I say “probably” because there are still many irrelevant factors which might influence the result. For instance, if you had a different teacher for each group, one group might make greater progress not because of the method but because of the teacher. This means either that one teacher should teach both groups, or that we should multiply the number of groups and the number of teachers, and have half the teachers teaching half the groups by one method, and the other half teaching by the other method. Of course here too the more we could multiply the number the better it would be. Even then there might be some reasonable question as to the validity of the experiment, for it might be that one method would tend to encourage faster progress at the beginning, but that the other would lead to greater progress in the long run. This could be determined only by carrying our experiment over a long period. And we might still have irrelevant factors, for the machines on which one group learnt to typewrite might be superior to those on which the other group learnt, and this factor would have to be eliminated in a similar way to the others. The experimental method has been well summed up by Thomson and Tait in their _Natural Philosophy_: “In all cases when a particular agent or cause is to be studied, experiments should be arranged in such a way as to lead if possible to results depending on it alone; or, if this cannot be done, they should be arranged so as to increase the effects due to the cause to be studied till these so far exceed the unavoidable concomitants, that the latter may be considered as only disturbing, not essentially modifying the effects of the principal agent.” In all experiments one must exercise ingenuity in finding other causes besides the one to be studied which may possibly influence a result, and in eliminating these. It might benefit the reader considerably if he were to think out for himself how he would apply experiment in its most thoroughgoing form to solve a given question, say the inheritance of acquired characteristics. * * * * * I have now cited enough methods to at least indicate what “thinking with method” means. To satisfy a certain human craving all of these have been named, though sometimes arbitrarily. Of course each may have to be modified to some extent to adjust it to different problems. I must repeat: there are methods numberless, and some problems will require methods all their own.