One of the least important of these factors is the variability of individual performance. In the beginning of the experiment each individual is more variable than at later points in the curve. This momentary variability need not be supposed to affect all the tests in the same way nor all individuals in the same direction. This fact may then tend somewhat to reduce the correlation of the preliminary trials and may in some cases materially affect the first five or ten trials. Beyond the twenty-fifth trial the variability in these tests is much reduced, and particularly so in the measures here used, which are in all cases, after the preliminary trial, the medians of five successive trials. Another factor that deserves mention is the possibility of change in the character of the tests themselves, through practice with them. It is quite probable, for example, that the opposites test comes, after many repetitions, to resemble more and more that type of process or function involved in color-naming. The responses become more and more intimately associated with the stimulus words, the suggested responses to each word become more and more limited in number and in most cases reduced to a single word for each stimulus. This state of affairs is true of color-naming at the very beginning of the experiment. As the order of the stimulus words is changed at each trial, the test may come to involve more and more the simple task of giving merely the quickest possible association of the right response, and the overcoming of inhibitions and interferences of a more or less general sort, with less and less emphasis on the element of selection. Much the same may also be true of the addition test. It is in these three tests that the increase in correlation is most marked, and the actual coefficients highest at the end of the experiment. Careful analysis of what takes place as one improves in these simple tests would no doubt yield interesting material.