In the course of a large twin study of Alzheimer's disease we used a two-stage telephone screening procedure. The modified Telephone Interview for Cognitive Status (TICS-m) served as an initial screen for dementia in 12709 individuals. The telephone Dementia Questionnaire (DQ) was then asked of collateral informants for subjects with TICS-m scores below 28, as well as for samples of persons with higher TICS-m scores. Based upon DQ responses, individuals with cognitive impairment not attributable to focal causes underwent assessment for the clinical diagnosis of Alzheimer's disease ('Alzheimer's dementia'), as did their twins. Well-defined Alzheimer's dementia was apparent in 39 subjects. Employing a cut-off of 27 or lower as indicative of cognitive impairment, the sensitivity of the TICS-m in the detection of Alzheimer's dementia was estimated at > 99% and specificity at 86%. Inclusion of the DQ increased the specificity at the 27/28 cut-point to 99%. The TICS-m score was associated with an area under the receiver operating characteristic (ROC) curve of 0.88 (95% confidence interval 0.81 to 0.94). The maximum number of cases of Alzheimer's dementia remaining undetected in the sample was estimated to be 34.