Deception has been extensively studied in many disciplines in social science. With the increasing use of instant messaging (IM) in both informal communication and performing tasks in work place, deception in IM is emerging as an important issue. In this study, we aimed to explore the online behavior of deception in a group IM setting. The empirical results from triadic groups showed that two types of online behavior under investigation could significantly differentiate deceivers from truth-tellers. The findings can potentially broaden our knowledge of behavioral indicators of deception in human interaction and improve deception detection in cyberspace.