I have already mentioned as the probable cause of this unsymmetrical distribution the peculiar variations in the effect of high degrees of concentration of attention and distraction. It would naturally be supposed that the position of the separate series within each test is the cause of the repeated piling up of values on each side of the average. If, in the case of a large test-series, the values are summed up for the first, the second, and third series, etc., and the average of each is taken, these average values vary greatly, as might be expected. The separate values are grouped about their mean with only tolerable approximation to the law of error, but yet they are, on the whole, distributed most densely in its region, and these separate regions of dense distribution must of course appear in the total result.
The following may be added by way of supplement: on account of the mental fatigue which increases gradually during the course of a test-series the mean values ought to increase with the number of the series; but this does not prove to be the case.
Only in one case have I been able to notice anything corresponding to this hypothesis, namely, in the large and therefore important series of 92 tests consisting of eight series of 13 syllables each. In this case the mean values for the learning of the 92 first series, the 92 second series, etc., were found to be 105, 140, 142, 146, 148, 144, 140 seconds, the relative lengths of which Fig 2 exhibits. For all the rest of the cases which I investigated the typical fact is, on the contrary, rather such a course of the numbers as was true in the case of the
