Page:Memory (1913).djvu/52
From Wikisource
This page has not been proofread.
Even in the case of the first mentioned contradictory group of tests a grouping of the separate mean values harmonising with the normal one occurs if, instead of all the 92 tests being taken into consideration at once, they are divided into several parts — i.e., if tests are combined which were taken at about the same time and under about the same conditions.
The conclusion cannot be drawn from these numerical results that the mental fatigue which gradually increased during the twenty minute duration of the tests did not exert any influence.
It can only be said that the supposed influence of the latter upon the numbers is far outweighed by another tendency which
 |
An image should appear at this position in the text, see the image guidelines If you are able to provide it, see Help:Adding images for uploading instructions. |
- ↑ If it should ever become a matter of interest the attempt might be made to define numerically the different effects of that tendency in different cases. For the probable errors of observation for the numerical values of series-groups afford a measure for the influence of accidental disturbances to which the memorisation is daily exposed. If now the learning of the separate series in general were exposed to the same of similar variations of condition as occur from test to test, then according to the fundamental principles of the theory of errors, a probable error of observation calculated directly from the spearate values would relate itself to the one just mentioned as 1 to √n, where “n” denotes the number of separate series combined into a test. If however, as is the case here, special influences assert themselves during the memorisation of these separate series, and if such influences tend to separate the values further than other variations of conditions would do, the “P.E.0“ calculated from the separate values must turn out somewhat too great, and the just mentioned proportion consequently too small, and the stronger the influences are, the more must this be the case.
An examination of the actual relations is, to be sure, a little difficult, but fully confirms the statements. In the 84 tests, consisting of six series of 16 syllables each, the √n = 2.45. We found 48.4 to be the probable error of observation of the 84 tests. The probable error of the 504 separate values is 31.6. The quotient 31.6: 48.4 is 1.53; therefore not quite 2/3 of the value of √n.
