Preliminary studies suggested that age at onset (AAO) may help to define homogeneous bipolar affective disorder (BPAD) subtypes. This candidate symptom approach might be useful to identify vulnerability genes. Thus, the probability of detecting major disease-causing genes might be increased by focusing on families with early-onset BPAD type I probands. This study was conducted as part of the European Collaborative Study of Early Onset BPAD (France, Germany, Ireland, Scotland, Switzerland, England, Slovenia). We performed a genome-wide search with 384 microsatellite markers using non-parametric linkage analysis in 87 sib-pairs ascertained through an early-onset BPAD type I proband (AAO of 21 years or below). Non-parametric multipoint analysis suggested eight regions of linkage with P-values<0.01 (2p21, 2q14.3, 3p14, 5q33, 7q36, 10q23, 16q23 and 20p12). The 3p14 region showed the most significant linkage (genome-wide P-value estimated over 10 000 simulated replicates of 0.015 [0.01-0.02]). After genome-wide search analysis, we performed additional linkage analyses with increased marker density using markers in four regions suggestive for linkage and having an information contents lower than 75% (3p14, 10q23, 16q23 and 20p12). For these regions, the information content improved by about 10%. In chromosome 3, the non-parametric linkage score increased from 3.51 to 3.83. This study is the first to use early-onset bipolar type I probands in an attempt to increase sample homogeneity. These preliminary findings require confirmation in independent panels of families.