Acting on Basis of Current Emotion
As we already said, the evidence for current emotional state causing behavior is quite sparse and weak. Nobody is saying that it never happens. Certainly people say or do things because they are angry, afraid, or happy. But how common is that? And, moreover, can that be the main function of emotion?
A meta-analysis by DeWall, Baumeister, Chester, and Bushman (2015) compiled evidence from several hundred papers that examined links between emotion and behavior. They took every article in the leading social psychology journal (the Journal of Personality and Social Psychology) that analyzed for whether emotion mediated between the study's independent variable and any outcome involving behavior or social judgment, from 1987 to the present (1987 was chosen as the start date because that was when new procedures for analyzing mediation started to become widely used, based on an article the previous year that explained and justified them, thereby bringing about a revolution in how social psychologists worked with their data; Baron & Kenny, 1986).
The idea of mediation by emotion goes something like this. First, something happens in the social situation. Second, the person experiences an emotion as a result of perceiving that. Third, the person does something. Experiments are published when the first step causes the third, that is, when the circumstance reliably causes a certain kind of response. The question is whether the person's emotional reaction to the circumstance contributes to causing what the person does. Someone insults you, and you get angry, and therefore, you act aggressively toward that person. If the insult had not elicited anger, you would not have responded to it with aggression. For evidence of mediation, to put it simply, the researcher must show that the circumstance is statistically linked to the behavior, that the emotion is also linked to the behavior, and that when you control for the emotion, the link from circumstance to behavior disappears. In that case, the emotion accounts for the link. (Researchers are also supposed to show that it doesn't work the other way around; namely that even if you control for the circumstance, the emotional state still predicts the behavior.)
A significant finding of mediation means that the emotion is the immediate cause of action. The causal link leads from the circumstance to the emotion and thence to the behavior.
The DeWall et al. (2015) meta-analysis showed that emotion sometimes causes behavior—but just barely. Of several hundred analyses, only 20% were statistically significant. Significance means that there is less than a 5% probability that the results would be found by random, chance variation. So one would expect 5% of the studies to yield significant results by random chance; 20% is thus not much better than chance.
That low rate is remarkable not just because it shows how infrequently researchers have been able to show that emotion causes behavior. Space in that journal is coveted by thousands of researchers, who compete fiercely to publish their best work in those pages. According to the logic of research design, nonsignificant findings are hard to interpret if not completely meaningless, and so most journals publish only studies that find significant results. That such a prestigious journal is so full of nonsignificant results is startling. The only explanation is that authors, reviewers, and editors all assume that emotion is a major if not the major cause of behavior, and so everyone must analyze the data to establish whether that is correct in the current case. If someone wants to show that behavior follows from something other than emotional states, it is necessary to provide data to show that emotion is not the real cause. And apparently the field has been slow to wake up to the fact that emotion is typically not the driving cause of behavior.
The meta-analysis also examined anticipated emotion. In contrast to the hundreds of analyses that looked for effects of people's current emotional state, only a handful examined anticipated emotional state. But those that did look at anticipated emotion yielded a much higher rate of success, near 90%. Much more research is needed, but the evidence available at present suggests that behavior is much more reliably based on how people expect to feel in the future than how they feel at present.
Moreover, and crucially, even the feeble results for current emotional state probably overstate the case. It is possible to get results that look like one's current emotional state is driving behavior—even when the true cause is anticipated future emotion. In particular, many consequences of unpleasant emotional states come about because people are trying to repair their unpleasant state and make themselves feel better.
This problem was identified by the eminent emotion researcher, Alice Isen (1984, 1987) in the 1980s. Surveying the research literature on emotion, she identified a key ambiguity in studies of bad moods and emotional distress. It is quite hard to disentangle the direct results of the emotion from the person's quest for an anticipated better state. Suppose, for example, that you ran a study in which you put people into good, bad, and neutral moods by random assignment and then measured how much ice cream they ate. You might find that the people who felt bad ate the most ice cream. Does that mean that bad moods cause people to eat sweets, perhaps to lose self-control, and to disregard their diets? Or, alternatively, does it mean that people in bad moods eat ice cream because they think the treat will cheer them up? Isen saw no solution to this problem and advocated that researchers shift to study positive emotions, which do not suffer from that problem of ambiguity. After all, people do not usually try to escape from feeling good. So if a researcher makes people happy and then observes changes in their behavior, it is fair to assume the behavior flowed directly from the happiness, rather than from efforts to change the emotional state, thus avoiding the problem that plagues studies of negative emotion. The rest of her career followed that path with considerable success.
