The General Health Questionnaire (GHQ) is a widely used measure of non-psychotic psychiatric disorder and a valued tool in epidemiological research. Its reliability and validity are well established, but not so its factor structure. In previous studies different researchers have found very different factor solutions. In the present study this problem was solved by using the FACTOREP procedure in which a series of analyses was conducted, rotating a reduced number of factors in each, until a structure was found which was replicated across four independent subject groups. The outcome produced a new 20 item version of the GHQ with a robust, replicable and reliable four-factor structure. Moreover this same four-factor structure can be obtained from the full 60 item GHQ.