114 



CARNEGIE INSTITUTION OF WASHINGTON. 



have been given. The chi-square (x^) test has been apphed to the data on 

 time and distance, to discover how unlikely it is that such differences as are 

 found between the test and control data are due to chance, instead of to the 

 treatment of the grandparents. In making the averages the following 

 groupings of rats have been used (in each of these the tests are always com- 

 pared with the corresponding controls) : strains separately with sexes sepa- 

 rately; strains together with sexes separately; strains separately with sexes 

 together; strains together with sexes together. In this way the unsettled 

 questions of the genetic differences between strains and sexes may be elimi- 

 nated. Briefly, these various groupings of the data show that the general 

 conclusions are independent of whatever strain and sex differences may exist. 

 "In table 1 are given extracts from the results, with strains and sexes 

 together. The greatest interest lies in the number of times the differences 

 between the averages of the tests and controls exceeds the probable errors of 

 these differences. When the quotient of the error into the difference is 3 or 

 higher the difference may be considered statistically significant, that is, the 

 probabiUty is at most only 1 in 20 that such a difference will be due to chance. 

 Plus differences indicate that the tests took more time or covered more dis- 

 tance than the controls. D/P.E. signifies the quotient of the difference 

 divided by the probable error of the difference. 



Table 1. — Time (in seconds) and distance averages of the grandchildren of alcoholized rats 

 (tests, 25 rats) compared with the grandchildren of the non-alcoholized grandparents (con- 

 trols 26 rats) on each day of the training in the maze. 



''The time, distance, and errors criteria agree in showing the following: In 

 all cases the averages indicate a greater capacity for learning the maze by the 

 controls; in the majority of cases these differences are great enough to be 

 considered statistically significant. The exceptions tend to appear in the last 

 part of training and in the retention test. It is suspected that the maze was 

 simple enough to be learned by all the rats, but that the alcoholism of the 

 grandparents of the tests tended to make them slower in learning it; in the 

 first part of the training, the differences are great enough to be significant, 

 while in the last part many of the rats do not continue to improve and the 

 differences in the averages are reduced, but they still He in the same direction. 

 However, the results subsequently obtained for the preceding (the filial) 

 generation conflict with this explanation, since there the first days of training 

 do not show any clear inferiority of the tests, but in the latter days the differ- 

 ences are statistically significant, favoring the controls. 



