Skip to main content Accessibility help
×
Home

Information:

  • Access
  • Cited by 11

Actions:

      • Send article to Kindle

        To send this article to your Kindle, first ensure no-reply@cambridge.org is added to your Approved Personal Document E-mail List under your Personal Document Settings on the Manage Your Content and Devices page of your Amazon account. Then enter the ‘name’ part of your Kindle email address below. Find out more about sending to your Kindle. Find out more about sending to your Kindle.

        Note you can select to send to either the @free.kindle.com or @kindle.com variations. ‘@free.kindle.com’ emails are free but can only be sent to your device when it is connected to wi-fi. ‘@kindle.com’ emails can be delivered even when you are not connected to wi-fi, but note that service fees apply.

        Find out more about the Kindle Personal Document Service.

        The determinants of food choice
        Available formats
        ×

        Send article to Dropbox

        To send this article to your Dropbox account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Dropbox.

        The determinants of food choice
        Available formats
        ×

        Send article to Google Drive

        To send this article to your Google Drive account, please select one or more formats and confirm that you agree to abide by our usage policies. If this is the first time you use this feature, you will be asked to authorise Cambridge Core to connect with your <service> account. Find out more about sending content to Google Drive.

        The determinants of food choice
        Available formats
        ×
Export citation

Abstract

Health nudge interventions to steer people into healthier lifestyles are increasingly applied by governments worldwide, and it is natural to look to such approaches to improve health by altering what people choose to eat. However, to produce policy recommendations that are likely to be effective, we need to be able to make valid predictions about the consequences of proposed interventions, and for this, we need a better understanding of the determinants of food choice. These determinants include dietary components (e.g. highly palatable foods and alcohol), but also diverse cultural and social pressures, cognitive-affective factors (perceived stress, health attitude, anxiety and depression), and familial, genetic and epigenetic influences on personality characteristics. In addition, our choices are influenced by an array of physiological mechanisms, including signals to the brain from the gastrointestinal tract and adipose tissue, which affect not only our hunger and satiety but also our motivation to eat particular nutrients, and the reward we experience from eating. Thus, to develop the evidence base necessary for effective policies, we need to build bridges across different levels of knowledge and understanding. This requires experimental models that can fill in the gaps in our understanding that are needed to inform policy, translational models that connect mechanistic understanding from laboratory studies to the real life human condition, and formal models that encapsulate scientific knowledge from diverse disciplines, and which embed understanding in a way that enables policy-relevant predictions to be made. Here we review recent developments in these areas.

Health nudge interventions to steer people into healthier lifestyles are increasingly applied by governments worldwide( 1 , 2 ). Nudges are approaches to law and policy that maintain freedom of choice, but which steer people in certain directions( 3 ); they consist of small yet relevant behavioural stimuli such as simplification of information and choices, framing and priming of messages, feedback to one's behaviour, defaults and reminders and similar behavioural cues. Much of the health burden is caused by modifiable behaviours such as smoking, unhealthy food consumption and sedentary lifestyles, but neither decades of health information and education, nor attempts at hard regulation (such as fat taxes or sugar taxes), nor voluntary self-regulation of industry have markedly promoted healthier lifestyles or helped to stop the rise of non-communicable diseases. At the same time, there is increasing evidence that the purposeful design of the living and consumption environments, the ‘choice architecture’, is a key to changing nutritional and activity patterns( 4 ) and to maintaining healthier lifestyles. There is mounting evidence for the usefulness of WHO's motto: ‘make the healthier choice the easy choice’, through easier access, availability, priming and framing( 5 ). More than 150 governments now use behavioral science, with an emphasis on nudges( 6 , 7 ). In these countries, nudging for health is regarded as an attractive option to make health policies more effective and efficient; a recent poll in six European countries found that health nudges are overwhelmingly approved by the people( 8 ). This is the backcloth against which we set out to test nudging tools that might be useful add-ons to traditional health policies.

However, to produce policy recommendations that are likely to be effective, we need to be able to make valid, non-trivial predictions about the consequences of particular behaviours and interventions. For this, we need a better understanding of the determinants of food choice. These determinants include dietary components (e.g. highly palatable foods and alcohol), but also diverse cultural and social pressures, cognitive-affective factors (perceived stress, health attitude, anxiety and depression), and familial, genetic and epigenetic influences on personality characteristics. Our choices are influenced by how foods are marketed and labelled and by economic factors, and they reflect both habits and goals, moderated, albeit imperfectly, by an individual understanding of what constitutes healthy eating. In addition, our choices are influenced by physiological mechanisms, including signals to the brain from the gastrointestinal tract and adipose tissue, which affect not only our hunger and satiety but also our motivation to eat particular nutrients, and the reward we experience from eating.

To develop the evidence base necessary for effective policies, we need to build bridges across different levels of knowledge and understanding. This requires experimental models that can fill in the gaps in our understanding that are needed to inform policy, translational models that connect mechanistic understanding from laboratory studies to the real life human condition, and formal models that encapsulate scientific knowledge from diverse disciplines and which embed understanding in a way that enables policy-relevant predictions to be made.

