17/5 DESIGN FOR A BRAIN 



As a clear illustration of such a process I quote from Lloyd 

 Morgan on the training of a falcon 



' She is trained to the lure — a dead pigeon . . . — at first with 

 the leash. Later a light string is attached to the leash, and 

 the falcon is unhooded by an assistant, while the falconer, 

 standing at a distance of five to ten yards, calls her by shout- 

 ing and casting out the lure. Gradually day after day the 

 distance is increased, till the hawk will come thirty yards or 

 so without hesitation ; then she may be trusted to fly to the 

 lure at liberty, and by degrees from any distance, say a 

 thousand yards. This accomplished, she should learn to 

 stoop to the lure. . . . This should be done at first only 

 once, and then progressively until she will stoop backwards 

 and forwards at the lure as often as desired. Next she should 

 be entered at her quarry . . . ' 



The same process has also been demonstrated more formally. 

 Wolfe and Cowles, for instance, taught chimpanzees that tokens 

 could be exchanged for fruit : the chimpanzees would then learn 

 to open problem boxes to get tokens ; but this way of getting fruit 

 (the 4 adaptive ' reaction) was learned only if the procedure for 

 the exchange of tokens had been well learned first. In other 

 words, the environment was beyond their power of adaptation 

 if presented as a complex whole — they could not get the fruit — 

 but if taken as two stages in a particular order, could be adapted to. 



4 . . . the growing child fashions day by day, year by year, a 

 complex concatenation of acquired knowledge and skills, adding 

 one unit to another in endless sequence ', said Culler. I need not 

 further emphasise the importance of serial adaptation. 



17/5. To what process in the multistable system does serial 

 adaptation correspond ? It is sullicient if we examine the 

 relation of a second adaptation to a first, for a series consists only 

 of this primary relation repeated. 



We assume then that the multistable system has learned one 

 reaction and that it is now faced with an environment that can 

 be adapted to only by the system developing some new reaction 

 that uses the old. It is convenient, for simplicity, to assume 

 here that the first reaction is no longer able to be disrupted by 

 subsequent events. The assumption demands little, for in the 

 next chapter we shall examine the contrary assumption ; and 

 there is, in fact, some evidence to suggest that, in the mammalian 



186 



