Av. gain in object couplets, 19 per cent. " " " movement couplets, 30 per cent. The first question which occurs in examining the foregoing tables is concerning the method of treating the indirect associations, _i.e._, obtaining the per cents. The number of couplets correctly recalled may be divided into two classes: those in which indirect associations did not occur, and those in which they did occur. Those in which they did not occur furnish us exactly what we want, for they are results which are entirely free from indirect associations. In them, therefore, a comparison can be made between series using objects and activities and others using images. On the other hand, those correctly recalled couplets in which indirect associations _did_ occur are not for our purposes pure material, for they contain not only the object-image factor but the indirect association factor also. The solution is to eliminate these latter couplets, _i.e._, subtract them both from the number correctly recalled and from the total number of couplets in the set for a given subject. By so doing and by dividing the first remainder by the second the per cents, in the tables were obtained. There is one exception to this treatment. The few couplets in which indirect associations occurred but which were nevertheless _incorrectly_ recalled are subtracted only from the total number of couplets in the set. The method by which the occurrence of indirect associations was recorded has been already described. It is considered entirely trustworthy. There is usually little doubt in the mind of a subject who comprehends what is meant by an indirect association whether or not such were present in the particular series which has just been learned. If none occurred in it the subjects always recorded the fact. That an indirect association should occasionally be present on one day and absent on a subsequent one is not strange. That a second term should effect a union between a first and third and thereafter disappear from consciousness is not an uncommon phenomenon of association. There were thirteen such cases out of sixty-eight indirect associations in the _A_, _B_ and _C_ sets. In the tables they are given as present because their effects are present. When the reverse was the case, namely, when an indirect association occurred on the second, ninth or sixteenth day for the first time, it aided in later recall and was counted thereafter. There were eight such cases among the sixty-eight indirect associations. Is it possible that the occurrence of indirect associations in, _e.g._, two of the four couplets of a series renders the retention of the other two easier? This could only be so when the intervals between two couplets in learning were used for review, but such was never the case. The subjects were required to fill such intervals with repetitions of the preceding couplet only.