The labor of reducing the system of equations would depend on their number, which must equal that of the conditions. This would doubtless be great--possibly twenty or twenty-five, but the work amounts simply to doing a great deal of figuring. I believe that this thing is worth trying, and I intend to try it myself as soon as I can secure the necessary help in doing the work of figuring, which in any case would not be nearly as great as that done to calculate a comet’s orbit. Physicists and astronomers are daily doing work of this kind, and doing it, too, on subjects regarding which there is quite as much reason to doubt the applicability of the method as in the present case. Why not try it? It admits of satisfactory “proving,” for if applied to two groups of libraries with absurdly different results, it would at once be shown to be faulty as so applied. I believe that we librarians use the experimental method too infrequently. When it is proposed to make some change or other, I constantly hear the objection, “That wouldn’t result at all as you expect; it would do so-and-so.” But why not try it? Try it and see what happens. That is the only real test Of course, if trying will cost a large sum, or involve some serious risk, we must count the cost, but in nine cases out of ten nothing is involved but a little extra work. In this case we are trying our experiments daily--we can’t help it. We have libraries running under all kinds of conditions and we have statistical reports of those conditions and of the resulting cost. It is surely worth while to see if we can not connect these costs and these conditions in some useful way.