Haploid parthenogenesis in facultatively apomictic Pilosella generated polyhaploid progeny (with half the maternal chromosome set) both in natural populations and garden experiments. Production of polyhaploids varied considerably among different species, hybridogenous species and hybrids. In the field (14 localities), the highest frequency of polyhaploids exceeded 80% of the total seed progeny produced by some recent hybrids. A similar diversity in the production of polyhaploids was also recorded in garden experiments. A two-step process by which new genotypes of both P. aurantiaca (tetraploid) and P. rubra (hexaploid) were formed under garden conditions during a polyploid–polyhaploid–polyploid cycle is described. In the first step, the maternal plants generated dihaploid and trihaploid F1 progeny, respectively. Although a substantive part of this polyhaploid progeny was either non-viable or sterile, the apomictic polyhaploids occasionally doubled their genome. Consequently, the F2 progeny resulting from the second step had a double ploidy level, identical to that of the original maternal parent. The complete process was autonomous, without contribution of pollen from parent genotype. This cycle necessarily implicates increasing homozygosity in F2 progeny compared to the original maternal polyploid plant. The probabilities of particular steps of this process occurring in Pilosella and the variation in polyhaploids are estimated and described, and the ability of polyhaploid plants to survive under field conditions discussed. Probability of the complete cycle (haploid parthenogenesis followed by doubling of the genome), which occurred under garden conditions in P. rubra, is estimated to be in the order of hundredths of percent. Despite this low probability, it can result in the production of new homozygous genotypes in populations of apomicts, especially in those occurring in disturbed habitats with little competition.