in determining just how past history is stored and exploited. The examination of specific operators can be expedited by subsuming and in a single overall diagram. Plans which satisfy this diagram and retain a recognizable variant of the "reproduction according to performance" procedures in or will be called plans of type .
(In steps 6 and 7 are amalgamated and the tests in 3 are unnecessary because exactly one new structure is formed per time-step.)
The next four sections will investigate the role of generalized genetic operators in plans of type . We will see that is used basically as a pool of schemata. (Recall from chapter 4 that this means that acts as a repository for somewhere between 2l and M·2l schemata; i.e., it contains instances of this many distinct schemata.) Past history is recorded in terms of the ranking (number of instances) of each schema in , much as discussed at the end of chapter 5. From this point of view crossing-over acts to generate new instances of schemata already in the pool while simultaneously generating (instances of) new schemata (see section 6.2). In general a total of 2l schemata will be affected by each crossing-over (see Lemma 6.2.1). Inversion (section 6.3) affects the pool of schemata by changing the linkage (association) of alleles (attributes) defining various schemata. In combination with reproduction, the net effect is to increase the linkage of schemata of high rank
and 
in a single overall diagram. Plans which satisfy this diagram and retain a recognizable variant of the "reproduction according to performance" procedures in 
.
steps 6 and 7 are amalgamated and the tests in 3 are unnecessary because exactly one new structure is formed per time-step.)
is used basically as a pool of schemata. (Recall from chapter 4 that this means that