Statistical significance, path dependency, and the culture of journal publication