A source of random error in using structured observations of behaviors, where subjects overachieve because they are aware they are being observed.