Finally, the intensity of the preceding sound was increased as well as the duration of the interval separating it from the following stroke.  The measure employed was the trochaic, the interval suffering change was that following the accented beat—­in this case, therefore, the intra-group interval.  The relations obtaining among the unchanged measures were, as to duration of accented and unaccented elements, 1.000:0.714; as to intensity, 0.875:0.250 inch.  Instead of a series, as in the preceding experiments, only one change in each direction was introduced, namely, an increase in duration of a single accented element of the series from 1.000 to 1.285, and an increase of the same element in intensity from 0.875 to 1.875 inch fall.  The results are given in the annexed table: 

TABLE LX.

Duration.                            Stress. 
Position      Interval Following Louder
in Series.          Judged to be                      Increased Stress.
+             =            —        Times Noted.    Not Noted. 
I.     8 per cent.  92 per cent.  0 per cent.   40 per cent.  60 per cent
II.    42    "       50    "       8    "        42    "       58    "
III.    57    "       36    "       7    "        54    "       46    "
IV.    67    "       26    "       7    "        62    "       38    "
V.    30    "       40    "      40    "        60    "       40    "

The figures show that in regard to the discrimination of changes in duration occurring in intervals internal to the rhythm group, as well as in the case of intervals separating adjacent groups, there is a progressive increase in sensibility to variations as the succession of sounds advances.  This increased sensitiveness is here complicated with another element, the tendency to underestimate the duration of the interval following a louder sound introduced into a series.  The influence of this second factor cannot be analyzed in detail, since the amount of underestimation is not recorded unless it be sufficient to displace the sign of the interval; but if such a quantitative method be applied as has already been described, the results show a continuous decrease in the amount of underestimation of this interval from the first position to the fourth, or penultimate, which presents the following relative values:  92, 66, 50, 40.  A phase of rapid increase in the amount of underestimation appears in the fifth or final position, represented on the above scale of relative values by 120.  This falling off at the end of the series, which appeared also in previous experiments, can be attributed only to an interference with the functions which the several measures bear in the process of comparison, and indicates that the accuracy of judgment is dependent on a comparison of the measure or element in question with those which follow as well as with those which precede it.