An effective approach to engineering new enzymes is to direct their evolution in vitro. By mimicking key processes of Darwinian evolution in the rest tube, the functions of enzymes can be explored free from the constraints of function within a living system. Efficient strategies for engineering new enzymes by multiple generations of random mutagenesis and recombination coupled with screening for improved variants have been developed. Our results with industrially important biocatalysts underscore the advantages of this ’evolutionary’ approach to protein engineering. (From a talk presented at the first National Academy of Engineering "Frontiers of Engineering" Symposium, 21 September 1995.)