Classical genetics depends upon investigation of function by random destruction with little information on structure. Modern mapping using random polymorphisms, cloning and sequencing investigates structure without function. The genome projects with their rapid gene discovery are, however, redefining classical genetic approaches. The efficient translation of this wealth of new information into insights in biological function at molecular, cellular and organismal levels requires large-scale approaches to the generation of mutants. Gene trapping in embryonic stem (ES) cells allows an efficient approach to the functional analysis of the murine genome. The usually separate processes of gene discovery, mapping, the observation of the expression pattern and the mutant phenotype in vivo, can be integrated by the use of an indexed library of insertionally mutated ES cell clones. It should be possible to generate mutants for a large proportion of the genes of the mammalian genome.