coupled with luxol fast blue, and thioflavine S; these techniques were performed on contiguous slides of the first temporal gyrus (Brodmann's area 22). The cost, difficulty and variability of the methods were evaluated. Modified Bielschowsky method revealed both amyloid and neurites. Cross, Bodian and Gallyas stained neurites preferentially and were more sensitive for neurofibrillary tangles than for senile plaques. Silver methenamine revealed amyloid in much the same way as thioflavine S. The highest count of senile plaques and neurofibrillary tangles was obtained with the modified Bielschowsky method. Best correlations between density of senile plaques and Blessed test score were obtained with modified Bielschowsky impregnation, silver methenamine, and Bodian and those for neurofibrillary tangles, with modified Bielschowsky, Naoumenko-Feigin, and Gallyas methods. Naoumenko-Feigin impregnation, even if fitted to the linear model, revealed very few changes (four times less than modified Bielschowsky) and appeared to be insensitive. All the techniques except Naoumenko-Feigin, were equally able to distinguish two groups of cases, i.e. those affected by Alzheimer's disease and those not affected. A multiple regression analysis indicated that the statistical relationship between the density of changes, evaluated after the modified Bielschowsky method, and the Blessed test score could not be improved by the additional use of other staining techniques. These data have direct implications for the evaluation of diagnostic criteria in Alzheimer's dementia. In this study, modified Bielschowsky impregnation was found to give the most complete picture of the lesions and, as such, could be considered a reference method. This technique is, however, expensive and difficult to perform. It stains a large number of normal structures and the recognition of the lesions is more subjective than with other selective stains. It is therefore difficult to recommend this technique for routine use. Other techniques alone, or in combination, can be recommended but the diagnostic criteria should be adapted for each of them.