The joining of DNA double-strand breaks in vivo is frequently accompanied by the loss of a few nucleotides at the junction between the interacting partners. In vitro systems mimic this loss and, on detailed analysis, have suggested two models for the mechanism of end-joining. One invokes the use of extensive homologous side-by-side alignment of the partners prior to joining, while the other proposes the use of small regions of homology located at or near the terminus of the interacting molecules. to discriminate between these two models, assays were conducted both in vitro and in vivo with specially designed substrates. In vitro, molecules with limited terminal homology were capable of joining, but analysis of the junctions suggested that the mechanism employed the limited homology available. In vivo, the substrates with no extensive homology end-joined with equal efficiency to those with extensive homology in two different topological arrangements. Taken together, these results suggest that extensive homology is not a prerequisite for efficient end-joining, but that small homologies close to the terminus are used preferentially, as predicted by the modified single-strand annealing model.