Heritable variation is at the heart of the process of evolution. However, variation is restricted in breeding for uniform crop populations using the pedigree line approach. Pedigree lines are successful in agriculture because synthetic inputs are used to raise fertility and control weeds, pests and diseases. An alternative method promoted for exploring the value of variation and evolutionary fitness in crops is to create composite cross populations. Composite cross populations are formed by assembling seed stocks with diverse evolutionary origins, recombination of these stocks by hybridization, the bulking of F1 progeny, and subsequent natural selection for mass sorting of the progeny in successive natural cropping environments. Composite cross populations can provide dynamic gene pools, which in turn provide a means of conserving germplasm resources: they can also allow selection of heterogeneous crop varieties. The value of composite cross populations in achieving these aims is dependent on the outcome of mass trials by artificial and natural selection acting upon the heterogeneous mixture. There is evidence to suggest that composite cross populations may be an efficient way of providing heterogeneous crops and of selecting superior pure lines for low input systems characterized by unpredictable stress conditions.