A selective procedure is described for the isolation of deletions in phage T4. This is based on the properties of partial diploids which possess a tandem duplication covering the rII region. The starting-point of the procedure is a particular s1231/3157 diploid which has a high segregation frequency and so possesses a long duplication. The replacement in this diploid of the large deletion 3157 by the small one 196 further increases the length of this duplication and removes the terminal redundancy of the resulting phage which are then non-viable. New compensating deletions which restore the terminal redundancy and thus the viability to these diploids are thereby selected. Nine new independent T4 deletions have been isolated by this procedure. The length of three has been estimated by terminal redundancy measurements and each found to be appreciably longer than the rII region. While developing the isolation procedure a key diploid intermediate was identified with the property that is was completely stable; that is, unlike typical diploids which continually generate haploid segregants during phage multiplications, this diploid strain produced no viable haploid segregants. This unexpected finding led us to examine the general problem of how tandem duplications can be stabilized. The solution we propose here is a structural one involving new stabilizing deletions which penetrate into one arm of the duplication removing certain essential genes. These stabilizing deletions are located in such a way that all haploid segregants formed by recombination are necessarily non-viable. Further investigations of the original stable diploid, and others isolated subsequently, validated this model for these phage diploids. It also led to the recognition of several novel genetic structures involving repeated DNA sequences which, together with the concept of the stabilizing deletions, could be of general significance.