The obesity epidemic can only have surged in recent decades because of a decline in physical activity levels, an increase in caloric consumption, or both. Although the contribution of physical inactivity to the problem is still unclear, both surveys of food consumption and data on the quantity of food distributed in the United States suggest that caloric consumption of the average American has increased during this time (3,4). This evidence should focus our attention on the most important question in this epidemic: why do people eat too much?

Beyond portion size, one principle is that the amount of food consumed increases as the effort to eat it decreases, even if the differences in effort are tiny. For example, office workers who had chocolate Kisses within reach on their desks ate an average of 5.6 more candies (total of 136 kcal) per day than did workers for whom candy was placed on a shelf 2 m away (14). The kilocalorie amounts demonstrated by these experiments (8-14) are significant, because a small caloric imbalance over time can produce obesity. In fact, by one estimate, the median weight gain in the obesity epidemic over the last two decades could be caused by a daily excess of only 100 kcal to 150 kcal (3).

The mere sight of food can stimulate people to eat. For example, Wansink et al showed that office workers ate 3.1 more chocolate Kisses (total of 75 kcal) when the candy was placed on their desks in transparent jars than when the candy was placed in opaque jars (15). In another clever experiment by Wansink et al, researchers secretly refilled bowls of soup while people ate from them and found that people ate 73% more soup when this occurred (16). Wansink's work demonstrates repeatedly that environmental cues influence the frequency and quantity of what people eat and that people do not typically recognize these cues (15-17).

Psychologists have shown that behavioral responses to environmental stimuli, whether the behaviors appear to be instinctive or deliberate, can be influenced by what has been termed priming, or the manipulation of decisions and judgments by the previous presentation of words, concepts, or images that are not perceived as being related to the task at hand (7,20-22). The effects of priming can be surprising. For example, Bargh et al primed subjects by having them solve word puzzles containing words related to anger, such as rude, impolite, and obnoxious; afterward, these subjects were more likely to interrupt a conversation than were people primed with neutral words or words related to politeness (23). North and colleagues showed that when they played French music in a wine store people bought more French wines, and when they played German music people bought more German wines, with little or no awareness of the effect of the music on their purchases (24). An example of how food consumption can be influenced by priming was shown in a study in which thirsty subjects shown a "happy face" subsequently drank more of a fruit-flavored drink and rated it more favorably than did subjects shown an "angry face" (25).

In recent years, psychologists have developed a greater understanding of the automatic behaviors, which can be defined as those that operate without cognitive direction (30,31). A great deal of mental effort is required to make conscious decisions and then implement them in the form of behaviors (30). Most of our responses to our environment can be understood as automatic behaviors. Human beings smile or laugh when amused, frown when annoyed, become startled when surprised by a loud noise, and tense their muscles when threatened, all without making any conscious decision or being aware of the behavior. An example of a more complex automatic behavior is social mimicry. In conversation, people copy others' mannerisms, such as smiling, rubbing their face, and shaking their feet, regardless of whether they are acquainted with the other people and without the slightest recognition that they are copying them (32).

Bargh has defined four characteristics of automatic behaviors: 1) they occur without awareness, 2) they are initiated without intention, 3) they continue once initiated without control, and 4) they operate efficiently or with little effort (33). However, not all of these criteria are required for a behavior to be considered automatic. Studies on food consumption indicate that eating should be viewed as an automatic behavior (8,9,12,14,15,34,35). People are generally not aware of how much they are eating. In studies demonstrating the influence of portion size on eating, people given large portions usually did not believe they had eaten more than people given normal-sized portions, and when surveyed afterwards, they did not report greater feelings of fullness compared with people eating smaller portions (8,9,12,34). Evidence that eating begins without conscious intent can be taken from both the tendency to eat any food that is in sight or at arm's length (14,15,35) as well as the finding that people are more likely to eat simply because it is mealtime than because they are hungry (36). Once people initiate eating, they usually continue until the food is gone or until some other external occurrence changes the situation. In one study, people were less likely to stop eating because they were full than because no food or drink remained, they had no time to eat more, or they had finished watching television (36). These studies also demonstrate that the natural trajectory of eating — that is, what takes place without conscious effort — is for it to continue. Effort is not required to continue eating when food is present; effort is required to refrain from eating when food is present.

It is intuitive that behaviors central to a human being's or animal's survival are automatic. Evaluations of the safety of our surroundings and judgments of the potential danger posed by strangers are often automatic and typically based upon stereotypes (23,37). The "fight or flight" response would be too slow to be protective if it required deliberate decision-making. Because a central evolutionary task of human beings has been to consume enough energy to live, it is not surprising that we are programmed to eat whenever food is within reach.

Characterizing eating as an automatic behavior does not mean that human beings cannot bring eating under volitional control. People certainly can refuse dessert or resist the temptation of the chocolates in the jar on the desk. All automatic behaviors can be controlled temporarily. Human beings can consciously prevent themselves from smiling when amused, frowning when annoyed, or tensing their muscles when threatened. It just takes effort. But the amount of effort required to refrain from eating when food is present is substantial, and it is nearly impossible to sustain over the long term. For example, in a study on self-control, Baumeister and colleagues allowed members of one study group to eat freshly baked chocolate-chip cookies while members of a second group given access to the cookies were told to refuse them and were allowed to eat only radishes; members of a third control group had no food to eat or to refuse (38). Afterward, the researchers asked the members of the three groups to work on an unsolvable puzzle. Members of the control group given no food worked on the puzzle for 21 minutes before quitting, and members of the group that were allowed to eat the cookies worked for 19 minutes. In sharp contrast, members of the group that had to refuse the cookies quit after only 8 minutes, and they reported more fatigue than members of the other groups. The work of refusing tempting food required mental effort — enough to deplete participants' ability to perform other higher-level processes.

In general, human self-control over automatic behaviors is limited. Self-control tires like a muscle and taxes our ability to perform other tasks (39). And just as refusing food depletes a person's mental reserves, tasks requiring mental effort can reduce the ability to resist the temptation of food. In one study, people trying to maintain a diet who were deliberately frustrated with an unsolvable problem increased their food intake compared with people who were not trying to control their eating (40). The high mental demands of dieting may partially explain the commonly observed pattern of dieters initially losing weight and then gaining it back (41).

Automatic behaviors share another important characteristic. Because people are unaware of the behaviors, they are also unaware that the behaviors are not under their control. Nisbett and Wilson have shown that people are often unaware of a particular stimulus that elicits a response, and even if they are aware of both the stimulus and the response, they may be unaware that the stimulus actually caused the response. Instead, people tend to fabricate reasons to explain their behaviors, typically choosing the most plausible, culturally acceptable theories (42). Bargh and Chartrand found that, even after people have been shown the results of experiments demonstrating the automatic nature of their actions, they steadfastly refuse to believe that those actions did not result from conscious choice (30). Our difficulty as a society in accepting how strongly our environment influences eating may stem from our inability to recognize and our refusal to accept our own eating as an automatic behavior. We blame our lack of willpower on the inability to maintain a diet, when it is more likely that our automatic responses to ubiquitous cues to eat and the availability of cheap, convenient, calorie-dense food are responsible.

Deborah Cohen, RAND Corporation. 1776 Main St, Santa Monica, CA 90407, Phone: 310-393-0411 ext 6023, Email: dcohen/at/rand.org.