In “Thinking, Fast and Slow” (2011), Nobel Prize winning psychologist Daniel Kahnemann tells an important story that illustrates an error you might be making, to your detriment and to that of others. The story: While talking with Israeli flight instructors, Kahnemann stated that the evidence is abundant that reinforcement (e.g., praise), not punishment, is the best way to teach new behaviors. An instructor disagreed with him, saying that almost every time he praised a trainee for outstanding performance in a drill, the trainee did worse the next time, and that almost every time he yelled at a trainee for making a boneheaded mistake, the trainee did better the next time. So, punishment is the more effective method of training!

Do you see the error in the instructor’s analysis, as it relates to extreme levels of performance? If you do, you deserve the Malouff Statistics Medal. The error involved failure to understand regression toward the mean. I call this the Extreme Behavior Error. Rather than try to explain this subtle concept in an abstract way, Kahnemann asked all the flight instructors to stand in a row, drew a line on the floor behind them, and asked them, one at a time, to toss a coin over their shoulder (while not looking) to try to land near the line. Some landed very close; some were far away; many were in between. Kahnemann measured the distance from the coin to the line for each toss and wrote that on the board, with the instructor’s name. He then repeated this process and showed that the extremely good performers and extremely poor performers from the first toss tended to be not as extreme the second time. That result, which occurs when chance factors affect a behavior or performance, is regression toward the mean.

How does the extreme behavior error apply to your life? You might be applying it to how you treat your children — using punishment because it “works” and not using reinforcement because it doesn’t “work.” That is the extreme behavior error, which might explain one reason parents so often use punishment, including spanking, with their children. Reinforcement generally works in improving behavior, but in the case of extreme behavior, it may not seem to work perfectly due to regression toward the mean. Got it?

You can make the same error with your students, your employees, and yourself! If you praise yourself only when you perform extremely well at some task, you may see that the next time you try the task, you don’t do so well and therefore you stop praising yourself for good behavior. Big mistake! If you usually scream at yourself for making an extreme error, you may see that you don’t make that mistake the next time and conclude that you need to scream at yourself every time you make a mistake. It seems as if the screaming is working. Actually, regression toward the mean is working.

So Kahnemann was right about the value of reinforcement over punishment in training. The assertive flight instructor was wrong — he made the extreme behavior error, which involves an illusion of what causes what.

The extreme behavior error applies only when a behavior includes chance elements. Feeling very depressed has chance elements, such as recent negative events, random key thoughts, and developing a minor physical illness. A group of extremely depressed individuals will tend to feel less depressed if they are assessed a month later. That is why treatment studies with assessment before and after treatment and no control group have limited value — the participants, extreme to start, tend to regress toward the mean, creating the illusion that the treatment worked. So, good studies use randomized control designs, where the control group is just as likely to regress toward the mean as the experimental-treatment group. The question then is whether the experimental group improved more than the control group. This design controls for (equalizes between groups) regression toward the mean.

A common example of the extreme behavior error is the belief that certain stock funds will do well in the future because they have done well in the past. Regression toward the mean will tend to occur for the funds that do extremely well because part of their past success was the result of chance factors. The much cited book “A Random Walk in the Woods” makes this point.  In a related matter, individuals who buy stock at the all time high for a stock exchange and sell at an extreme low may well be making the extreme behavior error — with a costly consequence.

Have you made the extreme behavior error? Seen others make it?

John Malouff, PhD, JD, Assoc Prof of Psychology