a single group, or test, is naturally indeterminate. It must be expected, however, that as the number increases, the correspondence between the distribution of the times actually found and those calculated in accordance with the law of errors will be greater. In practice the attempt will be made to increase the number to such a point that further increase and the closer correspondence resulting will no longer compensate for the time required. If the number of the series in a given test is lessened, the desired correspondence will also presumably decrease. However, it is desirable that even then the approximation to the theoretically demanded distribution remain perceptible.
Even this requirement is fulfilled by the numerical values obtained. In the two largest series of tests just described, I have examined the varying length of time necessary for the memorisation of the first half of each test. In the older series, these are the periods required by each 4 series of syllables, in the more recent series the periods required by each 3 of them taken together. The results are as follows:
1. In the former series: mean value (m) = 533 (P.E.o) = ± 51.
| Within the limits |
i.e. within the deviation |
Number of deviations | Of these deviations there occur | ||||||
| By actual count |
Calculated from theory |
Below | Above | ||||||
| 1⁄10 | P.E. | ± | 5 | 2 | 5 | .0 | 2 | 0 | |
| ⅙ | P.E. | ± | 8 | 4 | 8 | .2 | 3 | 1 | |
| ¼ | P.E. | ± | 12 | 6 | 12 | .3 | 4 | 2 | |
| ½ | P.E. | ± | 25 | 21 | 24 | .3 | 9 | 12 | |
| P.E. | ± | 51 | 48 | 46 | .0 | 24 | 24 | ||
| 1 | ½ | P.E. | ± | 76 | 61 | 63 | .4 | 30 | 31 |
| 2 | P.E. | ± | 102 | 76 | 75 | .6 | 37 | 39 | |
| 2 | ½ | P.E. | ± | 127 | 85 | 83 | .6 | 42 | 43 |
| 3 | P.E. | ± | 153 | 89 | 88 | .0 | 45 | 44 | |
