This paper introduces and validates quantitative performance measures for a rhythmic target-hitting task. These performance measures are derived from a detailed analysis of human performance during a month-long training experiment where participants learned to operate a 2-DOF haptic interface in a virtual environment to execute a manual control task. The motivation for the analysis presented in this paper is to determine measures of participant performance that capture the key skills of the task. This analysis of performance indicates that two quantitative measures—trajectory error and input frequency—capture the key skills of the target-hitting task, as the results show a strong correlation between the performance measures and the task objective of maximizing target hits. The performance trends were further explored by grouping the participants based on expertise and examining trends during training in terms of these measures. In future work, these measures will be used as inputs to a haptic guidance scheme that adjusts its control gains based on a real-time assessment of human performance of the task. Such guidance schemes will be incorporated into virtual training environments for humans to develop manual skills for domains such as surgery, physical therapy, and sports.