294 TEE AMERICAN NATURALIST [Vol.LVI 



first pair of curves, on no day in the second pair, on four 

 days in the third pair and one day in the fourth pair. 

 All the significant differences favor the controls. How- 

 ever, more important than the significance of individual 

 MAZE 

 TIME PER TRIAL 



differences as measured by the probable errors, is the 

 agreement in the direction of the differences on succes- 

 sive days. The fact that the differences on eight suc- 

 cessive days lie in the same direction probably has more 

 significance than that half of these taken separately may 

 be significant as judged by their probable errors. Con- 

 sidering the signs alone, in all the curves there are three 

 out of the 32 points of comparison showing the test av- 

 erages lower than the controls. One of these cases is 

 on the third day of training of the untreated rats from 

 treated parents, tlie other two cases are on the second and 

 third days of training of the treated rats from treated 

 parents. If chance alone is working, the probability of 



