D. O. HEBB 43 



order (thus providing 'negative reinforcement', or 'punishment' for 

 error) ; here the record is in terms of the number of digits correctly 

 repeated on each trial. In both cases it is clear that learning occurs. 



This can be seen in another v^ay. The forty subjects v^hose results are 

 diagrammed in Fig. i were each asked, following the nineteenth trial, 

 whether they had 'noticed anything unusual' about the procedure. 

 Twenty-two reported that there had been some repetition in the series 

 presented to them, and three of these could give the crucial series without 

 error. If the subject did not volunteer anything, he was asked explicitly 

 about the repetition; three more subjects reported that they had observed 

 it, and one of them could repeat the crucial series correctly. The remaining 

 fifteen subjects had not observed the repetition. (The questioning was 

 done between trials 19 and 20 in order not to direct attention to the 

 crucial scries, occurring on trials 18 and 21; none the less, the questions 

 may have affected the subsequent performance on trials 21 and 24, which 

 can be seen in Fig. i to show a further sharp improvement.) 



The implications of this rather simple-minded experiment are more 

 extensive than may be apparent at first. With such results, I can find no 

 way of avoiding the conclusion that a single repetition of a set of digits, 

 with or without the reinforcement of being told when an error has been 

 made, produces a structural trace which can be cumulative. I assume that 

 an activity trace may also be involved in the actual repetition, but it is the 

 structural change which is of interest here. 



It is important to note that we are dealing with highly practised material. 

 Associative connections already exist between any two digits, for the 

 educated subject especially. In addition to the very highly practiced 

 sequence 1-2-3-4 ..., the learning of historical dates, telephone numbers, 

 street addresses, quantitative values such as the speed of light or the number 

 of feet in a mile, and the batting averages of the Boston Red Sox in 1937 

 — all these varied uses of the nine digits mean that the subject has already 

 learned many sequences, in one or other of which any digit is followed by 

 any other digit. When he is given a specific series to repeat, the memory 

 for that scries must depend somehow on a further strengthening of the 

 connections already established. This is diagrammed in Fig. 3, where for 

 simplicity the trace systems or cell-assemblies of three numbers only are 

 represented. 1 has a strong connection with 2, and 2 with j (because of the 

 frequency with which the sequence 1-2-3 ■•• has been repeated in the 

 past) ; but 2 has connections with / as well as with j, and j with _' and i as 

 well as with 4 (not shown). 



Now let us suppose that the subject is given the sequence 3-2-1 ... to 



