D. O. HEBB 41 



THE NATURE OF THE TRACE IN SHORT-TERM MEMORY 



In principle, we may distinguish two ways in which a memory trace 

 can be estabhshed (cf. e.g. Eccles, 1953). One is a kind of after-discharge, a 

 reverberatory activity set up without necessarily depending on any 

 change in the units involved, other than the discharge of impulses; and 

 one consists of some change in the units which outlasts their period of 

 activity. The first can be called for convenience an actii'iry trace, the second 

 a stniaiiral trace. 



In discussing this point earlier (Hebb, 1949) I assumed that the repetition 

 of digits in a test of memory span provides a pure example oi the activity 

 trace. Here the experimenter presents verbally a series of digits, and the 

 subject is asked to reproduce them in the same order. After the subject has 

 attempted one series, the experimenter presents a second series and the 

 subject seems to forget the preceding series completely. He docs not get 

 the two mixed up just as, in a calculating machine, punching a second set 

 of numbers wipes out the preceding set completely. But is this what 

 happens? Is there no lasting after-effect: no structural change produced by 

 hearing and repeating the digits? 



With this question in mind, and with the intention of learning more 

 about the nature of the trace, in short-term memories for highly familiar 

 material, the following experiments were carried out. 



Method: the subjects were college students, each tested individually. 

 They were informed that the purpose of the experiment was to see 

 whether the memory span for digits would improve with practice. 

 Twenty-four series of digits were presented. On each of these trials, the 

 experimenter read aloud a series of nine digits at the rate of about one per 

 second. The subject was instructed to listen carefully and repeat the digits 

 in exactly the same order. 



Each series consisted of the digits from i to 9, in varying order, each 

 digit occurring once only. An example is 



591437826. 



There was, however, one special feature about which the subject was not 

 informed. On every third trial (3rd, 6th, 9th ... 24th) the same series was 

 repeated, and the object of the experiment was to discover what effect this 

 had. Would the repetition result in learning? If immediate memory for 

 digits in these circumstances is mediated solely by an activity trace, with 

 no structural changes, each new stimulus-event would be expected to 

 wipe the slate clean and set up a new pattern of activities. No cumulative 

 learning would occur. If such learning occurs, however, we may conclude 



