Using Monte Carlo method, the effects of marker density on the accuracy of QTL mapping, measured with mean squared error (MSE) of QTL position estimates, were investigated for a grand-daughter design under different population structures, heritabilities, sizes of QTL effect, and QTL locations. The optimal marker densities for QTL mapping for applying to marker assisted selection, (MAS) were analyzed from an economic point of view. The results show that, in general, with the increase of marker density MSEs decrease much more rapidly when all other factors are at a high level than at a low level. When the sample size reaches a certain level (e.g. 40 families with 100 sons per family), or when the QTL effect is large (e.g. the QTL explains 50% of the genetic variance), the decrease of MSE will be only slightly affected by other factors. For the best benefits of MAS, the optimal marker distances under different combinations of various factors were 5cM (small sample and large QTL, moderate sample, heritability and QTL, or large sample and small QTL), 10cM (moderate sample and high heritability or large QTL, or large sample and moderate heritability and QTL), and 20cM (large sample and high heritability or large QTL).