No. 645] ALCOHOL AND WHITE BATS 299 



as in the preliminary training, whereas the controls mod- 

 ify their behavior as soon as the change is made in the 

 procedure. 



TABLE I 



corresponding ones for the controls. 



In view of the above, the direct comparison of the av- 

 erages of the tests and controls in regular training would 

 lead to error unless the average performance in the pre- 

 liminary training happened to be the same for both sets. 

 In the long run this would undoubtedly be the case, but, 

 as it happens, the averages for the tests and controls do 

 not agree in the preliminary training. However, it was 

 found that this difference depended upon the rats with 

 strong right- or left-hand tendencies, for if these (those 

 tests and controls with more than 12 or less than 3 right- 

 end choices in the preliminary training) be omitted, the 

 average of all the rest "of the rat^ was the same for the 

 tests and controls. Using the rats wliose ])reliminary rec- 

 ords showed between 3 and 12 inchisive riglit-end choices, 

 the averages for the curves in F'm'. 3 ^v*■l•<' ol»tniiio 1. Start- 

 ing with the same average t^Midciirs \o ciitcr tlic right-end 

 compartment in the preliminar\ iraiuin-, tlie controls 



