In a recombinant inbred line (RIL) population of indica rice, two subpopulations composed of susceptible lines were selected for mapping of partial resistance to leaf blast with two isolates of the pathogen. A third subpopulation composed of susceptible lines with similar heading time was used for mapping of partial resistance to neck blast with a third isolate. The traits measured for partial resistance included diseased leaf area (DLA), lesion size (LS) and lesion number (LN) for leaf blast and lesion length (LL) and conidium amount (CA) for neck blast. A linkage map consisting of 168 DNA markers was constructed by using the whole RIL population. Quantitative trait loci (QTLs) conditioning these traits were determined at one-locus and two-locus levels. Eleven main-effect QTLs and 28 digenic interactions were detected by QTLMapper 1.01 b. Only three QTLs showing main effects were also involved in digenic interactions for the same trait. General contributions of epistatic QTLs of each trait ranged from 16.0% to 51.7%, while those of main-effect QTLs of each trait ranged from 4.7% to 38.8%. The general contributions of main-effect QTLs of most traits were smaller than those of epistatic QTLs, confirming the importance of epistasis as the genetic basis for complex traits. The general contributions of the main and epistatic effects of all QTLs detected for the two traits LL and CA of the partial resistance to neck blast reached 70.6% and 82.6% respectively, which obviously represented a major part of the genetic basis controlling partial resistance to neck blast. The results indicated the necessity for partial resistance mapping to use susceptible subpopulations where the interference of major resistance genes is avoided.