The aim of the present study was to evaluate the following parameters for the grain yield of silage maize: variability of inbred lines and their diallel hybrids, superior-parent heterosis and components of genetic variability and heritability on the basis of the diallel set. The two-year four-replicate trial was set up according to the randomized complete-block design at Zemun Polje. It was determined that a genotype, year and their interaction significantly affected variability of this trait. The highest. i.e. the lowest grain yield, on the average for both investigation years. was recorded in the silage maize inbred lines ZPLB402 and ZPLB405. respectively. The analysis of components of genetic variance for grain yield shows that the additive component (D) was lower than the dominant (H1 and H2) genetic variance, while a positive component F and the frequency of dominant (u) and recessive (v) genes for this observed trait point to prevalence of dominant genes over recessive ones. Furthermore. this is confirmed by the ratio of dominant to recessive genes in parental genotypes for grain yield (Kd/Kr> 1) that is greater than unity in both years of investigation. The estimated value of the average degree of dominance (H1/D)1/2 exceeds unity, pointing out to superdominance in inheritance of this trait in both years of investigation. Results of Vr/Vr regression analysis indicate superdominance in inheritance of grain yield. Moreover. a registered presence of non-allelic interaction points out to the need to study effects of epistasis, as it can have a greater significance in certain hybrids. A greater value of dominant than additive variance resulted in high values of broad-sense heritability for grain yield in both investigation years (98.71%, i.e. 97.19% in 1997, i.e. 1998, respectively). and low values of narrow-sense heritability (11.9% in 1997 and 12.2% in 1998).