In this connection, however, we must avoid the two extremes, uniformity of punishment and the so-called individualisation of punishment, the latter especially in fashion amongst American prison experts. No doubt it would be a desirable thing to apply a particular treatment to each convict, after a physical and psychological study of his individuality, and of the conditions which led him into crime; but this is not practicable when the number of prisoners is <p 227>very great, and the managing staff have no adequate notions of criminal biology and psychology. How can a governor individualise the penal treatment of four or five hundred prisoners? And does not the cellular system, which reduces the characteristic manifestations of the personal dispositions of prisoners to a minimum, levelling them all by the uniformity of routine and silence, render it impossible to observe and get to know the special character of each condemned person, and so specialising the discipline? Where, too, are we to find the necessary governors and warders who would know how to discharge this difficult duty? The solid fact that particular houses of correction or punishment are in excellent condition when their governors have the psychological intuition of a De Metz, a Crofton, a Spagliardi, or a Roukawichnikoff, and languish when he departs, strikingly demonstrates that the whole secret of success lies in the spirit of a wise governor, skilled in psychology, rather than in the slender virtue of the cell. Just as an imperfect code with good judges succeeds better than a ``monumental'' code with foolish judges, so a prison system, however ingenious and symmetrical, is worthless without a staff to correspond. And as the question of the staff is always very serious, especially for financial reasons, I believe that, instead of the impracticable idea of individualisation in punishment, we ought to substitute that of classification, which is equally efficacious and more easily applied. It cannot be denied that criminal anthropologists are not all agreed on the classification of <p 228>criminals. But I have already shown that the differences between proposed classifications are only formal and of secondary importance; and again, the number of those who agree to the classification which I have proposed increases day by day. Before inquiring how we can practically organise the positive system of social defence on the basis of this anthropological classification of criminals, we must bear in mind two rules, common to all the technical proposals of the same system. First, care must be taken that segregation does not become or continue to be (as it is too often at present) a welcome refuge of idleness and criminal association, instead of a deprivation.