Sequential pattern mining is a useful tool in understanding learning processes, but identifying the most relevant patterns can be a challenge. Typical sequential pattern mining algorithms and interestingness metrics mainly focus on finding behavior patterns common across all students. However, educational researchers also care about individual differences. This study proposes a method for finding sequential patterns which usage have high variation across students. This method borrows techniques from the field of lag sequential analyses and meta-analyses. It uses the log odd ratio to model the individuals' usage of a sequential pattern and the heterogeneity test to examine the usage variation. We applied this method to analyzing student action logs in a virtual experimental environment and present preliminary results illustrating how the identification of sequential patterns with high usage variation provides interesting information about students' learning behavior. The proposed approach adds a way for understanding individual differences in learning processes.