The doubled haploids open up the prospect of fixing heterosis by combining genes, which has  not   yet  been  reached  for  maize.  The  advantage  of  doubled  haploids  is  the  1:1 segregation,  whereby the frequency of polygenic combinations increases comparing to F2.  It  is  not  necessary   to  try  to  find  the  desired  genotype  immediately  in  the  first generation. The stepwise  breeding is possible, accumulating more and more favorable alleles in each cycle. The success is  possible only with dominant nature of heterosis, and not over-dominant. Therefore, to begin with,  the average degree of dominance in the loci controlling maize grain yield was estimated. Evaluation  was made in the doubled haploid populations derived from Rf7×Ku123 and MK01×A619 hybrids using the  NC III design, but the calculation method was modified to eliminate linkage bias. A total of 9  estimations were obtained, including those recalculated from literary sources. They equaled 0.53–  0.99, on average, 0.73, which confirms the dominant nature of heterosis in maize and suggests that  breeding for productivity of inbred lines per se may be successful. Three breeding cycles were performed in the Rf7×Ku123 combination and two cycles in the MK01×A619  combination. In the first cycles, doubled haploids were obtained directly from F1, and the  following cycles were carried out in two ways: 1) in the previous cycle, the  most  productive   line  was  selected  and  crossed  with  F1,  and  in  progeny  doubled haploids  were  produced;   2)  in  the  previous  cycle,  the  most  productive  backcross  of doubled haploid with the P1 or  P2 was selected, and doubled haploids were produced from it. Both methods showed about the same  efficiency. In 2017, doubled haploid lines of all cycles from Rf7×Ku123 were tested, the best parent (Rf7)  showed first ear productivity 103.9 g, the best line of first cycle - 112.5 g, the best line of  second cycle –121.2 g, the best line of third cycle – 135.6 g. The last line exceeded by 30% the  productivity of the best parent and was almost reached to F2  value (138.8 g). The average breeding  progress was 10% per cycle. In another combination MK01×A619, two  cycles  of  breeding  did  not   lead  to  significant  progress  due  to  low  efficiency  of colchicine doubling. The results showed that stepwise doubled haploid breeding can be effective even with a small volume  of each cycle (20–50 lines), but the doubling efficiency differs for different materials.