State-of-the-art

Although it seems self-evident that changes in body weight reflect the choices an individual makes about what food to eat, how much to eat and how much to exercise, the long-term balance between energy intake and energy output is mainly determined by interacting physiological systems. Since the discovery of leptin in 1994 and ghrelin in 1999, we have gained a partial mechanistic understanding of how homeostatic and hedonic influences are coded and how they impact on eating behaviour, and we have an emerging understanding of the mechanisms by which particular food constituents influence hunger and satiety. The strong evolutionary conservation of these mechanisms has meant that knowledge from animal models translates well into understanding of human physiology: for example, mutations in genes that affect signalling in these pathways have very similar effects in rodents and human subjects.

Animal studies and human genetics studies have also framed the contributions of genetic and epigenetic influences on body weight. Body weight in people is estimated (from twin studies) to be about 80 % heritable( 9 ) but the search for the genes responsible has (so far) revealed associations that account for only about 20 % of the inter-individual variation( 10 ). This has focused attention on other heritable mechanisms and particularly on the consequences of events in uterine and early post-natal life. Notably, stress and impaired nutrition during gestation and in early post-natal life are now known to have lifelong programming effects on physiology and metabolism.

Against this background of genetics and nurture, an individual's knowledge, preferences and behaviours, lifestyle and eating habits are all shaped by their environment. In our everyday consumption, we are far from rational agents; we do not use only evidence-based information when deciding which foods to buy, but are influenced by the wider information environment, which is shaped by cultural factors, including advertising and other media, and we are strongly influenced by earlier decisions and habits, even if these have not proven to be optimal.

Habits are preferences shaped by past choices. If dietary choices follow habitual patterns, then we need to understand how these arise. Children often have a say in what they eat (at school they often choose what to eat at lunch), but they may be unable to correctly assess the costs and benefits of different options. In that context, imitative or impulsive behaviour may dominate, making them vulnerable to peer pressure and the supply of food in their direct environment. Once habits are in place, they shape preferences and future choices. The habitual pattern of behaviour has implications for policy interventions: effective interventions must be continued for long enough to affect preferences in the longer run.

Emotional and environmental cues also have a large role. We are influenced by how product information is presented, even whether the name sounds healthy. At the point of purchase, a number of decision heuristics and biases undermine rational decision behaviour. The anchor effect leads us to overvalue the information we obtained first; the source effect draws greater attention to the source of information and leads to assumptions about its credibility that may be false; and herd behaviour makes us adopt products that others are purchasing. Furthermore, we are poor at estimating probabilities and objective risks; we overestimate our capacity for self-control, and underestimate the health risks associated with the choices we make. Conversely, we cheat in our mental book-keeping: ‘Today I ate too much, but I'll just eat less tomorrow’( 3 ). We tend to select current enjoyment (ice cream now) over conditions we wish for later (slim and fit), which behavioural economists explain in terms of the temporal discounting of future conditions( 11 ).

The decision-making situation has a large effect, as demonstrated in human ecology models. The triple A factors (affordability, availability and accessibility) have a major impact on decisions( 12 ), and help to explain the attitude–behaviour gap( 13 ). Marketers have long understood that how a product is positioned in the store (e.g. as a ‘stopper’ at eye level) has a major impact. The same is true for the perception of rapid availability (ready-to-eat dishes) and the brand's potential of reward. In fact, most preferences appear to be less stable than postulated in neo-classical models; many are formed at the place where the decision is made. This is why behavioral economists speak of constructive preferences.

Decision heuristics and biases apply in situations involving uncertainty, which is true of most real decision-making. In our everyday consumption we are far from rational (in the sense of following our best intentions). During the search phase of the consumption process, we only perceive selective product characteristics and because of our limited processing capacities, we restrict our search criteria to just a few (‘seven plus or minus two’). The presence of many alternatives is more likely to confuse us than to generate optimal decisions (choice overload or hyperchoice). Another key finding from behavioural economics is the power of default options, such as the standard menu in a cafeteria. People generally follow the default option, even when given an opt-out. This finding is robust in diverse decision areas as organ donation, purchase of organic apples and the use of green electricity, and across a wide range of methods (experiments, questionnaires, secondary evaluations). For this reason, a number of incentive systems have been developed based upon ‘hard’ and ‘soft’ defaults( 14 ).

Hedonic processes and reward are important drivers for our decisions and are strong enough to overrule homeostatic needs. Food selection and intake in human subjects is largely driven by an interaction of homeostatic control and reward signals. This interaction involves a complex involvement of higher cognitive functions including memory, learning and evaluation of different options.

