Psychology's relationship with the concept of self-esteem is ambivalent. For many, self-esteem is the central characteristic of personal well-being, high functioning, and psychological health (e.g., Shelley Taylor). For others, self-esteem has a dark side through its proximity to narcissism (e.g., Roy Baumeister). For a third group, self-esteem is mostly a by-product of other psychological processes, and not directly involved in important matters such as survival (Baumeister, Campbell, Krueger, & Vohs, 2003).

If self-esteem is not a commodity of much objective value, we should not spend too much energy pursuing it. More specifically, we should not pursue it as vigorously as we pursue sex or food. Sex and food are worthwhile for two reasons. Consumption feels good and it pays dividends in children had and years lived. Self-esteem just feels good. These considerations yield a testable hypothesis: If you ask people to evaluate sex, food, and self-esteem, self-esteem should come in last.

The study testing this hypothesis has been done and the data turn the predicted rank order on its head (Bushman, Moeller, & Crocker, 2011). Respondents liked and wanted self-esteem-related experiences (acing and exam, receiving a compliment) more than they liked and wanted their favorite sexual experience or favorite food. K'ching! One wonders if trading compliments among consenting adults could soon be illegal considering the enjoyment they're getting out of it. But I digress.

So here's a provocative finding that gets media attention (see for a blog at the NYT) and the psychologists' ambivalence seeks release by speculation. What does it all mean? What's wrong with people? Why do they overvalue this ‘commodity' that does so little for them?

Bushman et al. speculate that people (at least at this time, in this culture, and on college campuses) have become addicted to self-esteem. Addiction is a powerful metaphor and it should not be wielded carelessly. How is a preference for high self-esteem a sign of addiction? Recall that respondents rated how much they liked an experience and how much they wanted it. Normally, liking and wanting are well-aligned. We like what we want and want what we like. But sometimes, there is a discrepancy. Addiction involves a special kind of suffering because the wanting increases to become a craving, while the liking diminishes through habituation. The addict becomes ever more miserable because she knows she is irrationally hunting after something that is no longer enjoyable. It should be clear now: Addiction means that "Wanting > Liking," and Bushman and colleagues say so throughout their article. For example, they assert on p. 4 of the manuscript that "If self-esteem has become "needed" today, then people may "want" it even more than they "like" it. Wanting more than liking self-esteem would be one sign that people might be "addicted" to it."

Alas, this is not what they find. They find that Liking > Wanting for all experiences studied. It's just that the difference of Liking - Wanting is smaller for self-esteem related experiences than for sex or food related experiences. That should lay the addiction hypothesis to rest, no? No. Here's the result as reported by Bushman et al. cum interpretation (p. 9).

"A significant Reward • Value interaction, F(2,127)=18.51, p<.001, ή2=.23, showed that although liking exceeded wanting for all rewards, this difference was lower for self-esteem, t(128)=3.22, p<.002, compared with food t(128)=5.89, p<.001, and especially sex, t(128)=10.14, p<.001. One indicator of addiction is that people want something more than they like it (Robinson & Berridge, 2003). Thus, if people are addicted to anything, they are more addicted to self-esteem than to food or sex."

There's a slippery slope! If you give up the inequality of Wanting > Liking as definitional of addiction, you have just gone on record as saying that half of your preferences are addictions. It just depends on what experiences are being compared. If, for example, only sex and food were in the mix, Bushman study 1 would suggest that food is addictive (because [liking food - wanting food] < [liking sex - wanting sex]), whereas study 2 would suggest the opposite. Now do you want a research strategy that will pathologize half of your behaviors?

The pursuit of self-esteem may pose risks, as many studies by Bushman, Baumeister, and Crocker have shown, and it may be pathetic, but according to these data, it is not pathological.

There is a subtler problem in this research, and that is to think about addiction in terms of a difference score in the first place. When it is assumed that Liking - Wanting = Addiction, Addiction is being reified, it is treated as a thing that should be uniquely related to other things. Bushman et al. think so when correlating the differences between Liking and Wanting with other variables of interest (e.g., self-enhancing behaviors). The trouble is that difference scores cannot predict anything beyond that their components predict. Perhaps self-enhancement is just a matter of Liking or just a matter of Wanting, or some combination of the two. Taking the difference does not add anything, as it were.

This horse died long ago, but let me flog it once more. Suppose I want Andalusian ham more (5 on a 5-point scale) than I like it (3 on that scale) and so do you, but both your wanting (3) and liking (1) are lower than mine. Do we not agree that I am more of an addict than you are? Our behavior should reflect a critical difference. An addict actually consumes more of the troublesome commodity than the non-addict does. The difference-score measure does not pick up this effect. Multiple regression with interaction terms solves this problem. With this method we can test the following reasonable hypothesis: For people who dislike X but not for people who like X, increases in wanting are associated with greater misery. This implies that the difference between Wanting and Liking must be very large.

Regarding addiction, I am with Carl Gustav Jung, who maintained that "Every form of addiction is bad, no matter whether the narcotic be alcohol or morphine or idealism." In psychology, some practices may qualify as addictions, such as significance testing or the need to pathologize our research participants. I still suffer from the former but have overcome the latter.