Background The Alzheimer's Disease Assessment Scale (ADAS) is often used in international multicenter trials. Use across countries presupposes correct translation and adaptation of the scale, and maintenance of its psychometric properties. Objectives To compare the various translations of the ADAS used in Western Europe, to design internationally harmonized translations and to validate these. Setting International cooperative study in eight European countries. Methods An inventory was made of existing versions of the ADAS-Cog used in eight European countries, and adaptations were made. The concurrent validity of the harmonized versions of the ADAS was tested in 283 patients with probable or possible Alzheimer's disease. The Nurses Observation Scale for Geriatrics (NOSGER), CAMCOG-R and MMSE was used to assess concordance between cognitive and behavioral measures. Results Differences between the versions mainly involved object naming, items for verbal memory, such as the number of trials allowed, the imagery value of the words selected as targets or distractors, and the number of parallel versions. These differences were eliminated by adapting and harmonizing the various versions of the ADAS-Cog. Thereafter, only small differences between the different countries were found, and patterns of correlation between ADAS-Cog, and the NOSGER, CAMCOG-R and MMSE were consistent. Conclusions The study underlines the need to use harmonized versions of instruments for rating dementia in multinational studies. The findings indicate that the harmonization of the ADAS-Cog was successful.