The soil sampling method based on representativeness grades of sampling points has advantages at cost, applicability and mapping accuracy. However, it also needs to be researched, for example, sensitivity. In order to explore this method deeply, this study first used it to analyze representativeness grades of existing samples in a hilly area, and then studied relationships between representativeness grades of the samples and mapping accuracy. Results showed that: 1) representative environment chains with high representativeness grades existed widely, while chains with low grades were fewer, 2) representativeness grades of environment chains indicated variations of soil-forming environments in different degrees, 3) sampling results selected using this sampling method were much different from those using regular griding and purposive sampling methods, 4) mapping accuracy generally increased as more and more samples of lower representativeness grades were used for mapping, although the increases decreased during this process. These results illustrated that this sampling method is advantageous at application, cost and accuracy, and could be widely used. It is worth mentioning that taking samples of certain representativeness grade levels is necessary to avoid the decrease of mapping accuracy with more samples added for mapping. Besides, due to few good soil covariates in other topographic and geomorphological types, application of this method into such areas is limited to some extent.