Success and efficiency of Darwinian evolution is based on the dichotomy of genotype and phenotype: The former is the object under variation whereas the latter constitutes the target of selection. Genotype-phenotype relations are highly complex and hence variation and selection appear uncorrelated. Population genetics visualizes evolutionary dynamics as a process among genotypes. Phenotypes are represented only through empirical parameters. The quasispecies concept introduces the molecular mechanism of mutation. Optimization is seen as a process in genotype space. Populations optimize through adaptive walks. Selective neutrality leads to random drift. Understanding evolution will be always incomplete unless phenotypes are considered explicitly. At the current state of the art, almost all genotype-phenotype mappings are too complex to be analyzed and modeled. Only the most simple case of an evolutionary process, the optimization of RNA molecules in vitro, where genotypes and phenotypes are RNA sequences and structures, respectively, can be treated successfully. We derive a model based on a stochastic process which includes unfolding of genotypes to form phenotypes as well as their evaluation. Relations between genotypes and phenotypes are handled as mappings from sequence space onto the space of molecular structures. Generic properties of this map are analyzed for RNA secondary structures. Optimization of molecular properties in populations in modeled in silico through replication and mutation in a flow reactor. The approach towards a predefined structure is monitored and reconstructed in terms of an uninterrupted series of phenotypes from initial structure to target, called relay series. We give a novel definition of continuity in evolution which identifies discontinuities as major changes in molecular phenotypes.