In summary, we need to understand exactly what conscious and unconscious factors bias our choices and subvert our best intentions. We need to understand how our homeostatic and higher cortical processes support healthy eating, and how these mechanisms come to be undermined. Our policies on healthy eating must be framed in this setting if they are to be effective. It is also crucial to know what real individual responses to policy instruments and actions can be expected, and to customise our ‘policy toolbox’ accordingly.

The evidence-based policy approach, currently pursued at all policy levels, is based upon empirical data and valid models of behaviour and effect( 15 ). It relies on learning policy cycles of test–learn–adapt–share that tests policies in pilot applications and assesses their efficacy and cost-benefits before they are rolled out( 16 ). The most important policy measures are those that rely on optimized information (not more information, but more useful and intuitively understandable information). For an integrated, policy-focused understanding of food choices, we need to optimise information in four key areas: early life experiences; environmental factors and impulsive choice behaviour; emotions and decision making; and how choices change with age.

Early life experiences

Early life programming can influence stress responses, food choice and weight gain into adult life. The consequences of early life events for cardiovascular and weight-related morbidity have been studied in detail in the Dutch famine birth cohort, and are associated with changes in the methylation of certain genes in people conceived during the Hunger Winter of 1944–45( 17 ). However, even modest differences in food intake or food choices in early life may have lifelong repercussions, and the metabolic status of the mother during gestation influences the brain dynamics of the fetus( 18 ). Obesity is most prevalent in lower socio-economic groups, and this is likely to reflect genetics (assortative mating), epigenetics and environmental factors, including a childhood diet of energy-dense foods( 19 ).

Obesity has been rising among European children, and it disproportionately affects those in low socio-economic groups. However, we do not know the mechanistic link between stress and/or poor nutrition in early life and obesity in adult life, and in particular, we do not know whether this is mediated by programming effects on the reward systems that affect food choice in adult life. Understanding this is critical, for not only are children in low socio-economic groups most affected by obesity, but they are also particularly resistant to healthy food campaigns. In 2004, one London borough, after a healthy food campaign, introduced changes in the meals offered in primary schools, shifting from low-budget processed meals towards healthier options. The effect on educational outcomes was analysed using a difference in differences approach, using the neighbouring Local Education Authorities as a control group. Outcomes improved in English and Science, and authorised absences (linked to illness and health) fell by 14 %( 20 ). However, the children that benefited least were those from the lowest socio-economic groups; those most in need of support.

Stress in early life is also a concern, because it can have programming effects that heighten responsiveness to stress in adult life, contributing further to weight gain( 21 ). Stress is a feature of modern life, particularly in the workplace. Some people eat less when stressed, but most eat more: one large study over 19 years in more than 10 000 participants( 22 ) found that employees experiencing chronic work stress had a 50 % increased risk of developing central adiposity. How stress impacts on appetite and weight gain has been extensively studied in rodent models, which appear to mimic the human situation well. In rodents, whereas acute stress is anorexigenic, chronic stress can lead to weight gain( 23 ). Chronic stress is related to chronic stimulation of the hypothalamo–pituitary adrenal axis, comprising neuroendocrine neurons in the hypothalamus that regulate the secretion of adenocorticotrophic hormone from the anterior pituitary, which in turn regulates glucocorticoid secretion from the adrenal gland. The hypersecretion of glucocorticoids (cortisol in man, corticosterone in rodents) is implicated in obesity at several levels. Intake of high energy foods suppresses the hyperactivity of the hypothalamo–pituitary adrenal axis, leading to what has been called comfort eating. The underlying mechanisms are well established: glucocorticoids stimulate behaviours mediated by the dopamine reward pathway, resulting in increased appetite for palatable foods( 24 ); stress also releases endogenous opioids, which reinforce palatable food consumption and promote non-homeostatic eating. Conversely, comfort food ingestion decreases hypothalamo–pituitary adrenal axis activity( 25 ); thus if stress becomes chronic, then eating patterns become a coping strategy. Beyond stress, which affects most of the population at some time, about 7 % of the European population suffers from depression every year. A common symptom is an alteration in food intake, and this can result in a vicious circle of weight gain and depression( 26 ).

While early life experience has a major impact upon health throughout life, little is known about how stress, poor nutrition and metabolic challenges like gestational diabetes in early life influences later food selection and valuation, and this is key to defining the timing and nature of policy interventions.

Environmental factors, food reward and impulsive choice behaviour

Many aspects of modern diet might contribute to the obesity epidemic, including the composition and palatability of modern food, its availability and affordability, how it is marketed, the modern environment, contemporary food culture and gene–environment interactions. These impact on the reward component of eating that is key to impulsive choice behaviour; the behaviour that governs momentary choices to eat high or low energy foods. The motivation to eat competes with other motivations via a highly conserved neural circuitry, the reward circuitry. One key part of this is the nucleus accumbens, which integrates homeostatic, hedonic and cognitive aspects of food intake( 27 , 28 ), and this circuit involves the neurotransmitter dopamine. The nucleus accumbens receives a dense dopamine input from the ventral tegmental area. This does not code ‘reward’ in the sense of subjective pleasure; rather, it mediates incentive salience (attractiveness) and motivational properties of positive stimuli and events( 29 ). The dopamine system is regulated by cues that signal the availability of rewards as well as actual reward: dopamine neurons fire in a way that reflects the reward value and the dopamine that is released in the striatum has a key role in habit formation, while that released in the orbitofrontal cortex is involved in decision-making.

