118 



CARNEGIE INSTITUTION OF WASHINGTON. 



small, the probable errors are naturally very large; indeed the differences, 

 large as they seem, are more than three times their probable errors on only a 

 few days. These days are at the end of the training period (sixth, seventh, 

 and eighth days) and the last (fom'th) day of the retention test. The dif- 

 ferences between the averages on the earlier daj^s are much greater, but the 

 variability at this time is so much greater that the significance is much 

 reduced (see table 5). 



"The criterion of perfect trials fully supports the findings of the time data. 

 The tests made fewer perfect trials, required more trials before making a 

 perfect trial, and the average of the tim.e spent by the tests in running perfect 

 trials is gieater than the average for the controls, yet this difference is not 

 great enough to be significant. Plus signs indicate that the tests took more 

 time. 



*ExcIuding one abnormal rat. 



*Tf the hyi^othesis suggested as an explanation of the different results given 

 by the two generations compared above is at all correct, one would expect to 

 find in this case that both the first and later parts of the training gave real 

 differences, since there would be opportunity for both sorts of effect to be 

 manifest — the immediate somatic effect from the mother, and the germinal 

 effect from the grandparents upon the germ-cells of the parents. And the 

 table shows that the differences in the first days are greater than those dif- 

 ferences shown above (table 2) in the early days of the training of the first 

 generation, but the significance of the differences resembles the findings in 

 this first generation rather than those in the following generation, namely, 

 the last part of the training gives significant differences and the first part does 

 not. With such a small number of animals this result has no decisive sig- 

 nificance, but as far as it indicates anything it goes in the direction that the 

 suggested hypothesis demands." 



"Results of the multiple choice training. — The apparatus used in this 

 training consisted of a series of nine compartments, each of which had front 

 and back doors, operated at a distance by the observer; different sets of these 

 front doors were opened at different trials and the rat was given its reward of 



