Two distinct mechanisms can be envisioned for resistance of human immunodeficiency virus type 1 (HIV-1) reverse transcriptase (RT) to nucleoside analogs: one in which the mutations interfere with the ability of HIV-1 RT to incorporate the analog, and the other in which the mutations enhance the excision of the analog after it has been incorporated. It has been clear for some time that there are mutations that selectively interfere with the incorporation of nucleoside analogs; however, it has only recently been proposed that zidovudine (AZT) resistance can involve the excision of the nucleoside analog after it has been incorporated into viral DNA. Although this proposal resolves some important issues, it leaves some questions unanswered. In particular, how do the AZT resistance mutations enhance excision, and what mechanism(s) causes the excision reaction to be relatively specific for AZT? We have used both structural and biochemical data to develop a model. In this model, several of the mutations associated with AZT resistance act primarily to enhance the binding of ATP, which is the most likely pyrophosphate donor in the in vivo excision reaction. The AZT resistance mutations serve to increase the affinity of RT for ATP so that, at physiological ATP concentrations, excision is reasonably efficient. So far as we can determine, the specificity of the excision reaction for an AZT-terminated primer is not due to the mutations that confer resistance, but depends instead on the structure of the region around the HIV-1 RT polymerase active site and on its interactions with the azido group of AZT. Steric constraints involving the azido group cause the end of an AZT 5'-monophosphate-terminated primer to preferentially reside at the nucleotide binding site, which favors excision.