Human brain imaging studies using positron emission tomography and functional MRI (fMRI) confirm that these mechanisms function similarly in human subjects as in rodents. Thus the central nervous system response to palatable foods differs from that to bland foods and responses of subjects that crave palatable foods differ from those who do not. Importantly, cravings for palatable food activate similar brain regions and involve the same chemical messengers in human subjects as in rats. In the striatum, the availability of dopamine D2 receptors is reduced in severely obese subjects( 30 ), and people who show blunted striatal activation during food intake are at greater risk of obesity, particularly those with compromised dopamine signalling( 31 ).

Mammals pursue behaviour that is likely to yield them the greatest reward at that time; when fat stores are high, the rewarding power of food is less and they are more motivated to pursue other rewards. Thus hedonic and homeostatic mechanisms interact, and this takes place at defined brain sites. Importantly, endocrine signals such as ghrelin, insulin and leptin are not merely regulators of energy homeostasis, but also influence the reward circuitry to increase the incentive value of food( 32 34 ) and impulsive choice behaviour( 35 ). The consequences are striking: the one intervention of consistent effectiveness for weight loss in the morbidly obese is bariatric surgery and this works not by restricting intake or absorption, but by reducing the incentives to eat via changes in endocrine signalling to the brain( 36 , 37 ). This shows that morbid obesity is resistant to interventions because of a dysfunction of gut-brain signalling and is important for policy. Blame and shame strategies that deny the underlying pathology are destined to be ineffective, and may be counterproductive by promoting comfort eating. It is also important because these endocrine signals vary with time of day and according to the timing of meals. This opens a window of opportunity by which changing meal patterns, when we eat rather than how much, can influence both how we utilise the energy intake and our appetite.

Emotions and decision-making

Eating is triggered by many factors, including the sight, smell and memory of food, and anticipation of food is associated with activation of well-defined regions of the hypothalamus( 38 ). The sensory characteristics of food are also important in food choice, and these can be well studied by fMRI( 39 ). Visual attention can be rapidly cued by food items, particularly items with high calorific content, and attentional responding to these is magnified in overweight individuals, suggesting that heightened attention to high-energetic food cues promotes greater intake. Animal studies also indicate a major role for learning; associations are formed between the sensory characteristics of a food and its post-ingestive effects. Over time, these generate flavour preferences, and may also control meal size.

The sight of appetizing food modulates brain activity in consistent ways: viewing food items enhances activation both in visually-related brain regions and in regions associated with reward (orbitofrontal cortex, parahippocampal gyrus and the insula) in both adults and children( 40 , 41 ). Visually-driven responses to food are linked to increased connectivity between the ventral striatum, the amygdala and anterior cingulate in individuals at risk of obesity, hence differences in interactions within the appetitive network may determine the risk of obesity. Obese participants show greater visually-driven responses to food in reward-sensitive brain regions and, for obese individuals, greater responsiveness in these regions before weight-loss treatment predicts treatment outcome. Poor weight loss is also predicted by pre-treatment levels of activity to food stimuli in brain areas associated with visual attention and memory, consistent with the attentional effects of food being a predictor of weight loss success( 42 ).

However, we have a poor understanding of how valuation and selection of food are encoded neuronally. The orbitofrontal cortex, dorsolateral prefrontal cortex and ventral striatum are all implicated, but we have limited knowledge of what neuronal mechanisms are subserved by these structures. If we are to use functional neuroimaging studies to inform policies that promote healthier food choices, we need a better understanding of how health interventions impact on the brain mechanisms that control food selection and valuation. We need to address how molecular and cellular events, initiated by the exposure to food, translate into changes at the neuronal circuit level and how these translate to food decisions.

Physiological mechanisms of appetite control

In all mammals, appetite and energy expenditure are regulated by conserved neuronal circuitry using common messengers. Ghrelin, secreted from the empty stomach, reaches high levels after a fast and activates neurons in the hypothalamus that make a potent orexigen, neuropeptide Y. Leptin, secreted by adipocytes, reports on the body's fat reserves; it inhibits neuropeptide Y neurons, while activating others that express anorexigenic factors, notably neurons that express pro-opiomelanocortin. Pro-opiomelanocortin neurons and neuropeptide Y neurons are reciprocally linked, and which population is dominant determines how much (on average) an animal will eat. As an animal eats, neural and endocrine signals from the gut report on the volume ingested and on its composition, including its complement of fat, carbohydrates and protein. These signals, relayed by satiety centres of the caudal brainstem, converge on the ghrelin and leptin sensing circuits of the hypothalamus( 43 ). These in turn project to other limbic sites, including the paraventricular nucleus, which is the primary regulator of the sympathetic nervous system, and which also regulates the hypothalamo–pituitary adrenal axis. These pathways are powerful moderators of energy intake. Despite huge variations in day-by-day food intake, in the long term, the body weight of most individuals is remarkably stable. However, ‘crash dieting’ is an example of an intervention that reduces body weight in the short term, but as a result of the disruption of normal homeostatic mechanisms it has counterproductive effects in the long term.

