Using 'Galax urceolata' and 'Chamerion angustifolium ', I studied the evolutionary processes governing the origin and maintenance of autopolyploidy by examining the genetic relationships between diploid and tetraploids and the relative fitness differences and mating relationships among diploids, triploids and tetraploids. In 'Galax urceolata', cytotype frequencies differed significantly, with diploids most frequent in the north-east and least frequent in the south-west part of the range. Uniform populations were either diploid (81%) or tetraploid (19%), but never triploid. Overall, populations are predominantly diploid or tetraploid but rarely evenly mixed. RAPDs were used to investigate the genetic diversity and geographical relationships among four diploid and three tetraploid populations of ' G. urceolata'. Populations in close geographical proximity were genetically less similar than more distant populations. Patterns of genetic variation and geographical distribution support the role of environmental heterogeneity in cytotype structure. Diploids, tetraploids and hybrid triploids of 'Chamerion angustifolium ' were created and evaluated at six life stages to study relative fitness differences. Diploid offspring (from 2'x' * 2' x' crosses) had significantly higher seed production and lower biomass than tetraploid offspring (from 4'x' * 4'x' crosses). Relative to diploids, the cumulative fitness of tetraploids was 0.67. Generally, triploids (from 2'x' * 4'x', 4'x' * 2'x' crosses) had significantly lower seed production, lower pollen viability, and higher biomass than diploid individuals. Collectively, tetraploids exhibit an inherent fitness disadvantage, although partial triploid viability and fertility may help to reduce the barrier to tetraploid establishment. To quantify the possible mating relationships between cytotypes and the ploidy of their offspring, I cross-pollinated diploid, triploid and tetraploid individuals. Seed set differed significantly among cross types, with 2' x' * 2'x' crosses having the greatest seed set (0.71 ± 0.04). Several crosses that involved triploid parents resulted in little or no seed set. The mean seed set in diploid-tetraploid crosses was 2.06 (SE ± 0.16) with no differences between reciprocal crosses. Mean DNA content of the progeny differed significantly among cross types. The results indicated that tetraploids were produced through several pathways of mating including triploids; however, the recurrent production of tetraploids was insufficient to overcome the frequency-dependent mating disadvantage in diploid populations.