This paper examines the utility of restriction fragment length polymorphisms (RFLPs) for paternity analysis. While, on the average, 99% of falsely accused males can be excluded with the standard battery of blood group antigens, red cell enzymes, serum proteins, and HLA antigens, there are still mother-child pairs for whom the exclusion probability is not high. It has been suggested that additional resolution would be available with RFLPs. We have examined the strategic aspects of using RFLPs for paternity analysis, comparing the efficacy and cost of a multimarker haplotypic set with those of a comparable set of unlinked RFLPs, using published frequencies for the beta-globin complex, the serum albumin region, and the growth hormone region. There are four major findings. (1) Greater resolution is obtained with a carefully chosen set of tightly linked RFLPs producing chromosomal haplotypes than with a comparable set (same allele frequencies) of unlinked markers, but only if it is possible to establish linkage phase unambiguously. (2) Assay of linked sets is cheaper than is the assay of unlinked markers, but the cost advantage is optimized with sets of no more than two or three linked markers. (3) Also, with more than two or three tightly linked markers, the haplotypic frequencies are too poorly estimated to provide a reliable measure of the probability of paternity for unexcluded males, given the sample sizes likely to be available in the near future. (4) Optimal resolution, minimal cost, and acceptable accuracy are obtained with several independent sets of no more than two or three tightly linked RFLP markers each. With current technology, RFLP analysis is more expensive for the same level of genetic resolution than is the standard battery, but gradual replacement of the latter can be anticipated as economies of scale reduce the cost of the DNA technology.