The balance scale task has been frequently studied in developmental psychology. The aim of the present paper was to explore the latent structure of Chinese children’s performance on the balance scale task. In the first section, I review psychological studies using the balance scale task, including a description of Siegler’s(1976) rule assessment methodology (RAM). A critical analysis is provided based on problems RAM and the four basic rules proposed by Siegler. However, RAM was frequently criticized for lacking statistical background. It can be solved by latent class analysis (LCA) by which we can test statistically how many rules are needed to fit the data and which these rules are. The basic thoughts of latent class analysis and classification were introduced briefly. A test made of 25 items which belong to four different type of balance scale task, was developed for the present research. The 457 Chinese children aged 6 years to 15 years finished the complete test and the data was analyzed through latent class approach with MPLUS 6.0. Before analysis of combinations of items, it is essential to eliminate the one who could not use RULEⅠ. Those children were found by latent class analysis of 5 weight items. After that, there are 428 people in our dataset for further analysis of combinations of items. Latent class analysis of combinations of distance, conflict-balance and conflict weight items, two of each type, showed that LCA has several advantages in classification of cognitive rules. Model of sixed classes fit the data well, in which classes correspond RULEⅠ, RULEⅠ', RULEⅡ, compensation rule, RULE Ⅳ, distance dominant rule. Chinese children aged 6 to 9 mainly use RULEⅠ, and with increase of age, less children choose RULEⅠ, RULEⅠ', RULEⅡ to solve the test items. Children aged 10 to 13 mainly use compensation rule. Almost all Children aged more than 14 use RULE Ⅳ during the testing. The increase of proportion of one who using RULE Ⅳ was sudden between age 13 and age 14. The results of the present paper can be well explained by the Overlapping Waves Model. The rules utilized by Chinese children are discussed and compared to other researcher’s findings. Some advantages and shortcomings of LCA are also discussed. One hypothesis of using LCA to investigate the cognitive rules is that subject should utilize the same strategy during testing. That’s difficult to testing for LCA and may be well studied by more complex strategy shifting model.