We have genotyped 292 affected sibling pairs (ASPs) with Alzheimer's disease (AD) according to NINCDS-ADRDA diagnostic criteria and with onset ages of >/=65 years using 237 microsatellite markers separated by an average distance of 16.3 cM. Data were analysed by SPLINK and MAPMAKER/SIBS on the whole sample of 292 ASPs and subsets of 162 ASPs where both members possessed an apolipoprotein E (APOE)straightepsilon4 allele and 63 pairs where neither possessed anstraightepsilon4 allele. Sixteen peaks with a multipoint lod score (MLS) >1 either in the whole sample, the straightepsilon4-positive or -negative subgroups were observed on chromosomes 1 (two peaks), 2, 5, 6, 9 (two peaks), 10 (two peaks), 12, 13, 14, 19, 21 and X (two peaks). Simulation studies revealed that these findings exceeded those expected by chance, although many are likely to be false positives. The highest lod scores on chromosomes 1 (MLS 2.67), 9 (MLS 2.38), 10 (MLS 2.27) and 19 (MLS 1.79) fulfilLander and Kruglyak's definition of 'suggestive' in that they would be expected to occur by chance once or less per genome scan. Several other peaks were only marginally less significant than this, in particular those on chromosomes 14 (MLS 2.16), 5 (MLS 2.00), 12, close to alpha2-macroglobulin (MLS 1.91), and 21, close to amyloid precursor protein (MLS 1.77). This is the largest genome scan to date in AD and shows for the first time that this is a genetically complex disorder involving several, perhaps many, genes in addition to APOE. Moreover, our data will be of interest to those hoping to identify positional candidate genes using information emerging from neurobiological studies of AD.