On trying to sort them according to Purkenje's standards, I failed completely, and many analogous plans were attempted without success. Next I endeavoured to sort the patterns into groups so that the central pattern of each group should differ by a unit of "equally discernible difference" from the central patterns of the adjacent groups, proposing to adopt those central patterns as standards of reference. After tedious re-sortings, some sixty standards were provisionally selected, and the whole laid by for a few days. On returning to the work with a fresh mind, it was painful to find how greatly my judgment had changed in the interim, and how faulty a classification that seemed tolerably good a week before, looked then. Moreover, I suffered the shame and humiliation of discovering that the identity of certain duplicates had been overlooked, and that one print had been mistaken for another. Repeated trials of the same kind made it certain that finality would never be reached by the path hitherto pursued. On considering the causes of these doubts and blunders, different influences were found to produce them, any one of which was sufficient by itself to give rise to serious uncertainty. A complex pattern is capable of suggesting various readings, as the figuring on a wall-paper may suggest a variety of forms and faces to those who have such fancies. The number of illusive renderings of prints taken from the same finger, is greatly increased by such trifles as the relative breadths of their respective lineations and the differences in their depths of tint. The ridges themselves are soft in substance, and of various heights, so that a small difference in the pressure applied, or in the quantity of ink used, may considerably affect the width of the lines and the darkness of portions of the print. Certain ridges may thereby catch the attention at one time, though not at others, and give a bias to some false conception of the pattern. Again, it seldom happens that different impressions of the same digit are printed from exactly the same part of it, consequently the portion of the pattern that supplies the dominant character will often be quite different in the two prints. Hence the eye is apt to be deceived when it is guided merely by the general appearance. A third cause of error is still more serious; it is that patterns, especially those of a spiral form, may be apparently similar, yet fundamentally unlike, the unaided eye being frequently unable to analyse them and to discern real differences. Besides all this, the judgment is distracted by the mere size of the pattern, which catches the attention at once, and by other secondary matters such as the number of turns in the whorled patterns, and the relative dimensions of their different parts. The first need to be satisfied, before it could become possible to base the classification upon a more sure foundation than that of general appearance, was to establish a