When a research assistant (maybe) fabricates data

I was a young, untenured assistant professor, a few years into my first job. Investigating social interaction processes, my laboratory studies were especially effort-and time-intensive. Each 90-minute experimental session required three participants who would engage in live interactions with at least one of the others. So that participants would be strangers to one another, research assistants (RAs) scheduled participants by phone. Multiple RAs were needed for each session to prepare the audiorecording equipment and to run participants through procedures and individual verbal debriefings. It would take a full semester-sometimes longer-to run a sufficient number of sessions to test our hypotheses. So when we couldn’t schedule a full set of participants and had to cancel a session, we felt the loss; when a participant didn’t show for his/her scheduled session, we felt the loss; when the audiotape equipment failed and we had to discard a session, we felt the loss. We worked very hard to minimize those losses. One day my most trusted undergraduate research assistant - I’ll call this person Julie, to protect his/her identity - came by my office, looking quite stressed. Julie told me she had stopped by the lab control room a day or so before and thought she saw another, newer RA - I’ll call this person Mike - filling out a subject questionnaire, one of our main sources of data. Unfortunately, Julie had panicked and left without exploring further or questioning Mike, but, after further consideration, thought I should know what she thought she saw. Although unsure, Julie’s sense was that Mike was making up or altering data. In fact, she was pretty confident this was the case, but not positive. What does one do with such information?