Change in evolution is sometimes assumed to be purely unidirectional, but unidirectional change is not found in directional selection experiments, field studies, or fossil studies, and it should not be assumed in evolution. Nothing evolves in a straight line for long. Genetic models are based on generation-to-generation change and the models mislead us when long-term net rates of evolution are substituted for the generation-scale step rates required. The empirical step rates that biologists observe in field studies are much higher than any long-interval substitutes that paleontologists calculate from the fossil record. Thus the potential for random drift is also greater than most authors estimate. Random drift does not begin to equal its potential in evolution, which is evidence for the near ubiquity of natural selection. Directional and stabilizing selection can be quantified by comparing change observed in evolution to the change expected from random drift. If the observed change is greater, then the difference is attributable to directional selection. If the observed change is less, then the difference is a measure of stabilizing selection.