The authors have considered introducing a molecular breeding technique called ‘genomewide selection (GwS)’ to effectively improve the yield of maize for whole-crop silage use. The GwS for this purpose requires training populations (TPs) where individual plants having been developed from three-way crosses are examined in their phenotypes as well as in molecular-marker genotypes. The purpose of this study was to evaluate the (broad-sense) heritability of maturity- and yield-related traits in such TPs, because a former simulation study by the authors showed that the heritability should exceed 25% for the success of the GwS. Five preliminary TPs were grown to estimate the heritability of silking date, ear dry weight and culm length from 2011 to 2013 on two planting density levels, the higher of which was for productivity testing and the lower for accurate individual phenotyping. The results indicated that heritability exceeded 25% in all cases, and that the higher planting density did not result in lower heritability. Accordingly, it was concluded that the maize GwS to improve yield is sufficiently feasible in terms of maintaining high heritability in the TPs, and that planting density for TPs should be that for productivity testing because it is known to have remarkable interaction on yield with genotypes.