Ah, okay. I see where you're coming from. And that might be meaningful, but I still have issues ... 2 subjects in each group dropped out so really you should exclude them in your calculation (particularly given that 2 participants is 20% of the PS group and only 11% of the MPI group, so the proportions are biased due to unequal N). I am curious why they kept the 5 from the original study then randomly assigned 12 to MPI and 10 to PS. Why not 11 and 11, or 9MPI, 13PS (then you have a total of 14MPI and 13PS). It's just bizarre to me. And I think what is being missed is that for 9/25 of the subjects neither MPI or PS worked. These 9 were unable to meet the fluency requirements of their respective programs. I think that what this highlights is that there is not going to be one treatment that is superior to others for all people. If we borrow from counseling psychology literature the particular type of treatment provided doesn't matter that much to long-term outcomes. Zebrowski has argued the same phenomenon is true in stuttering therapy. It may very well be the case that MPI is superior to other forms of therapy. But there definitely needs to be more replications of this study against other forms of therapy as well before such a strong statement can be made. Think of is as an interesting Phase I clinical trial. It's interesting, for sure.