Studies have been conducted to determine the most popular PA participated by children, but little is known about association among PA participated by children. The purpose of this study was, using a new data mining technique, to find associations among PA participated by elementary school children. The data from the National Children and Youth Fitness Study II (N=4,678 students, 2,350 boys and 2,328 girls, Grades 1 to 4) were used for the study. Parents in that study were asked to report the five most frequent outside-school PA that their child had participated in over the previous year. The data was analyzed using the association rule, which uses an "if, then" criterion and calculates the confidence of the rule. If, for example, "A" occurs in the data greater than the minimum support level, it becomes a candidate to be considered for a rule. Occurrence of other events with "A" will then be examined based on a predetermined confidence level. For example, if "A" occurred 10 times and "B" was present in 9 of the 10, "A" and "B" are associated with a confidence of .9. The data mining software, WEKA (Waikato Environment for Knowledge Analysis, 3.2; developed by Witten & Frank, 1999), was used to discover association rules in the reported PA. The minimum support level was set at .05, and the confidence was set at .8. The data (93 activities) were analyzed by grade and gender, with a total of eight groups, and the association between the five most frequently participated PA (e.g., racing-sprinting) and other PA were analyzed for each group. Together, a total of 40 analyses were performed (4 Grades x 2 Genders x 5 PA). Fifty-three association rules were found and many of them reflected natural relationship among PA. Boys participated in individual activities (e.g., walking), for example, would likely to engage in other individual activities (e.g., racing-sprinting; Confidence=1). In contract, those, who played team activities, would likely to participate in another or more team activities (e.g., baseball-softball and soccer; Confidence=.82). Different patterns of team-activity participation, however, were found for girls. In fact, no rule could be found in girls°¯ participation in team activities. In conclusion, the results of association rule analyses provide us some better understanding on relationships among common PA participated by children. With such information, efforts can be made to promote specific activities to a target group of children.