A simulation study was carried out to generate 45,000 data for genetic marker and QTL of milk protein in which recombination rates varied from 0 to 0.4. An estimation of linkage was performed by the moment method and the maximum likelihood method(ML). Newton-Raphson algorithm was used in ML. The results obtained were as follows 1. 1n case of complete linkage(r=0), the number of QTL genotype in the sub-population of marker genotype was only one. Thus means and standard deviations of QTL and marker genotypes were equal. 2. Means of marker and QTL genotypes estimated by ML had no differences. However, the variances were higher than those of actual values. Means and variances of marker and QTL genotypes estimated by the moment method had marked differences from actual values. 3. The optimum sample size with unbiased estimates of recombination rates were 1000(r≤0.1) and 1500(r≥0.1), respectively. The homozygote with the lower mean had higher skewness, while the homozygote with the higher mean had lower skewness. 4. The skewness of two homozygote marker genotypes(MM, ㎜) can give indirect advantage of determining no differences between distribution of experimental data and population. Differences in skewness between two homozygote marker genotypes in each recombination rate were increased until r=1, and in case of above r=1, differences in skewness were decreased. The main cause of these results was that the frequency of QTL genotype(AA) within marker genotype(MM) group was decreased with increasing recombination rate. 5. When recombination rates were ranged from 0 to 0.4 by 0.05, mean gene effects were 2.104, 1.400, 1.225, 1.062, 0.896, 0.735, 0.626, 0.545 and 0.470, respectively. The closer to complete linkage, the higher mean gene effects. 6. It is evident that maximum likelihood method can markedly increase the power of marker-QTL linkage analysis. The results of this study will be useful as a basis for marker-QTL linkage determination in dairy cattle.