In a face-to-face classroom, the instructor can easily diagnose students' motivational status by observing their facial expressions and postures, but such cues are absent in an online classroom. Therefore, online instructors often estimate students' motivational level based on their online behavior such as the number of messages they post, and look for effective strategies to help them actively participate in online dialogues. One such strategy is contract learning which facilitates self-directed behaviors through structuring an agreed learning process. This study reports a contract learning strategy in a graduate-level online class, examining whether a sample of 28 students' motivation could indeed be predicted by their online behavior. Results from the study found that the students' online behavior was not a predictor for their motivational status, though there were age and gender differences in their online behavior. The students felt more self-directed and motivated during contract learning, but what they really liked was being able to select assignments that were relevant to their interests and needs. This paper concludes by discussing practical implications of the findings at the end. (Contains 4 tables.)