We have developed a general method for determining whether a gene of Escherichia coli is essential for viability. The method requires cloned DNA spanning the gene in question and a reasonably detailed genetic and physical map of the cloned segment. Using this information, one constructs a deletion of the target gene in vitro. For convenience, the deletion can be marked by an antibiotic resistance gene. A DNA segment containing the deletion is then cloned onto an att delta phage lambda vector. Integration of this phage, by homologous recombination at the target locus, and subsequent excision provide an efficient route for crossing the marked deletion onto the bacterial chromosome. Failure to delete the target gene indicates either that the resulting deletion was not viable or that the desired recombinational event did not take place. The use of prophage excision to generate the deletion allows one to estimate the fraction of deletion-producing events by analysis of the other product of the excision, the phage produced on induction of the prophage. In this way one can determine whether failure to recover a particular chromosomal deletion was due to its never having been formed, or, once formed, to its failure to survive. Applying this method to the polA gene, we found that polA is required for growth on rich medium but not on minimal medium. We repeated the experiment in the presence of plasmids carrying functional fragments of the polA gene, corresponding to the 5'-3' exonuclease and the polymerase-3'-5' exonuclease portions of DNA polymerase I. Surprisingly, either of these fragments, in the absence of the other, was sufficient to allow growth on rich medium.