It seems that dietary decisions can be regulated by circulating metabolic hormones, including those that signal to brain areas involved in food intake and appetitive behaviours. One example is ghrelin, an orexigenic hormone that increases anticipatory( 44 ) and motivated behaviour for food, notably for fat( 45 ) and sugar( 46 ). Ghrelin enhances the reward value of foods and hence increases their consumption( 32 ). Recently, ghrelin has been shown to guide dietary choice, but not entirely as expected for a reward-promoting hormone. For example, rats offered a free choice of lard (100 % fat), sucrose and chow increased their lard consumption over 2 weeks; ghrelin administration changed this food choice and they started to consume chow. Interestingly, these effects of ghrelin diverge from those of fasting, after which the consumption of energy-dense foods is prioritised( 47 ). The pathway from the ventral tegmental area to the nucleus accumbens appears to be engaged by ghrelin to change food choice( 47 ) and reward-linked behaviour( 48 ). Several other gut- and fat-derived hormones also impact on food reward circuitry. Leptin, for instance, affects food reward encoding by dopamine neurons of the ventral tegmental area( 49 ).

While morbid obesity is characterised by dysfunctional gut–brain signalling, a key stage in the progression to obesity is the development of leptin resistance. As a consequence, dietary restriction has a limited effect on obesity; long term compliance is poor, and those who lose weight are likely to swiftly regain it and may even overshoot after the end of a diet. Normally, eating is most rewarding when there is energy deficiency, and least in an energy-replete state, but leptin resistance develops in both the appetite circuitry and in the reward circuitry, so food remains rewarding despite a state of energy excess. Imaging studies have confirmed the impact of hormones in the recruitment of both hypothalamic and reward circuits. For example, when subjects are infused with peptide YY (a postprandial gut-derived satiety factor) the changes in activity in the caudolateral orbital frontal cortex predict feeding, whereas when levels of peptide YY are low, hypothalamic activation predicts food intake( 50 ). Insulin, which is released in the periphery after food ingestion is also a potent modulator of brain activity. In recent years it has become clear that, just as peripheral insulin resistance develops in association with obesity, so does insulin resistance in the brain( 51 ).

Thus, paradoxically, one of the strongest predictors of weight gain is weight loss dieting. One of the biggest studies to demonstrate this was the Growing Up Today Study, a prospective study of >16 000 adolescents( 52 ). At the 3-year follow-up, adolescents that were frequent or infrequent dieters had gained significantly more weight than non-dieters. The study controlled for BMI, age, physical development, physical activity, energy intake and height change over the period. The longest study that demonstrates this is Project EAT (Eating and Activity in Teens and Young Adults), a population-based study of middle and high school students( 53 ). This study, which controlled for socio-economic status and initial BMI, again showed that the strongest predictors of weight gain were dieting and unhealthy weight control behaviours. The behaviours associated with the largest increases in BMI over a 10-year period were skipping meals, eating very little, using food substitutes and taking ‘diet pills’.

This raises the concern that emphasising the health risks of obesity may lead to behaviours that exacerbate the problem. This worry is compounded when one looks at the media response in the UK to recent publicity, where concerns about the effects of excessive weight gain in pregnancy were translated as concern about obesity in pregnancy. These are very different; while excessive weight gain in pregnancy is detrimental, so is weight loss, even from a condition of obesity. Physiologically, dietary restriction during pregnancy can lead to starvation of the fetus, as homeostatic mechanisms defend maternal body weight at the expense of the fetus. Thus, how advice related to healthy eating and lifestyles is formulated and disseminated needs careful attention. There has been little work on food choice in children, and this is important to explore because of the weaker self-control capacity of children, which is coupled to the maturation of their prefrontal cortex( 54 ). This has a bearing on in-store marketing (and legislation on that) and the development of interventions aimed at preventing childhood obesity.

The neuroimaging of food choice

Human associational and behavioural studies have many potential confounding factors, so interpreting them depends on inferences from our understanding of the neurobiology of appetite. However, there is a disconnect between our mechanistic understanding and our ‘softer’ knowledge of individual consumer behaviour, which makes these inferences unsafe. We need to create bridges in our understanding, enabling us to integrate behavioural and observational studies with neurobiological studies in a way that can be used to educate stakeholders and inform policy.

Human neuroimaging is an emerging technology that can be used to define the neural circuits involved in food valuation and selection. Food decision-making has been studied surprisingly little; most neuroimaging studies use passive viewing paradigms in which participants are exposed to food: they study food cue reactivity rather than the ensuing decision-making processes. Combining different imaging techniques can optimise the temporal and spatial description of the neuronal circuits underlying food valuation and selection during hunger and satiety. Recent developments in fMRI include (a) combining diffusion tensor imaging with resting state analysis to determine network structures and changes during different physiological states; (b) high-resolution anatomical MRI to improve investigation of hypothalamic and midbrain responses; and (c) arterial spin labelling techniques to establish a quantitative neural activity measure of hunger and satiety. In addition, developments in magnetoencephalography and electroencephalography include: extraction of resting state dynamics with high temporal resolution and combination with diffusion tensor imaging; and application of Bayesian-based source localisation to define the temporal and spatial network involved in food selection. Most fMRI studies that link a given brain circuit with cues associated with food or with the choice for a particular food are based on correlations between an event and a recorded brain activity. To determine causality, we need to be able to change brain activity and determine its impact on behaviour. In human subjects, defined neuronal structures can be manipulated using transcranial magnetic stimulation or direct current stimulation to either facilitate or attenuate cerebral activity.

Along with the rise in the number of neuroimaging studies there have been many neuroimaging data-sharing initiatives, and several databases contain resting fMRI data and anatomical MRI data from thousands of individuals. For functional imaging, things are more complicated but there are notable efforts of sharing fMRI datasets (openfmri.org), unthresholded statistical maps (neurovault.org) and coordinate-based data synthesis (neurosynth.org). However, the value of such databases depends on the available metadata, and existing databases lack most or all of the metadata necessary for research on food choice, such as weight( 54 ), restraint eating status( 55 ) and personality characteristics( 56 ).

For policies to be built on robust evidence, it is essential that the evidence is developed in a way that facilitates meta-analysis. There is great variability in neuroimaging results, and this is especially true for fMRI tasks involving complex stimuli such as food stimuli( 40 , 41 ). Bennett & Miller( 57 ) showed that the reproducibility of fMRI results was only 50 %, even for the same task and stimuli in the same participants. This was confirmed by a meta-analysis of fMRI studies of responses to food pictures: measurements for the brain areas that were most consistently activated by looking at food v. non-food pictures were only reported in fewer than half of the studies included( 40 , 41 ). Reproducibility can be improved by standardising measures, but there are no standardised fMRI protocols for assessing food responsivity and food choice for different food categories. To filter out effects due to subject characteristics rather than methodological difference, standardisation of instruments and measures is crucial for data sharing and pooling across studies( 58 ). Recently, researchers have begun to share (standardised) food images for use in experimental paradigms (e.g. 59 , 60 ) and tools for standardised collection of food-related subject characteristics( 61 ).

To connect data from human imaging studies with neurophysiological data from rats, we must improve and adapt high-field rodent fMRI technology in a setting that allows to map involvement of neural circuits in food valuation and selection. Small rodent resting state and pharmacological fMRI is an emerging technology that has not yet been applied to address how brain activity changes upon food restriction and food anticipation. Thus, it is not known, for example, how brain activity is changed upon food restriction in rodents or how gut peptides like leptin and ghrelin affect functional connectivity between brain regions. Small rodent fMRI bridges the gap between neuronal activity at the cellular level with fMRI measures in human subjects, making it possible to connect molecular and cellular data with fMRI measures.

Novel technologies to understand the brain mechanisms underlying food choice

There is a poor understanding of what underlies the responses quantified in neuroimaging studies. By combining in vivo electrophysiology with optogenetics or pharmacogenetics, it is now possible to record from and interfere with defined neurons involved in food valuation and choice, and this is key to unravelling what underlies the responses recorded by neuroimaging. Optogenetics takes advantage of genes that encode light-sensitive channels and these channels can be expressed conditionally in specific neurons. These neurons can then be either activated or inhibited by shining light on them. This technical approach requires that these neurons express the cre recombinase enzyme. Targeting cre for instance to tyrosine hydroxylase (the rate limiting enzyme for dopamine production) neurons such as in (germline) tyrosine hydroxylase-cre rats, allows these light-sensitive channels to be expressed only in midbrain dopamine neurons. To achieve this, light-sensitive channels are cloned into a recombinant viral vector such that, only upon expression of cre, the channels are expressed in dopamine neurons( 62 , 63 ). This makes it possible to activate precise populations of neurons in rodents, and to compare observations with brain responses observed by neuroimaging. Similarly, subpopulations of dopamine neurons can be targeted with viruses to express novel receptors that are not endogenously present; these can then be specifically activated (or inhibited) by systemically applied drugs that act on those novel receptors (e.g. 64 ).

How the life-long learning process contributes to food selection and valuation

The sensory characteristics of food are important in food choice, but learning also has a major role( 65 ). Associations are formed between the sensory characteristics of a food (the conditioned stimulus) and its post-ingestive effects (the unconditioned stimulus). Over time, these flavour-nutrient associations generate flavour preferences and they also control meal size. In human subjects, fundamental questions remain about the nature of the unconditioned stimulus and how this is combined with sensory signalling from the tongue to the brain.

In adult human subjects, flavour-nutrient learning is notoriously difficult to observe under controlled laboratory conditions, although in non-human animals this form of learning is extremely reliable. Several examples of flavour-nutrient learning have been reported in children, and this may be because most dietary learning occurs in early life. By adulthood, we have encountered so many foods and flavours that our capacity to learn new associations might be saturated. If so, this reinforces the importance of childhood as a critical period during which our dietary behaviours are established. A further consideration is the complexity of the modern Western dietary environment. Human subjects are now exposed to a wide range of foods, in numerous different brands and varieties. This may limit our opportunity to learn about individual foods, which has the potential to promote overconsumption( 66 ).

Learned beliefs impact our dietary choices directly. Typically, we decide how much we are going to eat before a meal( 67 ). These decisions are often motivated by a concern to avoid hunger between meals, and the learned expected satiety of individual foods is important in this. Low energy-dense foods tend to have greater expected satiety, and such foods are often selected to avoid hunger between meals. Increasingly, portions are also determined by external agents such as restaurants or retailers. Recently, it has become clear that larger portions not only increase our food intake but also affect choice. This is because larger portions are likely to satisfy our appetite between meals and, in the absence of concerns about satiety, decisions tend to be motivated primarily by palatability.

A further possibility is that satiation or the absence of hunger between meals is itself valued( 68 ). The results of human appetite studies suggest that both oral and gastric stimulation are needed for optimal satiety( 69 71 ). However, the underlying process also involves integration of explicit knowledge about the food and amount that has been consumed( 72 , 73 ). Consistent with this, several studies show that satiety and satiation are reduced when eating occurs in the presence of cognitive distraction( 74 ). Eating ‘attentively’ appears to have the opposite effect( 75 ), and food properties like viscosity can increase perceived fullness for otherwise similar foods( 76 ). Despite its importance, the process by which interoceptive signals are integrated remains unclear. This merits attention because some studies indicate that differences in interoceptive awareness are a predictor of adiposity in human subjects( 77 ).

How physiological, psychological and emotional factors predispose people to unhealthy eating

One key question in the effects of sensory, nutrient and satiety contributions to reward is whether the initial response to certain stimuli remains after repeated exposure. Does the response to a low-energy beverage with artificial sweeteners stay the same, or do people slowly learn that ‘diet’ products contain lower energy content? For this case, it is hard to demonstrate such dietary learning( 78 ) although there is some evidence for detection of food energy content in the mouth( 79 , 80 ). Another important consideration is whether it makes a difference whether one goes from, for example, 836·8 → 209·2 kJ (200 → 50 kcal), or from 627·6 → 0 kJ (150 → 0 kcal) In both cases, there is a reduction of 627·6 kJ (150 kcal), but in the case of 836·8 → 209·2 kJ (200 → 50 kcal), there is still energy left in the stimulus, whereas in the case of 627·6 → 0 kJ (150 → 0 kcal), there is no energy left. It has been argued that the absence of any energy content will lead to a lower reward value after repeated exposure. Conversely, most ‘light'products still contain energy, albeit less than their regular counterparts, with soft drinks a notable exception.

In both human subjects and rodents, the motivation to choose one food over another is driven by the emotional, hedonic and metabolic properties of the foods. The dopamine system is critically involved in this, and is essential for associating rewards with environmental stimuli that predict these rewards. Activity of this system is affected by both metabolic information and emotional and cognitive information. The hypothalamus, amygdala and medial prefrontal cortex play important roles in, respectively, feeding behaviour, emotional processing and decision-making. Manipulation of the dopamine system can be achieved by nutritional interventions and reducing dopamine levels in lean and obese subjects leads to decreased activity in the reward system( 81 ).

There is also evidence that incidental emotions can affect food choices. Sadness leads to greater willingness to pay for unnecessary consumer goods( 82 , 83 ) and increased consumption of unhealthy food items( 84 ). However, the biological mechanisms linking affective states to food choices are unknown. Recent work has begun to investigate the underlying neural mechanisms of dietary choice in human subjects using neuroimaging and brain stimulation techniques together with validated choice paradigms and behavioural trait measures (e.g. 84 88 ).

