【Objective】 This study was conducted to investigate influences of sample size and spatial distribution on prediction of soil mapping, which is contributive to formulating soil sampling strategies scientifically and improving soil prediction accuracy effectively. 【Method】 Out of 5 403 soil samples, a validation dataset and training sub-datasets different in number of soil samples were derived randomly, and each subset encompassed five examples different in sampling site spatial distribution pattern. Influences of sample size and spatial distribution on predictive soil mapping, embodied by spatial distribution characteristic and prediction accuracy, were explored on the basis of the prediction of organic matter content (OMC) in topsoil layer with the Ordinary Kriging (OK）or Inverse Distance Weighted (IDW) interpolation method. 【Result】 Results show that when the number of soil samples decreased from 5 000 to 39 the OMC predictive maps based on OK or IDW interpolation was gradually losing details of local variation, and when the number dropped down to 20 or 10, the predictive maps became distorted. When the number varied in the range of 5 000~1 250, the predictive maps based on OK interpolation were quite similar in accuracy with r varying in the range of 0.55～0.59, and RMSE in the range of 3.03～3.15, but when the number dropped down to 625, the predictive maps based on OK interpolation declined significantly in accuracy, and varied sharply between the five groups different in spatial distribution pattern of sampling sites even in the same training subdset. The predictive map based on IDW interpolation varied in accuracy with the number and the distribution pattern of soil sampling sites on a trend similar to that based on OK interpolation, except that the predictive map based on IDW interpolation declined significantly in accuracy with the number of samples starting to drop from 1875 and varied sharply with spatial distribution pattern of the sampling sites. On average, the predictive maps based on OK interpolation were obviously higher than those based on IDW interpolation in accuracy before the map began to turn distorted. 【Conclusion】 All the findings in this study indicate that both sample size and spatial distribution pattern have certain impacts on predictive soil map, and the impacts are quite limited when the sample size is large enough. However, when the sample size drops below a certain level, the predictive maps will lose details of some local variations and prediction accuracy as well, while spatial distribution of sampling sites turns up to be the main affecting factor. Compared with OK, IDW is lower in accuracy and responds to the changes in sample size and spatial distribution earlier before the predictive map is distorted.