Key Problem: because subjects in the real world do not assign themselves to groups randomly, there are many other differences among the groups that could contribute to differences in the dependent measures (e.g., brain damage may be due to poor diet, accidental head injury, other drug use, etc.)