a set of probability distributions over , one of which is selected by each application of a stochastic operator .
assigns to each basic structure the stochastic operator which is to be used to modify A.
an arbitrary operator from W which determines, from and a selection i(t), a distribution over .
Once the set of structures has been given, along with an observation procedure which assigns a payoff µE(A) to each trial of a structure , a reproductive plan of type is determined by specifying the functions r and {w}. The algorithm proceeds as follows:
Algorithms of type are strictly sequential in the sense that one individual A'(t) is tested at a time. serves as a reservoir of information about the environment and as a basis for generating new trials. remains constant in size because each new individual A'(t) replaces an individual already in the population. Under the operators W of interest (particularly the generalized genetic operators), A'(t) can be looked upon as the "offspring" of Ai(t)(t), retaining many (but generally not all) of the attributes of Ai(t)(t). Via the function r each structure in the population carries a specification of the operator appropriate to it (a kind of "species" designation). (The apparent generalization to stochastic selection of one of a set of
, one of which is selected by each application of a stochastic operator 
.
the stochastic operator 
and a selection i(t), a distribution 
over 
is determined by specifying the functions r and {w}. The algorithm proceeds as follows: