As elaborated throughout this summary, much of the workshop discussion revolved around how on the one hand, food triggers satiety signals from the digestive system to the brain indicating fullness, or hunger, while on the other hand, food and food cues also send sensory signals that activate the brain reward system. In his closing presentation, Edmund Rolls brought these ideas together and explored how sensory and satiety signals are integrated in the brain. Moreover, bringing the discussion full circle and touching on the concept of context that Laurette Dubé had introduced earlier in the workshop, Rolls explored how “higher-level” cognitive factors influence how people actually perceive and respond to all of these signals being received by the brain. This chapter summarizes his closing presentation and the discussion that followed. Box 5-1 highlights key points made by Rolls.
FOOD REWARD, APPETITE, SATIETY, AND OBESITY1
Rolls hypothesized that obesity is related to an imbalance between the satiety and sensory-produced reward neural systems, with the latter overriding the former as a function of individual differences in cognitive control. He described eating as an output of the interaction in the brain of satiety signals, all of the various sensory inputs (taste, smell, texture, sight) that are con-
______________
1 This section summarizes the concluding presentation of Edmund T. Rolls, D.Phil., D.Sc., Hon.D.Sc., M.A., Oxford Centre for Computational Neuroscience, Oxford, United Kingdom. Rolls mentioned Fabian Grabenhorst as a major contributor to the work that he discussed.
BOX 5-1
Key Points Made by the Concluding Speaker
- Edmund Rolls hypothesized that obesity is related to an imbalance between the satiety and sensory-produced reward neural signals, with the latter overriding the former. There are individual differences in the potency of reward signals. Moreover, cognition can influence the reward effects of food. Evidence from nonhuman primate and human studies suggests that all food-related sensory stimuli, including taste, olfactory, and visual stimuli, converge in the brain in the orbitofrontal cortex and amygdala, where the reward value of the food is computed and then modulated by gut satiety signals. Top-down cognitive control appears to be expressed in the same area of the brain as well.
- In Rolls’s view, given that humans are generally dominated by sensory inputs with high reward value, a key challenge for the food industry is to create highly palatable foods with low energy density.
verted into reward signals, and cognitive factors that modulate food reward (see Figure 5-1). He noted that satiety signals have been “genetically set” for tens of thousands of years, but that many of the sensory and cognitive effects of food have changed considerably over just the past 30 years with the increased availability of a wide variety of very highly palatable foods.
As far as which parts of the brain play a role in food intake, Rolls pointed to the orbitofrontal cortex and amygdala in the temporal lobe as being important sites of convergence for sensory inputs from the primary taste cortex (taste), the olfactory cortex (smell), the visual cortex (sight), and the somatosensory cortex (texture). For example, the taste pathway in primates, which enters via the brain stem and passes through the primary taste cortex, ends in the orbitofrontal cortex and amygdala. Neurons in the primary taste cortex do not actually respond to the input they are receiving. They simply “tell” the brain what the stimulus is independently of how “nice” or “pleasant” it is. Only when that signal reaches the orbitofrontal cortex and amygdala is its reward value represented. The same is true of other sensory signals as well.
In primates, the same areas of the brain—the orbitofrontal cortex and amygdala—also receive satiety signals, which enter by way of the lateral hypothalamus and brain stem to influence the reward value of food-related sensory stimuli. Also in primates, “top-down” cognitive processes have major influences on activity in the orbitofrontal cortex and amygdala and affect the way taste and other sensory stimuli are processed. Cognitive modulation of reward value is important, Rolls said, in determining how people respond in any situation.
FIGURE 5-1 Rolls’s framework for integrating all of the different food-related inputs in the brain and their effects on eating behavior.
SOURCE: Rolls, 2011. Reprinted by permission from Macmillan Publishers Ltd. on behalf of Cancer Research UK: International Journal of Obesity (London) (Rolls, E. T. 2011. Taste, olfactory and food texture reward processing in the brain and obesity. International Journal of Obesity [London] 35[4]:550-561), copyright 2011.
Sensory Signals and Reward
Sensory neurons are highly selective. Rolls showed data illustrating the selectivity of a single taste neuron that responds to glucose sweetness but not to salt, sour, or water and that can even differentiate between glucose sweetness and fruit juice. In a study of the macaque orbitofrontal cortex, Rolls and colleagues (1989) demonstrated that a single glucose-sensitive neuron increased from a baseline firing rate of a few spikes per second to 20 spikes per second when the monkey was fed glucose. As the monkey was fed more glucose, eventually the firing rate of that same neuron decreased to zero, as did the monkey’s feeding behavior. But the same neuron tested with fruit juice (black currant juice) continued to respond even after it had stopped responding to glucose. Rolls and his team interpreted these results as evidence for sensory-specific satiety, with the reward value of the glucose decreasing to a point where the monkey started rejecting the glucose but still responded to fruit juice. According to Rolls, the same phenomenon has been demonstrated repeatedly in humans.
Of importance, it is only in the orbitofrontal cortex, not the primary taste cortex, where taste reward value is represented. In a related study, Rolls and colleagues (1989) demonstrated that when fed glucose to satiety, neurons in the primary taste cortex continued to fire at the same rate even as those in the orbitofrontal cortex slowed to zero. Rolls described the orbitofrontal cortex as the “reward antenna on the world.” It “tells” the brain about the various details of food based on the responses of all the different orbitofrontal cortex neurons, with each neuron responding to a different combination of taste, odor, texture, and temperature stimuli. For example, one neuron might respond to glucose and fruit juice but not to sodium and monosodium glutamate, another to viscosity but not to taste, another to fat texture but not to viscosity, and so on (Verhagen et al., 2003). Together, the population of neurons in the orbitofrontal cortex produces information about a rich variety of reward stimuli and provides for sensory-specific satiety related to specific combinations of stimuli.
As another example of the selectivity of sensory neurons, Rolls and colleagues (1999) demonstrated that not only does the orbitofrontal cortex contain oral fat texture-specific neurons, but those neurons contribute to sensory-specific satiety. When test monkeys were fed cream, their fat texture-specific neurons fired until the monkeys became satiated on the cream. But after the neurons stopped responding to the cream, they continued to respond to glucose. In Rolls’s opinion, the fact that some neurons in the orbitofrontal cortex respond to fat texture but not to other textures has important implications for the food industry: low energy-dense foods could be made highly palatable through the incorporation of substances that trigger a fat texture response (Rolls et al., 1999, 2003; Verhagen et al., 2003).
Rolls emphasized that food rewards in the orbitofrontal cortex are a neuronal representation of stimulus value that have nothing to do with behavioral responses (Grabenhorst and Rolls, 2011; Rolls and Grabenhorst, 2008). In other words, the responses tell people “how nice things are, but not what to do about them.” At least that is the case with nonhuman primates, based on the evidence.
As far as human evidence goes, in one of Rolls’s early human neuroimaging studies, people were provided either tomato or chocolate to taste and later fed to satiety with that food (Kringelbach et al., 2003). The researchers found for both foods that the measured signal in the orbitofrontal cortex was high before the individuals became satiated and then decreased upon satiation, as did the pleasantness rating of the food. Moreover, either food could trigger sensory-specific satiety. After subjects were fed to satiety with tomato, their brain still responded to chocolate, and vice versa.
In another human imaging study, de Araujo and Rolls (2004) observed significant representation of oral fat texture in the anterior cingulate cortex
and ventral striatum that was independent of viscosity. They also observed a convergence of responses to oral fat and sucrose in the most anterior part of the cingulate cortex. Grabenhorst and colleagues (2010) found a linear relationship between a fat texture signal in the orbitofrontal and anterior cingulate cortex measured with magnetic resonance imaging (MRI) and the signal’s subjective pleasantness rating.
The Processing of Reward: Modulation by Cognition
If an individual is provided with an ambiguous stimulus—such as the chemical compound isovaleric acid, whose odor is similar to that of some cheeses, such as brie, but is also similar to body odor—the brain responds differently depending on how the stimulus is described (de Araujo et al., 2005). With isovaleric acid, for example, if one were told that what one was smelling was brie, one would believe that it was brie. But if one were told that it was soldiers’ socks, one would believe that it was soldiers’ socks. Based on these different responses to the same stimulus, Rolls wanted to know where in the brain sensory and reward information is being influenced by cognition. Does the flavor reward representation travel up into some cognitive area, or does cognition travel down into the reward system? Evidence indicates the latter.
By delivering isovaleric acid into the nose of test individuals using an olfactometer, Rolls and colleagues observed greater activation in the pregenual cingulate and orbitofrontal cortex when individuals were told that what they were smelling was cheese compared with when they were told it was body odor (de Araujo et al., 2005). In a control condition, in which individuals were provided with no actual odor but told that what they were smelling was either cheese or body odor, the response to cheese was still greater than the response to body odor, but there was a much lesser top-down effect than when there was also a bottom-up signal from the odor. Rolls interpreted these results to mean that cognitive effects, such as descriptions of food in advertising, affect the parts of the brain that represent pleasantness and reward value.
Top-down attention matters, too. In another human imaging study, participants were delivered monosodium glutamate taste and were told to rate either its pleasantness or intensity (Grabenhorst and Rolls, 2008). The instructions were intended to set the brain up for performing different tasks with the same taste. The researchers observed a greater brain response in the orbitofrontal cortex and pregenual cingulate cortex among individuals instructed to rate pleasantness, with the signal being linearly related to the pleasantness of the taste. When subjects were instructed to rate intensity, on the other hand, the researchers observed a greater response in the insular taste cortex, again with the signal being linearly related to the intensity of
the taste. In other words, Rolls explained, when one is interested in how intense a taste is, or what a taste is, one’s processing in the primary taste cortex is turned up, whereas when one is interested in how pleasant a taste is, one’s processing in the orbitofrontal cortex is turned up. In Rolls’s opinion, again, these results have implications for the food industry: it is important to know how drawing attention to any one property of a food is going to impact hedonic versus perceptual processing in the brain.
Rolls also has been curious about where in the brain differences among individuals in liking a food are represented. Are those differences represented in the sensory processing system (e.g., primary taste cortex) or in the hedonic system (ventral striatum, pregenual cingulate cortex, orbitofrontal cortex)? Rolls and McCabe (2007) conducted an experiment with chocolate cravers, identified as such using a standard food-craving questionnaire. They delivered liquid chocolate into the mouth and measured responses to the sight of chocolate as well as to chocolate in the mouth. They found no difference between cravers and noncravers in the primary taste cortex, suggesting that whatever separates cravers from noncravers is not involved with sensory processing. However, there was a difference in the pregenual cingulate cortex, with chocolate in the mouth producing a much larger response in cravers than in noncravers in the ventral striatum, and the sight of chocolate producing a much larger response in cravers than in noncravers in the pregenual cingulate cortex and orbitofrontal cortex. These results suggest to Rolls that individual differences in whether a food is liked are expressed in the hedonic system but not in the sensory analysis system. Rolls opined that understanding individual differences in brain responses to highly pleasant foods may help scientists understand the mechanisms that drive the liking of particular foods, food decision making, and food intake.
Summary and Future Directions
In summary, Rolls noted that all of the various food-related sensory stimuli, including taste, olfactory, and visual stimuli, converge in the brain in the orbitofrontal cortex and amygdala, where the reward value of the food is computed and modulated by gut satiety signals. Moreover, sensory rewards are biased by top-down cognitive or attentional control. The next step for the brain is decision making, said Rolls: “Once you have computed that something is nice, you then make a decision about what you are going to do about it.” He did not elaborate, but referred workshop participants to his book Emotion and Decision-Making Explained (Rolls, 2014).
Rolls identified several topics to consider for future discussion, all of which revolved around the problem of obesity and whether obesity overstimulates the food reward system in the brain (Rolls, 2011, 2012, 2014). He pointed to early work by Stanley Schachter suggesting that people who
are obese may be more sensitive to the reward properties of food. In Rolls’s opinion, there is considerable individual variation in sensitivity to different types of reward stimuli as a result of natural selection. In other words, he said, “We are all slightly different in the rewards that we find attractive.” He suggested future large-scale imaging studies aimed at examining brain responses in relation to body mass index (BMI). It would also be extremely interesting, in his opinion, to know whether people who are obese are more sensitive to the reward value of food, and perhaps less sensitive to the satiating property of food.
Rolls identified palatability and variety as two additional factors to consider when evaluating whether obesity may be related to overstimulation of the food reward system. Enhanced palatability in the human diet may lead to an imbalance between sensory reward and satiety signals, and enhanced variety may lead to increased food intake as a result of satiety’s being partly sensory-specific.
In general, Rolls opined, “humans are dominated by these sensory inputs produced by food that make it pleasant and rewarding.” It can take humans a week or two to adjust their response to satiety signals following a change in the energy density of their diet.
Most of what Rolls described during his talk was what he called the brain’s “pleasure system,” which computes how “nice” something is and gives rise to goal-directed action. But humans also have what he called an explicit “reasoning system.” For example, a person can decide not to eat ice cream because doing so may lead to obesity. The two systems have different goals. The goal of the reasoning system is to produce long-term optimal behavior using advance planning. The goal of the pleasure system is short-term reward, with the reward value of a stimulus being influenced by its adaptive value during evolution. The two systems may be in competition, said Rolls. Again, individual variation is important, in his opinion: “Some individuals may be more susceptible to advice and operation of the explicit, cognitive, reasoning control system.”
In conclusion, Rolls opined that the “mismatch hypothesis” appears reasonable. According to this hypothesis, food palatability, availability, variety, and exposure through advertising have increased food reward in the past 30 years, while satiety signals have remained unchanged, and this “mismatch” is contributing to overeating. The challenge for the food industry, assuming that the mismatch hypothesis holds, is to create “healthier” but still highly palatable foods, said Rolls.
CONCLUDING DISCUSSION WITH THE AUDIENCE
The workshop concluded with an open discussion in which all speakers were invited to the stage to answer questions from the audience. Questions addressed a range of topics.
Sensory Versus Metabolic Effects of Stimuli: What Does the Science Say?
There was some discussion of the relative importance of sensory versus metabolic effects on food preferences. Richard Mattes commented on what is now a substantial literature showing that for salt and fat and, less compellingly, sugar, preferred levels are determined more by sensory exposure than by metabolic effect. For him, this suggests that there is nothing “special” about high-salt versus low-salt or high-fat versus low-fat diets; rather, preference is determined by what one has been exposed to and is familiar with. Mattes thinks this is an important point to address given that some people suggest there is a special quality about salt, fat, or sugar that may be driving eating behavior.
Rolls replied, “The primary determinant of reward value is what comes into the mouth in tiny quantities.” Based on sham feeding experiments with rats, animals show a preference for sucrose even when no food is reaching the gut. If the concentration of sucrose is increased, consumption will also increase until it reaches a sickly sweet point at which consumption drops. Rolls said these kinds of results are essential to understanding how rewards guide behavior (see Rolls, 2014). Tiny quantities of substance in the mouth act as a potent reward signal. In the gut, on the other hand, large quantities are needed to produce a reward. That said, although fundamental reward selection is produced by sensory receptors in the mouth, preferences can be conditioned by postingestive consequences such as sickness or by the metabolic energy value of food.
Dana Small agreed that sensory information is critical in determining behavior. But the reason it is critical, in her opinion, is that it is a conditioned cue associated with postingestive metabolic effects. If a rat is exposed to artificial sweetener over time, especially in a hungry state, and learns that the artificial sweetener is not associated with a positive postingestive effect, the dopamine response to that artificial sweetener will disappear, and the rat will stop consuming it.
The Slow Process of Weight Gain: Implications for the Concept of “Food Addiction”
Mattes commented on the small percentage of people in the general population, about 5 percent, who would be classified by the Yale Food
Addiction Scale (YFAS) as “food-addicted,” compared with the 65 percent of the general population who are overweight or obese. Not only would only a small proportion of people who are overweight or obese be classified as food-addicted, but also, in his opinion, most people who are obese are not “gluttonous.” Most people reach that weight category by gaining half a kilo to a kilo per year. Addiction, on the other hand, refers to an overconsumption that Mattes termed “really remarkable.” Nor is weight gain in people who are obese steady. There are long periods of time when their weight is stable, and then it goes up, then another stable period, and so on.
Ashley Gearhardt responded that it is important to distinguish between someone who is addicted and the impact of an addictive substance on public health. With every addictive substance, only a small percentage of people, 5 to 10 percent, become fully addicted. If the focus was only on that small percentage, the actual public health cost would not be captured in any way, especially for legal, cheap, easily accessible, and heavily advertised substances. Most people who consume an addictive substance do not become addicted, but do overconsume and do have a tendency at times to overindulge in a way that is more likely to occur than with something that is not addictive. Gearhardt’s concern with ultra-processed foods is that if they have the capacity to trigger an addictive response, most people will show a subclinical level of overconsumption that will have a significant public health cost.
Is the Focus on Addiction Diverting Attention from Other Biological Processes at Play?
Laurette Dubé suggested that the focus on addiction may be distracting experts from considering and understanding all of the biological processes at play. For example, she emphasized the importance of reinforcement in how people react to food. “We learn to like what we are exposed to,” she said. Food is different from nicotine, opium, and other addictive substances in the sense that it is the only substance connected to sensory processing.
Food Versus Alcohol
When asked about what can be learned from alcohol given that it can be considered both a drug of abuse and a food, Dubé replied that alcohol has some of the same sensory aspects as food but that its “nutrient dimension” is much less complex. Rolls added that alcohol intake obviously is not driven by the energy one derives from drinking it. Charles O’Brien added that alcoholism is among the most highly gene-driven addictions and that some alcoholism in families is influenced by variants of the opioid receptor.
Mention of alcohol prompted an audience member to comment on a
study showing increased alcohol consumption in bariatric surgery patients 2 years after surgery. The phenomenon appears to be limited to gastric bypass patients. Gastric banding, a purely restrictive procedure that does not involve rewiring the gastrointestinal tract, has not been associated with increased uptake of alcohol. The researchers in the cited study interpreted their findings to mean that gastric bypass patients, but not gastric banding patients, experience more rapid alcohol absorption and speculated that perhaps a more rapid rate of absorption provides greater reward. The audience member asked the panel to comment on (1) whether they agreed with the interpretation that increased alcohol intake following gastric bypass surgery could be due to faster absorption of alcohol, and (2) why the phenomenon occurs 2 years, and not 1 year, after surgery.
Timothy Moran agreed that the researchers’ interpretation was one possibility and noted that with some other addictive compounds, the more rapidly they are absorbed, the more addictive they are. In terms of the time period, putting a large amount of calories into the intestine quickly immediately after gastric bypass produces a very negative sensation. Food consumption following surgery tends to increase significantly at about the 1-year mark. The 2-year time period may simply reflect the system being accustomed to that kind of more rapid delivery of calories.
Hisham Ziauddeen remarked that the case of gastric bypass is probably more complicated on several levels. For example, he noted all the hormonal changes that occur fairly early following the surgery and even before any weight loss. He suspects that much of what is going on during those 2 years is not yet understood. He mentioned recent studies reporting an increased risk of suicide in patients who had undergone gastric bypass surgery, which he believes may reflect more pathologies beyond what is happening in the gut or with substances.
The Food Environment on College Campuses
When children leave home and go to college, they generally gain weight. The panel was asked how they would advise college administrators to change the food environment on college and university campuses. Dubé commented on the complexity of the problem. There are ways to design the environment in a way that encourages healthier eating—for example, by not placing high-fat/high-sugar foods near the cashier—but such changes need to be made in light of what is necessary to maintain sufficient sales. Creating demand for healthier eating in the individual is also important. College is a highly stressful experience. Students need to be nurtured and educated about healthy eating. Then, at the policy level, there needs to be greater investment in those who are financially able to create healthy foods in a way that meets customer demand.
Rogers described the “freshman 15” as a natural experiment that demonstrates how eating and body weight are under environmental control. There is a long history of research in this area, he said. Such changes in weight are viewed as a function of how different the new environment is compared with the old, modified by individual experience. Rogers added that going to college is a major change in lifestyle that represents, for some people, a “step” in the stepwise trajectory to obesity.
Moran added that college cafeterias are very different from what they were like 40 years ago. Then, students were presented a tray and were served. Now, colleges are competing to provide extensive choices.
For Dubé, the complexity of the issue highlights the need for more evidence not just on the brain, food, or nutrition, but also on all of the complex factors that make up society. She said, “I would strongly promote the idea that the science and evidence that are needed need to be expanded to those domains in a very urgent manner.”
“Food Addiction” as a Risk Factor for Obesity
An audience member asked what weight of evidence would be needed to support the hypothesis that addiction is a risk factor for obesity. Gearhardt replied that at this point, scientists do not have enough evidence regarding the potential level of impact of an addictive-like process on obesity. The field is still in its infancy, she said. Results of some studies suggest a risk. But most of these studies are small, not nationally representative, and not longitudinal. In Gearhardt’s opinion, the field needs larger studies that are nationally representative and longitudinal. Ziauddeen cautioned, however, that scientists have a long way to go in terms of defining endpoints before those studies can be conducted.
Pursuing Pleasure or Avoiding Displeasure?
According to an audience member, experiments with rats have demonstrated that the drive to avoid displeasure is stronger than the drive to pursue pleasure. But is this true in humans? When patients with addictions of one form or another struggle with compulsive behaviors that enable them to avoid displeasure, what is the decision-making process that is occurring at that time? What is known about the avoidance of displeasure during the decision-making process that takes place in choosing between a healthy food and a high-fat/high-sugar food?
Avena replied that her work with rats has shown that animals actually are willing to inflict displeasure on themselves to get to M&Ms and other foods. But when restricted from sugar or highly palatable foods, they show behaviors suggesting that they are not happy. Rather, they show signs of
depression, anxiety, and distress; the symptoms are mild, but they are present. Is that displeasure perhaps fueling poor food choices the next time they have access to food? It is a good question, Avena said—one in need of empirical study.
Rolls noted (as in Rolls, 2014) that the field of neuroeconomics is beginning to tackle just these sorts of issues. That is, how are benefits and costs weighed during the decision-making process, what trade-off occurs in the brain between planning for the future and seeking short-term gain, and what is the genetic basis of impulsive decision making?
Ziauddeen mentioned George Koob’s description of the avoidance of the dysphoria of withdrawal as “the dark side of addiction” (Koob and Le Moal, 2005). It is an interesting idea to consider, he said. Studies with drugs have shown that continued drug seeking in people who are addicted to drugs helps ameliorate many negative effects associated with not taking drugs. This is true even with drugs like cocaine, which is not associated with prominent physical withdrawal. It is probably a relevant phenomenon to consider with food given the multiplicity of factors involved in eating behavior.
In Rogers’s opinion, a desire to eat chocolate probably is not overridden by concern that one may develop heart disease in 20 or 30 years. But it may be overridden by concern that one may gain some weight in the near future as a result. Rogers encouraged a closer examination of how the experience of eating and the experiences that people have immediately after they have eaten can be used to help gain control over what are perceived as problematic foods. For example, the sequence of events that some people experience when eating “forbidden” or “naughty” foods includes not only pleasure but also regret soon after having eaten. In the United Kingdom, efforts are under way to encourage healthy eating by communicating messages such as “eating healthy will make your skin look better.”
