If the association test is to be useful in the study of pathological conditions, it is of great importance to have a reliable measure of the associational value of a pair of ideas. Many attempts have been made to modify and amplify the classical grouping of associations according to similarity, contrast, contiguity, and sequence, so as to make it serviceable in differentiating between normal and abnormal associations. In this study we attempted to apply Aschaffenburg's [1] classification of reactions, but without success. Our failure to utilize this system of classification is assigned to the following considerations: (1) Distinctions between associations according to logical relations are extremely difficult to define; in many cases there is room for difference of opinion as to the proper place for an association, and thus the application of a logical scheme depends largely upon the personal equation of the observer; that even experienced observers cannot, in all cases, agree in placing an association is shown by Aschaffenburg's criticisms of the opinions of other observers on this point.[2] (2) Logical distinctions do not bring out clearly the differences between the reactions of normal subjects and those of insane subjects; logically, the reaction _bath--ink_, which was given by a patient, might be placed in the class with the reaction _bath--water_, although there is an obvious difference between the two reactions. (3) Many of the reactions given by insane subjects possess no obvious logical value whatever; but since any combination of ideas may represent a relationship, either real or imagined, it would be arbitrary to characterize such a reaction as incoherent. [Footnote 1: Experimentelle Studien uber Association. Psychologische Arbeiten, Vol. I, p. 209; Vol. II, p. 1; Vol. IV, p. 235.] [Footnote 2: Loc. cit, Vol. 1, pp. 226-227.] The criterion of values which is used in this study is an empirical one. As has already been explained (p. 8), every word contained in the frequency tables possesses a value of at least 0.1 per cent, and other words have a zero value. With the aid of our method the difficulty of classifying the reactions quoted above is obviated, as it is necessary only to refer to the table to find their proper values: the value of the reaction _bath--water_ is 33.9 per cent, while that of the reaction _bath--ink_ is 0. Logically the combination _health--wealth_ may be placed in any one of four classes, as follows: / intrinsic / causal dependence health--wealth / \ coordination \ \ extrinsic / speech reminiscence \ sound similarity