THE TPIREE-LEVEL SYNTHESIS 



each level interacting with the others. The simplest example of this 

 interaction of levels is seen in the principle that the chromosomes 

 are controlled by the genotype. They themselves are as near to the 

 genotype as materials can be, yet, at the cell-level, they are in fact 

 part of the phenotype: they are subject to the reaction of heredity 

 and environment. In consequence the heredity which controls 

 evolution is itself subject to evolution. Its mechanism varies, is 

 selected, and is adapted, sometimes for the short-term advantage of 

 immediate progeny, as with apomixis, and sometimes for the long- 

 term advantage of an indefinite posterity, as in the evolution of 

 crossing-over. 



There is an abiding conflict in the operation of selection between 

 long- and short-term advantages. It is a conflict which is represented 

 by the opposite advantages of the fitness of the individual and the 

 flexibility of the population. This again is a two-level aflair. What 

 fitness is for the fleeting individual, flexibility is for the enduring 

 population. Two-level equilibria can never be stable and the see-saw 

 of fitness and flexibility is probably one of the major factors of 

 instability in the races, stocks, and strains of all living communities. 



The Three-Level Synthesis 



Our three-level analysis is made possible only by a rigorous appli- 

 cation of the principle of determinacy. From the determinacy of 

 genetics proceeds its rigour and coherence as a basis of prediction. 

 This assumption of determinacy is itself justifiable by prior 

 consideration as well as by practical success. 



In passing from one level to another we change the laws of 

 behaviour. We also regulate at a higher level the disorderliness of 

 the lower level. Apparent indeterminacy is reduced to determinacy, 

 as Schrodinger has pointed out, when the gene is incorporated in 

 the chromosome. But it is also similarly reduced when the cell is 

 incorporated in the body and again when the body of the apparently 

 indeterminate individual is incorporated in the population. 



It would be idle to pretend that the application of prediction in 

 a many-levelled system is a simple one, or that its principles are well 

 understood. But it is clear that they demand a reconstruction of 

 single-level notions. In the course of this book some examples will 



371 



