Maximum Likelihood fits of five models to the mean family size distribution F(n) in genomes from (A) 92 Bacteria and (B) 79 Archaea clustered at E
= 1e-20 and f
= 0.7. BDI3 is the best fit for the constituent genomes. As there are few families larger than n = 20, the data point at n = 20 shows the sum of all families with n ≥ 20, and the theory points at n = 20 show the sum of the predicted frequencies of all families with n ≥ 20. Hence the apparent spike in these distributions.