By adding thirty-one markers in the pre- vious linkage map, a new genetic linkage map con- taining 205 markers was constructed, spanning a total of 2305.4 cM with an average interval of 11.2 cM. The genotypic errors in the whole genome were de- tected by the statistical method and removed manu- ally. The precision of the linkage map was improved significantly. Main and epistatic QTL were detected by R/qtl, and main QTL were confirmed and refined by multiple interval mapping (MIM). Finally, MIM de- tected seven QTL for rows number, and five QTL for each grain yield, kernels per row and 100-kernel weight. The contribution to genetic variations of QTL varied from 35.3% for grain yield to 61.5% for rows number. Only kernels per row exhibited significant epistatic interactions between QTL. Twenty-four epistatic QTL were detected which distributed on almost all the ten chromosomes. About two-third epistatic QTL were observed between main QTL and another locus, which had no significant effects. These results indicate rather clearly that there are a number of QTL affecting trait expressions, not directly but indirectly through interactions with other loci. Thus, epistatic QTL effects may play a crucial role, if not more important than main QTL effects, in the genetic variation for the measured traits in present study. By adding thirty-one markers in the pre- vious linkage map, a new genetic linkage map con- taining 205 markers was constructed, spanning a total of 2305.4 cM with an average interval of 11.2 cM. The genotypic errors in the whole genome were - tected by the statistical method and removed manu- ally. The precision of the linkage map was improved significantly. Main and epistatic QTLs were detected by R / qtl, and main QTLs were confirmed and refined by multiple interval mapping (MIM). Finally, MIM de- tected seven QTL for rows number, and five QTL for each grain yield, kernels per row and 100-kernel weight. The contribution to genetic variations of QTL varied from 35.3% for grain yield to 61.5% for rows number. Only kernels per row exhibited significant epistatic interactions between QTL. Twenty-four epistatic QTL were detected which distributed on almost all the ten chromosomes. About two-third epistatic QTL were observed between main QTL and another locus, which had no significant effect These results indicate rather clearly that there are a number of QTL affecting trait expressions, not directly but indirectly through interactions with other loci. Thus, epistatic QTL effects may play a crucial role, if not more important than major QTL effects, in the genetic variation for the measured traits in present study.