>All three steps in the genetic algorhythm exploit the above-average building blocks to produce a new generation.
>Step (1), reproduction according to fitness, causes all schemata to be treated according to the heuristic based on the estimation of schema averages.  Above average schemata have more instances in the next generation, below average ones have fewer.
>In step (2), crossover generates offspring that are different from their parents, producing new combinations of the schema passed on by step (1).  Crossover sustains the increased use of shorter, above-average schemata, but may disrupt longer extant schema, particularly those not using shorter, above-average schemata as building blocks.
>Schemata not tried before may be generated by recombination of fragments when crossover disrupts extant schemata.  That is, crossover may generate new schemata even as it recombines those already present.  
>Mutation acts in step (2) to provide an insurance policy against loss of alleles, and it can also generate new schemata by altering the defining positions of extant schemata.
> Finally, in step (3), the offspring replace strings already in the population.  This process introduces a “death rate” just sufficient to keep the population at a constant size.
>These combined effects of the genetic algorhythm are summarized in mathematical form by the Schema Theorem. 
>The most important feature of a genetic algorhythm is its ability to carry on this sophisticated manipulation of building blocks by acting only on whole strings.