In the genome-wide association study (GWAS) on schizophrenia [O'Donovan et al. (2008); Nat Genet 40:1053–1055] a UK-sample of 479 cases with DSM-IV schizophrenia was genotyped in comparison to control subjects with follow up of 12 putative loci in international replication sets of approximately 15,000 cases and controls. In these cohorts and a combined bipolar and schizophrenia UK-sample, six single nucleotide polymorphisms (SNPs) supported association, with the strongest evidence for SNP-marker rs1344706 at the zinc finger ZNF804A locus on chromosome 2q32.1 (P = 1.61 × 10−7). We attempted replication of these findings in a German population of 2,154 individuals (632 with affective disorders, 937 with schizophrenia, and 585 controls), but found none of the GWAS risk alleles significantly associated with psychosis. Particularly rs1344706, initially surpassing the genome-wide significance level in an extended phenotype of schizophrenia and affective disorder, produced consistently negative results. At the ZNF804A locus estimated Odds ratios reached 1.08 (0.93–1.26 95% CI) for the schizophrenia sample and 1.04 (0.90–1.20 95% CI) for the combined set of cases with schizophrenia and affective disorder. The main limitation of our study may be the reduced power of the sample size, but our data may be useful for future meta-analysis of GWA data sets. Although GWAS have proven extraordinary successful in identifying susceptibility genes for complex genetic disorders, the hypothesis of common genetic variants in the complex group of the schizophrenic psychoses with small effect size but relatively high frequency is still put to further scrutiny. © 2011 Wiley-Liss, Inc.