Within the unit group itself the values of the mean variation show here, as in the preceding forms, a progressive increase in sensitiveness to temporal variations from first to last of the component intervals.  The proportional values for the four intervals in order are, 1.000, 0.786, 0.771, 0.666.  The distribution of these relative values, however, is not uniform for all four rhythmical forms, but falls into two separate types in dependence on the position of the accents as initial or final, following the discrimination already made.  The figures for the four forms separately are as follows: 

TABLE LXXVIII.

  Stress. 1st Interval. 2d Interval. 3d Interval. 4th Interval.

Initial,    9.57 per cent.   5.53 per cent.  5.83 per cent.  6.57 per cent. 
Secondary, 13.23    "       10.60    "      12.93    "       9.50    "
Tertiary,   9.00    "        8.70    "       2.00    "       4.90    "
Final,     11.45    "        9.00    "      12.60    "       7.85    "

In the first type (Rhythms I. and III.) appear a descending curve followed by an ascending; in the second type (Rhythms II. and IV.) a second descending curve follows the first.  The changes in the first type are not cooerdinated with a similar curve of variation in the intensive magnitude of the beats.  It is to be noted here that the smallest mean variation presented in this whole set of results is found in that element of the first form which receives the stress, an exception to the general rule.  The variations in the contrasted type have their maxima at those points on which the group initiation—­ primary or secondary—­falls, namely, the first and third.

As in preceding rhythmical forms, while the separation of accentual stress from primacy in the series tends to increase the mean variation of that element on which this stress falls and to raise the index of mean variation for the whole group, yet the mean variation of the initial element is also raised, and to a still greater degree, reinforcing the evidence that primacy of position is a more important factor of instability than the introduction of accentual stress.

In the investigation of mean variations for units (if we may call them such) of more than four beats only a modicum of material has been worked up, since the types of relation already discovered are of too definite a character to leave any doubt as to their significance in the expression of rhythm.  The results of these further experiments confirm the conclusions of the earlier experiments at every point.

These higher series were treated in two ways.  In the first the reactor beat out a rhythm consisting in the simple succession of groups of reactions, each of which contained one and only one accent.  These units in each case were marked by initial stress, and were composed of five, six, seven, eight and ten beats respectively.  The results are given in the following table, which contains the series of mean variations in duration both for single intervals and for total groups.