Since the controversy level α in our model affects the variance in the empirical data (α is the distance between the peaks of the two distributions, and is thus related to the variance, but the balance p is also a factor), we can hope that at as α increases one obtains qualitative properties consistent with the data: first a unimodal distribution with peak between the means of f and g, and then a local minimum near the mean of h. In fact, this is precisely what happens. The main result is the following