18/6 DESIGN FOR A BRAIN 



cause for its own correction : step-functions change value, and that 

 particular form of interaction is destroyed. Then the step-func- 

 tions' new values provide new forms of interaction, which are again 

 tested against the environment. The process can stop when and 

 only when the step-functions have values that, acting with the 

 environment, give behaviour that keeps the essential variables 

 within normal limits. Interactions are thus as subject to the 

 requirements of ultrastability as are the other characteristics of 

 behaviour. 



Ultrastability thus works in all ways towards adaptation. 

 The only question that remains is whether it is sufficiently effective. 



18/6. Is the principle of ultrastability really sufficient to over- 

 come the tendency to chaos ? Is it really sufficient to co-ordinate 

 the activities of, say, 10 10 neurons when they interact with an 

 extremely complicated environment ? Let me admit at once that 

 the problem will require a great deal of further study before a final 

 answer can be given. The mathematical study of such systems 

 has yet hardly begun, so no rigorous proof can be given. The 

 available physiological evidence is slight, and the physiologist 

 who tries to get direct evidence will encounter formidable diffi- 

 culties. Nevertheless, we are not wholly without evidence on the 

 subject. 



Consider first the spinal reflexes. If we examine a mammal's 

 reflexes, examining them in relation to its daily life, we shall 

 usually find, not only that each individual reflex is adapted to the 

 environment but that the various reflexes are so co-ordinated in 

 their interactions that they work together harmoniously. Nor 

 is this surprising, for species whose reflexes are badly co-ordinated 

 have an obviously diminished chance of survival. The principle 

 of natural selection has thus been sufficient to produce not only 

 well-constructed reflexes but co-ordination between them. 



A second example is given by the many complex biochemical 

 processes that must be co-ordinated successfully if an organism is 

 to live. Not only must a complicated system like the Krebs' 

 cycle, involving a dozen or more reactions, be properly co-ordinated 

 within itself, but it must be properly co-ordinated into all the other 

 cycles and processes with which it may interact. Biochemists 

 have already demonstrated something of the complexity of these 

 systems and the future will undoubtedly reveal more. Yet in 



196 



