The presence of nuclear androgen binding proteins measured by nuclear androgen exchange in rat spermatids arid spermatozoa was re-examined. Specific binding was observed to be related to less dense contaminating particles when sonicated testes were fractionated by isopcynic centrifugation through a 40–61% Nycodenz linear gradient. No specific binding was observed in a pure preparation of epididymal spermatozoa collected by retrograde perfusion of the cauda epididymidis, even when a nuclear exchange assay of superior sensitivity [17] was used. Contamination could easily be induced by adding prostatic tissue to epididymal spermatozoa prior to sonication. Despite this strong evidence that the measure of nuclear androgen receptors by nuclear exchange in germ cells is artefactual, the persistence of high endogenous concentrations of 5α-dihydrotestosterone and 5α-diols (but not testosterone) in spermatozoa of castrated rats argues for the opposite conclusion.