When I approached the question in the Harvard psychological laboratory, it was clear to me that the degree of attention and carefulness which the court may presuppose on the part of the customer can never be determined by the psychologist and his experimental methods. It would be meaningless, if we tried to discover by experiments a particular degree of similarity which every one ought to recognize or a particular degree of attention which would be sufficient for protection against fraud. Such degrees must always remain dependent upon arbitrary decision. They are not settled by natural conditions, but are entirely dependent upon social agreement. A decision outside of the realm of psychology must fix upon a particular degree in the scale of various similarity values as the limit which is not to be passed. The aim of the psychologist can be only to construct such a scale by which decisions may be made comparable and by which standards may become possible. The experiment cannot deduce from the study of mental phenomena what degrees of similarity ought to be still admissible, but it may be able to develop methods by which different degrees of similarity can be discriminated and by which a certain similarity value once selected can always be found again with objective certainty. After many fruitless efforts I settled on the following form of experiments, which I hope may bring us nearer to the attainment of the purpose. A group of objects is observed for a definite time and after a definite interval another group of objects is offered for comparison. This second group is identical with the first in all but one of the objects, and this is replaced by a similar one. The question is how often this substitution will be noticed by the observers. I may give in detail a characterization of the set of experiments in which we are at present engaged. We are working with picture postal cards, using many hundred cards of different kinds, but for each one we have one or several similar cards. As postal cards are generally manufactured in sets, it is not difficult to purchase pairs of pictures with any degree of similarity. Two cards with Christmas trees, or two with Easter eggs, or two with football players, or two with forest landscapes, and so on, may differ all the way from a slight variation of color or a hardly noticeable change in the position of details to variations which keep the same motive or the same general arrangement, but after all make the card strikingly different. The first step is to determine for each pair the degree of similarity, on a percentage basis. To overcome mere arbitrariness, we ask thirty to forty educated persons to express the similarity value, calling identical postal cards 100 per cent and two postal cards as different as a colored flower piece and a black picture of a street scene O. The average value of these judgments is then considered as expressing the objective degree of similarity between the two pictures of a pair.