THE NEURAL BASIS OF LEARNING 



1475 



with food (in salivary conditioning) or with shock (in 

 avoidance conditioning). Responses eventually appear 

 of course, to the bell but not to the metronome. It is 

 possible to set up a still more complex discrimination 

 between two stimuli. One might shock the left leg, for 

 example, if a dog does not lift it upon hearing the 

 bell, but shock the right leg if he does not lift that one 

 upon hearing the metronome. In this case, a con- 

 ditioned differential response is learned. 



In differential conditioning procedures, the re- 

 inforced stimulus is known as the 'positive stimulus' 

 and the one not reinforced is the 'negative stimulus.' 

 On the other hand, the response the subject is trained 

 not to make is regarded as being inhibited, and when 

 the subject learns not to respond to the negative 

 stimulus, the subject is said to have developed a 

 'conditioned' or 'internal inhibition.' Furthermore 

 the response made to a positive stimulus may be 

 called an 'excitatory reflex' while the response that is 

 not made to a negative stimulus may be termed an 

 'inhibitory reflex.' These procedures and terms just 

 described are most in vogue among the Russian and 

 Pavlovian workers. 



American psychologists, on the other hand, in 

 general prefer to develop discriminations in animals 

 by using 'avoidance conditioning' and 'discrimination 

 learning 1 methods. In using 'avoidance conditioning 1 

 to set up a discrimination, some stimulus, sa) .1 tone 

 of 1000 cycles, is presented continuously .mcl the 

 animal does not respond to it. When the tone is 

 changed, however, say to 1020 cycles, the animal 

 must learn to lift its paw (or run from one end of a 

 box to another) to avoid shock. It is the change in the 

 signal to which the animal is conditioned to respond. 



In the type II situations, discrimination learning is 

 actually another type of trial-and-error learning. The 

 animal is presented with two stimuli and required to 

 choose between them. Animals can be trained to 

 make such choices by rewarding correct responses 

 with food and punishing incorrect ones with shock. 

 Many such techniques for discriminative learning, 

 however, involve only reward, not punishment. 



In the study of learning, it is often desirable to have 

 problems of such difficulty that they tap the 'higher 

 mental processes' of the animal — imaginal and ide- 

 ational processes of the kind human beings use in 

 language and thinking. For this purpose, the discrimi- 

 nation technique just described can be altered so that 

 the animal must make a 'delayed response." In this, 

 the animal is presented with two stimuli, shown 

 which one is correct and then is forced to wait for 

 some interval, usually 10 sec. or more, before being 



permitted to make a choice. The position of the 

 correct stimulus must of course be varied from trial to 

 trial or the problem is no more than one of making a 

 simple discrimination. Primates solve this kind of 

 problem much more readily than do subprimate 

 animals. The ability also depends, as we shall see, on 

 the functioning of certain neural structures. 



other methods. One of the earliest devices to be 

 employed in learning studies was the maze. Its chief 

 advantage is that it can be varied in difficulty from 

 a simple T, in which the animal makes one single 

 choice, to extremely complicated patterns entailing 

 20 or 30 choices in a series. Mazes can also be planned 

 in such a way as to constitute very ditlicult problems 

 that can be solved only by using 'higher mental 

 processes.' 



Some of the more recent techniques for studying 

 behavior are the so-called operant (Skinnerian) 

 methods. These are Type II procedures that permit 

 extremel) delicate objective measures of the learned 

 responses, the central correlates of which are our chief 

 concern; lor this reason we ma\ expect to see them 

 used increasingly - As an example of the niceties of the 

 technique, consider the following (2171. Rats can be 

 taughl i" |nr^s .1 lever to turn off a shock that would 

 otherwise be delivered to the feet. If a clock in the 

 circuit determines that the shocks will be delivered 

 every 20 see. unless the lever is pressed (in which case 

 the clock is reset to zero), the rat soon learns to space 

 its responses reasonably accurately at nearly 20-sec. 

 Intervals. This learned 'timing behavior" iv stable over 

 man) hours and from day to day. Automatic re- 

 cording of the responses frees the experimenter from 

 many unrewarding features of experimentation while 

 at the same time providing a clear objective account 

 of the ongoing behavior. The reader interested in this 

 technique will wish to consult Ferster & Skinner (60). 



Summary 



This account of methods and terms used in the 

 study of learning is necessarily quite brief. Learning 

 has been studied in literally hundreds of other ways. 

 We have selected only those terms, methods and con- 

 cepts that provide the necessary background for the 

 data to be presented on neural correlates of learning. 

 For a more detailed discussion of methods, the reader 

 should consult Hilgard's chapter in the Handbook of 

 Experimental Psychology (96) and other useful sources of 

 information (98, 170). For a treatment of the phe- 

 nomena and concepts of learning, see Deese (46), 



