68 K. S. LASHLEY 



" P. L." a f g i: c d: b e h j 



1st P. R a g f i: c d: b e h j 



The order is changed in this case only between the groups f 

 and g, and the difference between them is not great enough to 

 be significant in either case. There is essential agreement in 

 the results obtained by the two criteria. 



As will be noted in the table and from the coefficients of 

 regression, the difference between the groups is greatest when 

 measured by the difficult standard of three perfect trials. 1 Are 

 these differences more significant on this account? At first sight 

 it might seem so. The number of animals considered remains 

 constant and hence, other things being equal, the ratio of the 

 difference to its probable error increases. But the probable 

 errors are dependent also upon the amount of variability and a 

 further analysis of the data shows that the coefficient of varia- 

 tion remains constant or is even increased when the more diffi- 

 cult standard is used. The figures in table 2, which are taken 

 from groups c and d, illustrate this. The probability that 

 the first difference in the table (3.54) is due merely to chance 

 is about 1/3; that the second (4.12) is due to chance is 1/1 or 

 greater. A glance at the probable errors for the averages of all 

 the rats (page 70) shows that these are quite consistent with 

 the results for the smaller groups. 2 The coefficient of variation 

 in the number of trials preceeding the first errorless run is .5900, 

 for those preceeding " perfect learning " is .6107 and the prob- 

 able error of the average of the latter is proportionately greater 

 than that of the former . If two such groups were compared by 

 the two criteria the differences obtained would obviously bear 

 the same relation to their probable errors as do those in the 

 smaller groups. 



The general results of this analysis point to the following 

 conclusions : 1 . Where there is a difference in the average capacity 

 of two groups of animals for habit-formation, the more difficult 

 the problem that they are required to learn the greater will be 

 the apparent difference between the groups in the practice re- 



1 Some exceptions occur, but this is to be expected from the small number of 

 animals included in the groups. 



2 No great importance could be ascribed to this fact alone as it does not follow 

 that there is any correlation between the variability within the subordinate groups 

 and the variability of all the animals taken together, but the fact that the same 

 results are obtained for both the small and large groups does seem significant. 