Around the same time, however, another researcher, Robert Cialdini, created a way to tease them apart. Cialdini had been locked in an ongoing debate with Dan Batson about the causes of helping. Are people ever really altruistic, in the sense that they do things purely for the benefit of others? Or do they just help because it makes them feel good? Cialdini suspected that most if not all the evidence purporting to establish human altruism was actually just a sign that people want to feel good. They help others because doing so cheers them up. (It works: Helping others and doing good deeds do in fact generally make the helper feel good.)
Matters came to a head with evidence that ordinary people who find themselves momentarily in sad, unhappy moods are especially helpful. Does sadness cause helping? Or does sadness make people want to feel better, so they help in the hope and expectation of attaining a cheerier mood?
The ingenious solution devised by Cialdini and his colleagues (Manucia, Baumann, & Cialdini, 1984) was the mood-freezing pill. "Freezing" in this case had nothing to do with being cold: It simply meant that whatever emotional state you have when you take the pill would continue for a while and could not be changed. In reality there is no such pill, but it wasn't necessary to actually have such a pill. All the researchers had to do was convince people that they had one.
The beauty of the mood-freezing pill is that once someone has taken it, there is no point in trying to repair your mood or alter your emotional state. Again, this does not refer to actual pills but rather to people's beliefs. If you think your mood can't change no matter what you do, any attempts to change it are doomed to fail, so it makes no sense to try. Therefore, this clever ruse can solve Isen's problem. Sad people who believe they have taken a mood-freezing pill know there is no point in trying to cheer themselves up. If sadness causes behavior directly, the mood- freezing pill won't matter, and people will continue to perform the same actions regardless of whether they have taken a pill. But insofar as behavior is designed to cheer oneself up, the mood-freezing pill should put a stop to it.
The researchers gave what were in fact placebos (pills with no effects) to college students after inducing various mood states. The students were told that the researchers were testing new pharmaceutical products and that preliminary results had shown that these had the side effect of locking one's emotional state wherever it happened to be, for about an hour. To make it plausible (as no known drugs actually accomplish this), the experimenter explained that many drugs have emotional side effects, such as cannabis, which intensifies emotional reactions. This new drug, they said, would not intensify emotions but simply keep them about the same for about an hour.
The findings were dramatic. Without the mood-freezing pill, sad moods led to more helping, as in previous studies. But the sad people who had supposedly taken the mood-freezing pill failed to help.
This pattern is relevant because it further undermines the evidence that current emotional state causes behavior. In such a study, one could get evidence that sad moods mediated statistically between circumstances and helping behavior. But in reality, the helping was not a direct result of sadness. Sad people only helped when they thought that helping might cheer them up. So a correct understanding of the phenomenon would show that anticipated emotion was the true cause. Presumably, a substantial number of the 20% of significant findings in the meta-analysis that purported to show current emotion causing behavior could in fact reflect acting on the basis of anticipated emotion.
There have been further findings confirming the mood-freeze pattern. Sad people do eat more sweets than happy people, but not if their moods have been frozen (Tice, Bratslavsky, & Baumeister, 2001). Upset people seek immediate rather than delayed gratification, but again, not if their moods are frozen (Tice et al., 2001). Perhaps most dramatically, anger fails to cause aggression when moods are frozen (Bushman, Baumeister, & Phillips, 2001). The latter is especially remarkable because the idea that anger leads to aggression is one of the foundations of research on aggression, and some experts have asserted that anger contains within the emotional state the incipient muscle movements for aggression (Berkowitz, 1989). But apparently, angry people lash out mainly because they think they will feel better as a result. Remove the expectation that attacking someone will make them feel good and reduce anger, and anger ceases to cause aggression.
There are profound implications of this view that human behavior is guided by anticipated emotions. Whatever else people are doing, they are often also acting in ways that will improve their future emotional state. They choose actions that they expect will bring positive emotions and reduce or avoid negative ones. In an important and emotional sense, most behavior is thus guided by prospection. Deciding what to do is commonly informed by expectations about the future and the emotions associated with this or that anticipated outcome.
If that view is correct, then emotional prospecting is crucial. A false prediction about how one might feel will lead to a wrongheaded, counterproductive action. That raises the question of how well people can predict their emotional outcomes, to which we turn in the next section.