292 WALTER S. HUNTER 



thus avoid it. There is no necessity to assume conscious as 

 opposed to automatic recognition. The fact that an animal's 

 behavior in successive trials stamps a certain pathway as untrue 

 is sufficient. 



In the present experiment when this point was reached, i. e., 

 when the curve began to fall permanently, the errors were 

 eliminated as follows: After those of returning over the true 

 pathway had gone, there followed a stage in which the errors 

 of entering blind alleys nos. i, 3 and 6 were almost the only 

 ones made. Here a curious, although not necessarily a totally 

 unpredictable fact came to light. The tendency to make errors 

 at I and 6 was stronger than at first. The pecuHar nature of 

 these cul-de-sacs is that each precedes an exactly similarly 

 situated alley which is the true one. Thus the reaction called for 

 is a control of the impulse to turn until a few more inches shall 

 have been traversed. To the writer and other observers, it ap- 

 peared that after the pigeon had learned that the required turns 

 were to the right and left respectively, it took the first oppor- 

 tunity to make such reactions rather than control the impulse 

 a little longer. Such behavior indicates the influence of a kinaes- 

 thetic factor, a visual one, or both. In most cases the error at 

 cul-de-sac no. 3 was the next eliminated and was followed in 

 order by those at nos. 6 and i. 



The data which were secured upon the question of the value 

 of previous training in the subsequent learning of a new maze 

 are as follows: Pigeons nos. 5 and 8 learned the problem 

 equally as soon as did nos. i, 2 and 7. The curves, figs. 6 and 

 7, for the two groups of birds, however, reveal marked differ- 

 ences. Neither the time nor the error graphs for pigeons 5 

 and 8 rise as high as do the respective graphs for pigeons i, 2 

 and 7. Yet the time curve for the latter group descends perma- 

 nently sooner than the corresponding curve for the former group. 

 The most probable conclusions to be drawn from these facts 

 would link up previous training and subsequent learning as 

 follows: (i) The habits acquired in labyrinth A by birds nos. 

 I, 2 and 7 interfered A\'ith their learning of B and resulted in a 

 slow elimination of errors. (2) The training in A made it pos- 

 sible for nos. I, 2 and 7 to reduce their time records permanently 

 sooner than did nos. 5 and 8. Just why previous training 

 should have had this effect on the time records the writer is 



