26 RUTLEDGE T. WILTBANK 



at work as in the case of groups 1 and 2. Of the 239 errors 

 made by the group, 137 were made in the first blind alley, 

 or 56.9 per cent of the total number; and the errors per- 

 sisted there as with the other groups. In the case of this 

 group, however, the tendencies were not quite sufficient 

 to turn the tide of positive transfer. This reversal of such 

 justifiable expectations as those above might be due to 

 an increase in the number of deviations from the rule of 

 turning to sniff at the door as the rats learned an increasing 

 number of mazes. That there was an increase in these 

 exceptions cannot be asserted, for no record of the number 

 of sniffings was kept; but it would seem probable that the 

 futility of such behavior would lead to its becoming less 

 frequent. It would not take many deviations of this 

 sort from the usual behavior to account for the difference 

 between the 3.58 per cent of negative transfer shown 

 by group 1 and the 4.33 per cent of positive transfer shown 

 by group 5. 



The negative instances dealt with above indicate cer- 

 tainly that, when three mazes are learned in succession, 

 the transfer is not a function of the last two mazes alone, 

 but of all three; and they probably indicate that, when 

 four or five are learned in succession, the transfer is not a 

 function of the last two alone, but of all the mazes involved. 

 They also prove — it needs hardly to be remarked — that, 

 although transfer is generally positive, there may be a 

 particular kinaesthetic habit occurring in a particular maze- 

 situation so as to produce negative transfer. 



The Question as to Whether the Transfer is Cumulative 



The experiment demonstrates that the transfer as between 

 the first and second mazes of a series is positive, and that 

 the transfer as between any other two contiguous mazes 

 of the series is also positive, judging by all the three criteria 

 employed, except in the two negative results in trials. 

 The query which next presents itself is whether this gen- 

 erally persistent positive transfer increases with each addi- 

 tional maze-learning, or in other words whether it is 

 cumulative. 



There was no series in which the transfer as estimated by 



