STUDIES OF CEREBKAL FUNCTION. IX 29 



tion of cases is not normal either with respect to extent of 

 lesion or learning scores, and since the slight increase in the 

 magnitude of the measure of association given by the cor- 

 relation ratio would be meaningless at present, we have not 

 computed these constants. 



Comparison of table 3 with tables 13 and 14 shows that for 

 operated animals the intermaze correlations are significantly 

 higher than the correlations between extent of lesion and the 

 criteria of learning. Since the intermaze correlations for 

 normal animals are very low, the intermaze correlations for 

 operated animals must be ascribed to some effect of the lesion, 

 and it appears that the effective agent in cerebral lesion is 

 more accurately measured by maze performance than by 

 measurement of the surface area of the lesion. This may be 

 due either to the failure of our methods of measurement of 

 the lesions to express all of the significant characteristics of 

 the lesion or to the fact that the correlation coefficient is a 

 better expression of the relationship in one case than in the 

 other. The latter possibility has certainly played some part 

 in the matter, for the intermaze relationship is rectilinear, 

 whereas the cortex-learning relationship is not. However, 

 since the distribution of lesions is not normal, no correlation 

 method can be depended upon to give a certain picture of 

 the relationship and there is no present method of finding the 

 causes of the discrepancy. 



Continuity of the mass relationship 



As a further test of the validity of our conclusions based 

 on correlations, and to determine whether the correlations 

 represent a continuous relationship or are due to the destruc- 

 tion of some critical amount of tissue, we have divided the 

 cases by class intervals of 10 per cent destruction and com- 

 puted the average practice for learning required by the ani- 

 mals in each class interval. The results of this analysis are 

 presented in table 15. The last two intervals (40 to 49, and 

 50 per cent) are based upon too few cases to have significance. 

 In the remaining four intervals there are only six inversions 



