In the 1960s a teacher was given a roster showing the actual I.Q. test scores of the students of one class, and for another class a roster in which the I.Q. column had been (mistakenly) filled in with the students locker numbers. The teacher assumed that the locker numbers were the actual I.Q.s of the students when the rosters were posted at the beginning of the semester.
After a year it was discovered that in the first class the students with high actual I.Q. scores had performed better than those with low ones. But in the second class the students with higher locker numbers scored significantly higher than those with lower locker numbers!