In response to reported schizophrenia linkage findings on chromosomes 3, 6 and 8, fourteen research groups genotyped 14 microsatellite markers in an unbiased, collaborative (New) sample of 403-567 informative pedigrees per marker, and in the Original sample which produced each finding (the Johns Hopkins University sample of 46-52 informative pedigrees for chromosomes 3 and 8, and the Medical College of Virginia sample of 156-191 informative pedigrees for chromosome 6). Primary planned analyses (New sample) were two-point heterogeneity lod score (lod2) tests (dominant and recessive affected-only models), and multipoint affected sibling pair (ASP) analysis, with a narrow diagnostic model (DSM-IIIR schizophrenia and schizoaffective disorders). Regions with positive results were also analyzed in the Original and Combined samples. There was no evidence for linkage on chromosome 3. For chromosome 6, ASP maximum lod scores (MLS) were 2.19 (New sample, nominal p = 0.001) and 2.68 (Combined sample, p = .0004). For chromosome 8, maximum lod2 scores (tests of linkage with heterogeneity) were 2.22 (New sample, p = .0014) and 3.06 (Combined sample, p = .00018). Results are interpreted as inconclusive but suggestive of linkage in the latter two regions. We discuss possible reasons for failing to achieve a conclusive result in this large sample. Design issues and limitations of this type of collaborative study are discussed, and it is concluded that multicenter follow-up linkage studies of complex disorders can help to direct research efforts toward promising regions.