To determine causation – rather than just correlation – the team devised a second study in which workers were given a series of tasks and the opportunity to cheat on them. The half who knew they were under surveillance, they found, were more likely to cheat.