Restriction fragment length polymorphism (RFLP) linkage maps have been constructed in several major diploid crops. However, construction of RFLP maps directly in polyploids has lagged behind for several reasons: (1) there are a large number of possible genotypes for each DNA probe expected in a segregating population, and these genotypes cannot always be identified readily by their banding phenotypes; and (2) the genome constitutions (allopolyploidy versus autopolyploidy) in many high polyploids are not clearly understood. We present here an analysis of these problems and propose a general method for mapping polyploids based on segregation of single-dose restriction fragments (SDRFS). SDRFs segregate 1:1 (presence: absence) in gametes of heterozygous plants. Hypothetical allopolyploid and autopolyploid species with four ploidy levels of 2n = 4x, 6x, 8x, and 10x, are used to illustrate the procedures for identifying SDRFs, detecting linkages among SDRFs, and distinguishing allopolyploid versus autopolyploids from polyploids of unknown genome constitution. Family size required, probability of linkage, and attributes of different mapping populations are discussed. We estimate that a population size of 75 is required to identify SDRFs with 98% level of confidence for the four ploidy levels. This population size is also adequate for detecting and estimating linkages in the coupling phase for both allopolyploids and autopolyploids, but linkages in the repulsion phase can be estimated only in allopolyploids. For autopolyploids, it is impractical to estimate meaningful linkages in repulsion because very large family sizes (>750) are required. For high-level polyploids of unknown genome constitution, the ratio between the number of detected repulsion versus coupling linkages may provide a crude measurement of preferential chromosome pairing, which can be used to distinguish allopolyploidy from autopolyploidy. To create a mapping population, one parent (P1) should have high heterozygosity to ensure a high frequency of SDRFs, and the second parent (P2) should have a low level of heterozygosity to increase the probability of detecting polymorphic fragments. This condition could be satisfied by choosing outcrossed hybrids as one parental type and inbreds, haploids, or doubled haploids as the other parental type.