Why Positive Reinforcement Doesn't Seem to Work

All you have to do to get people to behave better is praise what they're doing right, or so you've heard. Everybody knows positive reinforcement is the way to go. But if you've ever actually tried it — if you're a parent or an employer or manager — you've probably discovered it doesn't seem to work very well. And yet plenty of studies show that it does. So what is going on?

To find out, Paul Schaffner designed an ingenious experiment. Let's say you were one of the participants. You walk in and sit down and he explains the rules. You are to play the role of a teacher. You will be able to punish a student for being late and reward him for being on time. Or you can choose to do nothing at all. You are presented with a series of the arrival times of a student for fifteen days in a row, one day at a time. Sometimes your student arrives late and sometimes he arrives on time.

You want to get a good result so you punish the lateness and reward the punctuality. Almost everyone in the experiment does this. What you don't know is that the student is arriving late and on time at random. The arrival times were all decided before the experiment started and your punishment and rewards are not influencing the student at all.

But it very much seems to you that your rewards and punishments are having an influence — a bad influence. Your student seems to do better after being punished, and worse after being rewarded. Most of the participants in this experiment concluded that punishment works better than rewards. It was the obvious conclusion, given what they experienced.

But their conclusions were wrong. Rewards do work, and punishment doesn't work as well. They got the wrong impression because of something called "statistical regression." This means that statistically, an exceptionally good performance is usually followed by a performance that is not as good. A high point, in other words, is statistically likely to be followed by a regression of some sort. Of course! It is "exceptionally" good because it is exceptional. It is the exception.

And in the same way and for the same reason, an exceptionally bad performance is usually followed by something somewhat better.

This is illustrated by what's called the "Sports Illustrated jinx." Many athletes believe it is bad luck to appear on the cover of Sports Illustrated. They think it will jinx them. And given how statistical regression works, you can now understand why they think that way. Most people appear on the cover of Sports Illustrated after an amazingly good performance. In fact, that's why they were chosen for the cover. So statistically, an athlete's next performance will very likely be worse than the performance that got them on the cover. So it seems perfectly obvious that you jinx yourself by being on the cover.

Sports Illustrated wanted to put the swimmer, Shirley Babashoff, on their cover before the 1976 Olympics, but rumor has it she tried to talk them out of it. She didn't want to jinx herself. They convinced her it would not be bad luck by pointing out that Mark Spitz was on the cover and came home from the 1972 Olympics with seven gold medals! The jinx is an illusion.

We can see this jinx easily by looking at it differently. The top score you can get bowling is 300. It is the league tournament, and John bowls a 300. His teammates are ecstatic! His league manager makes him team leader. But the very next time John bowls, is he likely to bowl another 300? Not likely. He will probably bowl something less than that. Why? Because he got so much praise? No, it's just statistical regression.

In other words, the short-term impression you get is that positive reinforcement doesn't work. But that is illusory. In the long run, it works very well indeed.

And you can also get a false impression that punishment works well when in the long run it doesn't work well at all. If a manager decides to let her people make their own decisions, and only intervenes when things go wrong, she would probably see improvement after her angry outbursts every time. Ken Blanchard called this management style the "leave alone — zap" style. It is also known as Seagull Management (swooping in out of nowhere, making a mess, and then disappearing).

It is statistically likely that the employee's behavior will "regress to the mean" no matter what the manager does, or even if she does nothing at all. If the employee's behavior was worse than usual, it will regress to the mean, or in other words, regress to the average. In this case, it will improve. The manager with the leave alone — zap style of management would very definitely get the impression that her interventions were effective. But over time, that kind of managing creates resentment. The immediate results, however, can mislead her into thinking she's doing the right thing.

If you can remind yourself of the misleading nature of the regression effect, you will have more success in using positive reinforcement with your employees or your children.

For example, if you're trying to get your son to keep his room cleaner, and one day you see he has made an effort, you might praise him. Don't expect things to just get better and better. After your praise, his room might get worse. Statistically, that is bound to happen. But it doesn't mean your praise had no influence. People change and improve in a roller coaster fashion — not in a straight line.

If you were watching a roller coaster, and you praised it when it went up and criticized it when it went down, you'd get the impression that praising its upward movement made it go down, and criticizing its downward movement made it go up. But statistical regression is misleading. It causes people to draw the wrong conclusions about what really works.

And to make matters worse, your brain's negative bias will make negative results more noticeable and memorable. (Read more about that here.)

So now you know why positive reinforcement doesn't seem to work. We not only have a negative bias, but statistical regression makes your results misleading. It is difficult to see the true situation: That rewards and acknowledgements really do work, and they work better than punishment.

Adam Khan is the author of Principles For Personal Growth, Slotralogy, Antivirus For Your Mind, and co-author with Klassy Evans of How to Change the Way You Look at Things (in Plain English). Follow his podcast, The Adam Bomb.