NRG1 is one of the most researched genes associated with schizophrenia. Although three meta-analyses in this area have been published, the results have been inconclusive and even conflicting. Family based studies can be problematical due to the difficulty of synthesizing them with case-control studies. Heterogeneity is another persistently difficult problem which tends to be side-stepped in genetic studies. To deal with these points, we performed a meta-analysis of 26 published case-control and family-based association studies up to September 2008 covering 8049 cases, 8869 controls and 1515 families. The matrix of coancestry coefficient was also calculated using population genetics. Across these studies, the conclusions are as follows: Firstly, only SNP8NRG221132, 420M9-1395(0) and 478B14-848(0) showed significant association in the relatively small sample size. Secondly, we applied both Kazeem's and Lohmueller's methods for synthesizing family and case control studies and there was no statistically significant difference between the results from the two methods, suggesting that either method can be used. In addition, the association analysis of case-control studies was statistically consistent with that of family studies. Finally, the matrix of coancestry coefficient suggested obvious population stratification. The study reveals that one SNP of the NRG1 gene does not contribute significantly to schizophrenia and that population stratification is evident. In future genetic association analysis on complex psychic diseases, haplotype blocks and population structure should be given greater consideration.