D. O. HEBB 39 



ganglia, by cutting oft the head — the anterior four and a halt segments. 

 The animal continued to respond correctly, showing that there were 

 suiTicient synaptic modifications in the remaining ganglia to mediate the 

 response — until the new head regenerated, at which time the habit was 

 lost. The noise generated by the new ganglia, the irrelevant neural activity 

 of the uneducated brain, was sufficient to disrupt learning completely. 



In this case we are dealing with the effects of noise on an established 

 habit, and in a lower animal. It seems clear that the potential disrupting 

 effects must be as great or greater while learning is still going on, and in 

 the large brain of the higher animal. If then the rate of learning reflects the 

 noise level in the system, it becomes intelligible that the higher animal 

 does not learn faster than the lower, when each is given a task to which it 

 is constitutionally adapted. The rate of learning per sc is not an index of 

 intelligence or level in the phylctic scale (Lashley, 1929); we may note, 

 for example, the occurrence of one-trial learning, by inspection only, in 

 the solitary wasp (Baercnds, and Tinbcrgen and Kruyt, cited by Tinber- 

 gen, 195 1). There is of course no reason why modifications of the individual 

 synapse should be made more quickly in a system with many synapses 

 than in one with few. When receptors, effectors and intervening neural 

 structures are adapted to a small number of acquired responses, and there 

 is little or no irrelevant concurrent activity, the necessary synaptic 

 modification may occur very rapidly. Much that is considered to be 

 innate, because there is little evidence of practice, may in fact depend on 

 immediate learning at first exposure to the stimulating conditions. 



The problem of excess activity or noise in the larger brain is character- 

 istic of those regions of divergent (rather than parallel) conduction in 

 which learning is supposed to occur. If the experimental stimulation is to 

 result in a modification of function at the appropriate synapses, the 

 excitation produced by the stimulus must be conducted to those particular 

 synapses. From the sensory surface to the cortical projection area there is 

 no difficulty; here conduction is in parallel and the same population of 

 units can be reliably excited, time after time, since the units which begin 

 together (and thus are simultaneously excited) end together, and reinforce 

 one another's action at the next synaptic level (Hebb, 1958). But this 

 condition does not hold for the further cortical transmission, especially 

 where the mass ratio of association to sensory cortex is large (the A/S ratio: 

 Hebb, 1949). A relatively small sensory projection area cannot dominate a 

 large region of divergent conduction, maintaining on successive trials the 

 same conditions of transmission with respect to single units. This applies 

 to the large thalamo-cortical sectors which comprise the so-called 



