In rabbits, 1112 diploid (2N) spermatozoa were recognized by their large heads which were found to have an area in optical projection of 1.532 ( 0.034 SE) times the area of those deemed to be normal haploids (iN). This ratio is of the same order as the theoretical figure of 1.587 for a ratio of quadratic dimensions involving objects expected to differ twofold in volume. 2N spermatozoan heads are stockier in shape and more pear-shaped than those of haploids. A frequent partial or complete doubling of the tail (37% of 2N spermatozoa) indicates some prior cytological doubling event consistent with diploid status. Partial tail doubling is nearly always proximal, in the midpiece segment, and arguments relating to centriolar function can be developed. The heads of 2N spermatozoa can bear one, two, or (very exceptionally) three separate tails. Motile specimens have been observed of all types of 2N spermatozoa with partial or complete tail doubling, or with single tails. Twentyseven percent of 2N spermatozoa are “live” (unstained by nigrosin-eosin) and bear one tail. A further 19% are “live” but with partial or complete tail doubling. A survey of eight potential sources of variation in AS-strain rabbits aged 29 weeks showed that the incidence of 2N spermatozoa (mean 0.31%) varied between brothers but otherwise exhibited little variation in relation to kinship, nor between duplicate ejaculates or duplicate miscroscope preparations, and the heritability estimate in this partially inbred strain was zero. In a search for material with a higher incidence of 2N spermatozoa, three strains (AS, AD and R) were sampled from 26 to 192 weeks of age and marked strain and age effects were apparent, with a prominent strain/age interaction such that young males of AD-strain had the unusually high incidence of ca. 1.5% 2N spermatozoa. These young AD-strain males provide the highest known natural incidence of 2N spermatozoa. Colchicine injection failed to increase the incidence. When whole semen of young AD-strain males (average incidence 1.6% 2N spermatozoa in controls) was centrifuged in a dextran-based density gradient, an upper fraction with 0.4% 2N spermatozoa was recovered from the tubes, and a lower fraction with 2.9%. The two figures differ significantly from each other and from the control figure for whole semen. The technique was highly repeatable and constitutes a success in physical separation of living spermatozoa in accordance with their genetic content. Improvement in the degree of separation is under investigation. The fraction with a high incidence provides a new source of 2N spermatozoa for further experiments. In the fraction with a low incidence, genetically deleterious spermatozoa have been removed from mammalian semen by artificial means, and the results can be viewed as a pilot experiment of interest in a medical context.