On the basis of the earlier work of Goss and Harris, Cox et al. introduced radiation hybrid (RH) mapping, a somatic cell genetic technique for constructing fine-structure maps of human chromosomes. Radiation hybrid mapping uses X-ray breakage of chromosomes to order a set of genetic loci and to estimate distances between them. To analyze RH mapping data Cox et al. derived statistical methods that employ information on sets of two and four loci, to build an overall locus order. Here we describe alternative nonparametric and maximum-likelihood methods for the analysis of RHs that use information on many loci simultaneously, including information on partially typed hybrids. Combination of these multipoint methods provides a statistically more efficient solution to the locus-ordering problem. We illustrate our approach by applying it to RH mapping data on 14 markers in 99 radiation hybrids for the proximal long arm of human chromosome 21.