Labeled antibodies, as well as their fragments and antibody-derived recombinant constructs, have long been proposed as general vectors to target radionuclides to tumor lesions for imaging and therapy. They have indeed shown promise in both imaging and therapeutic applications, but they have not fulfilled the original expectations of achieving sufficient image contrast for tumor detection or sufficient radiation dose delivered to tumors for therapy. Pretargeting was originally developed for tumor immunoscintigraphy. It was assumed that directly-radiolabled antibodies could be replaced by an unlabeled immunoconjugate capable of binding both a tumor-specific antigen and a small molecular weight molecule. The small molecular weight molecule would carry the radioactive payload and would be injected after the bispecific immunoconjugate. It has been demonstrated that this approach does allow for both antibody-specific recognition and fast clearance of the radioactive molecule, thus resulting in improved tumor-to-normal tissue contrast ratios. It was subsequently shown that pretargeting also held promise for tumor therapy, translating improved tumor-to-normal tissue contrast ratios into more specific delivery of absorbed radiation doses. Many technical approaches have been proposed to implement pretargeting, and two have been extensively documented. One is based on the avidin-biotin system, and the other on bispecific antibodies binding a tumor-specific antigen and a hapten. Both have been studied in preclinical models, as well as in several clinical studies, and have shown improved targeting efficiency. This article reviews the historical and recent preclinical and clinical advances in the use of bispecific-antibody-based pretargeting for radioimmunodetection and radioimmunotherapy of cancer. The results of recent evaluation of pretargeting in PET imaging also are discussed.