1418 
These results confirm the hypothesis put forward heretofore, that 
the accurate reproductions, which in M.’s case are invariably depend- 
ent upon memory, and consequently comprise a recognition of the 
reproductions, facilitate a later recognition, as appears from a shorten- 
ing of the reaction time. 
As it might be supposed, that the quickening of the reactions in 
the case of an accurate reproduction in the interval, is due to a 
deeper impression of that fragment of the figure that was exhibited 
at the second sitting, with a view to facilitate a reproduction, we 
must observe that there is no reason whatever, why it should not 
affect the more or less inaccurate reproductions, as well as the 
accurate. Moreover, our control-tests have proved that a quickened 
reaction is undoubtedly due — anyhow chietly due — to the effect 
of the accurate reproduction. This is borne out by the fact that in 
some of these tests the observer was shown, during the interval, a 
fragment of the primary stimulus, while he was asked to avoid as 
much as possible any representation of the specified stimulus. We 
have calculated the recognition time for 25 of such experiments 
upon each subject; we subjoin the results together with the times 
needed for a preceding accurate reproduction. 
TABLE V. (time, seconds). 
| 
‚Arithmetica!, : | Mean 
| mean Median | deviation 
Accurate reproduction in | M 12 1,2 0,22 
the interval llp 1,9 1,5 0,64 
Exhibition of a fragment ( M 1,4 155 0,32 
of the figure, without re- | | 
production in the interval |( P 3,9 3,7 | 1,88 
The experiment upon P. also shows, in the reaction times of the 
recognition with preceding inaccurate reproduction, the different 
effects of the inaccurate memory-image and of the images of imagi- 
nation upon the later recognizing process. We wish to call partic- 
ular attention to the very considerable mean deviation: 
Arithm. av. 4.7; median 4.5; mean deviation 2.35. 
The average time needed for recognition is almost the same for 
fairly accurate and for inaccurate reproductions preceding the recog- 
nition, whereas the mean deviation is considerably larger here. This 
tallies with our hypothesis that the calculated times constitute the 
average of two very distinct groups of data, whose differences, of 
