By pure endpoint diagnosis of the disease, the risk of developing schizophrenia has been repeatedly associated with specific variants of the neuregulin1 (NRG1) gene. However, the role of NRG1 in the etiology of schizophrenia has remained unclear. Since Nrg1 serves vital functions in early brain development of mice, we hypothesized that human NRG1 alleles codetermine developmentally influenced readouts of the disease: age of onset and positive symptom severity. We analyzed 1,071 comprehensively phenotyped schizophrenic/schizoaffective patients, diagnosed according to DSM-IV-TR, from the GRAS (Göttingen Research Association for Schizophrenia) Data Collection for genetic variability in the Icelandic region of risk in the NRG1 gene. For the case-control analysis part of the study, we included 1,056 healthy individuals with comparable ethnicity. The phenotype-based genetic association study (PGAS) was performed on the GRAS sample. Instead of a risk constellation, we detected that several haplotypic variants of NRG1 were, unexpectedly, less frequent in the schizophrenic than in the control sample (mean OR=0.78, range between 0.68 and 0.85). In the PGAS we found that these "protective" NRG1 variants are specifically underrepresented in subgroups of schizophrenic subjects with early age of onset and high positive symptom load. The GRAS Data Collection as a prerequisite for PGAS has enabled us to associate protective NRG1 genotypes with later onset and milder course of schizophrenia.