Natural selection should be studied as an end in itself, and this requires rigorous experimental tests of theoretical models linking molecular phenotypes to differences in fitness. We describe the experimental verification of one such model and thereby demonstrate that the causal relations between genotype and fitness need not be as hopelessly complex as many have assumed. The model uses metabolic control theory to link enzyme activity to metabolic flux and then assumes that fitness is proportional to flux. The model was tested using the pathway for the uptake and metabolism of growth-rate-limiting concentrations of lactose in E. coli inhabiting chemostats. Many of the properties expected of natural selection are manifest in this system.