>Second: The second reason that crossover’s disruption of longer schemata is not so troubling stems from the observation that more complicated schemata are typically formed from combinations of shorter, well-established schemata.
>More complicated building blocks are usually formed by combining simpler building blocks.  This fact reflects our earlier observation that technical innovations such as the internal combustion engine tend to involve a particular (re)combination of relatively simpler widely used building blocks.  Moreover, devices such as the internal combustion engine become the centerpiece for a wide range of still more complex devices.
>The result is a kind of hierarchy wherein building blocks at one level of complexity are combined to form the building blocks at the next higher level of complexity.
>Under a genetic algorhythm a similar hierarchy forms wherein the higher-level (longer) schemata are typically composed of well-tested above average schemata from the lower level of complexity.
>This hierarchy ameliorates the disruptive effects of crossover.  First of all, under a genetic algorhythm, above-average schemata come to occupy a large proportion of the population, because of above-average replication in step 1 (Reproduction according to fitness).
>Consider two parent strings that contain identical copies of the same schema.  Crossover cannot disrupt the schema, even if the crossing over takes place inside the schema’s outer limits.  The alleles exchanged will be replaced by identical alleles from the opposite chromosome.
>It follows that crossover rarely disrupts longer schema composed of particular combinations of shorter, above average schemata.  If some of these longer schemata are in turn above average, they will spread through the population.
>The hierarchy becomes more elaborate, providing for the persistence of still longer schemata.  A hierarchy of disruption-resistant schemata emerges, similar to the way default hierarchies emerge.