times in classes separated by 10[sigma].  Of thirty-one reactions, seven are here in the class 170[sigma].  This is the model class, and the mean gotten by taking the average of 31 reactions is 162[sigma].  If the mode had been taken to represent the usual reaction time in this case, there would have been no considerable error.  But suppose now that in the series there had occurred a reaction of 800[sigma].  Should it have been used in the determination of the mean?  If so, it would have made it almost 30[sigma] greater, thus removing it considerably from the mode.  If not, on what grounds should it be discarded?  The fact that widely varying results are gotten in any series of reactions, points, it would seem, not so much to the normal variability as to accidental differences in conditions; and the best explanation for isolated reactions available is that they are due to such disturbing factors as would decrease the strength of the stimulus or temporarily inhibit the response.  During experimentation it was possible to detect many reactions which were unsatisfactory because of some defect in the method, but occasionally when everything appeared to be all right an exceptional result was gotten.  There is the possibility of any or all such results being due to internal factors whose influence it should be one of the objects of reaction-time work to determine; but in view of the fact that there were very few of these questionable cases, and that in series I, for instance, the inclusion of two or three reactions which stood isolated by several tenths of a second from the mode would have given a mean so far from the modal condition that the results would not have been in any wise comparable with those of other series, those reactions which were entirely isolated from the mode and removed therefrom by 200[sigma] have been omitted.  In series I alone was this needful, for in the other series there was comparatively little irregularity.

The results of studies of the reaction time for the one-cell electric stimulus appear in Table XI.  The first column of this table contains the average reaction time or mean for each subject.  Nos. 2 and 4 appeared to be much less sensitive to the current than the others, and few responses to the first stimulus could be obtained.  Their time is longer than that of the others, and their variability on the whole greater.  Individual differences are very prominent in the studies thus far made on the frog.  The one-cell stimulus is so near the threshold that it is no easy matter to get a mean which is significant.  Could the conditions be as fully controlled as in human reaction time it would not be difficult, but in animal work that is impossible.  No attempt has thus far been made to get the reaction time in case of summation effects except in occasional instances, and in so far as those are available they indicate no great difference between the normal threshold reaction and the summation reaction, but on this problem more work is planned.