A natural assumption would be that the physiological and psychological reactions to an affective state use the same neural pathways to influence food choices. However, Maier et al. ( 24 ) have recently shown using fMRI that experiencing an acute stressor leads to changes in two separate and dissociable neural pathways: one associated with the physiological reaction to stress, and the other with the conscious perception of being stressed. The physiological response was measured by sampling salivary cortisol, the psychological experience was recorded using a visual analog scale on which participants indicated how they felt right after the stress induction. Cortisol was associated with signals about the reward value of food: individuals with a higher cortisol response showed a higher representation of taste in the ventral striatum and amygdala, and amplified signalling between ventral striatum/amygdala and the ventromedial prefrontal cortex when a tastier food was chosen. Yet the subjective perception of being stressed did not correlate with the strength of this connection. Instead, the perceived stress level (but not the cortisol reaction) was associated with the connectivity strength between left dorsolateral prefrontal cortex and the ventromedial prefrontal cortex: the more stressed participants had felt, the weaker was the connectivity between these two regions when self-control was needed to overcome taste temptations in order to choose the healthier food. A series of studies have demonstrated that connectivity between dorsolateral prefrontal cortex and ventromedial prefrontal cortex relates to the degree to which individuals use self-control in dietary choice( 89 92 ). This connection in the prefrontal cortex may maintain a goal context that promotes focusing on long-term outcomes such as future health, whereas sensory and motivational signalling from subcortical areas may promote information about more immediate choice outcomes. Thus, self-control in dietary choice may depend on a balance of signalling and information exchange in value computation networks and disruptions to this balance during highly affective states may lead to impaired self-control.

Modeling the interactions between physiological, psychological and emotional factors related to feeding behaviour

An ultimate ambition must be to generate formal models that encapsulate scientific knowledge from diverse disciplines, and which embed understanding in a way that enables policy-relevant predictions to be made. Modelling is a natural way of working together to provide added value; it expresses intrinsically the need to make links between levels of understanding. Most importantly, it takes seriously the issue of how to generate policy guidelines that have a robust scientific basis, by providing a common framework of understanding across disciplines.

Modelling provides a logically coherent framework for a multi-level analysis of food choice, integrating measures of the neural components of the appetitive network with whole-system output (behavioural experiments) in a framework consistent with the neural homeostatic and hedonic mechanisms, and providing a test-bed for studies of behavioural interventions. The first phase in modelling is a scheme that embodies constructs that explain behavior by describing a causal chain of events. A computational model expresses these mathematically, usually as differential equations. Typically such differential equations are (a) coupled (expressing interdependence between factors) and (b) non-linear (expressing complex dependencies between variables). To be useful, a model must be developed at a level of detail appropriate for the data it is informed by, and the type of prediction that it is called upon to make. It must be complex enough to satisfy the former, but simple enough to satisfy the latter: making models over-complex is counterproductive, as such models are not predictive( 93 ).

For example, oxytocin neurons are well established as playing an important role in satiety( 94 , 95 ) and, according to recent studies, in food choice( 96 , 97 ). These neurons respond to signals from the gut that control meal size, and exactly how they respond has been analysed at the single-cell level. Their behaviour can be captured in detail by biophysical (Hodgkin–Huxley style) models, that can then be approximated by simpler models that capture the essential behaviour while being better suited for modelling networks of neurons( 98 ). Decision making at the level of the neuron networks that oxytocin engages can be modelled by biologically realistic ‘winner-takes-all’ networks, which provide predictive models of how continuous variables lead to categorical decision making, and such network models can be fit to human brain imaging data by mean field approximation( 99 ). Such models can link brain imaging data with experimental behavioural data in a predictive way, as in the spiking search over time & space model that has been developed to analyse attentional processes( 100 ). Relatively simple mathematical models can capture important features of value-based decisions well, and in a similar way for food-based decisions as for social decisions, indicating that there is a common computational framework by which different types of value-based decisions are made( 101 ). At a high level, the aim must be to generate agent-based models that describe by a set of explicit rules all the factors that influence food choice, validating each rule by a mechanistic understanding of the neurobiological and physiological mechanisms that implement these rules. It is a long goal, but working towards it provides a unified framework for multi-disciplinary research.

Conclusions

Clearly we need a more sophisticated understanding of the determinants of food choice, an understanding consistent with many different types of evidence. To translate this into policy recommendations will involve further challenges: we must be aware of the potential for unintended consequences, of the likely need for policies tailored to specific populations, and of the difficulties in achieving compliance and measuring outcomes. The nudge approach to behavioural change appears at present to be most likely to be fruitful; small interventions that can be trialled for effectiveness in controlled settings. To develop these policy tools we need to identify a set of specific proposed interventions that are aimed at particular target groups. We must identify the evidence that suggests that these will be effective, and identify the gaps in our knowledge that may make our predictions uncertain before deciding which interventions to trial, and exactly how to implement them.

Financial Support

The research leading to these results has received funding from the European Union's Seventh Framework programme for research, technological development and demonstration under grant agreement no 607310 (Nudge-it).

Conflict of Interest

Peter Rogers has received funding for research on food and behaviour from industry, including from Sugar Nutrition UK. He has also provided consultancy services for Coca-Cola Great Britain and received speaker's fees from the International Sweeteners Association.

Authorship

All authors participated in writing this review.

References

1. Halpern D for VicHealth (2016) Behavioural Insights and Healthier Lives. Melbourne: Victorian Health Promotion Foundation.
2. Matjasko, JL, Cawley, J, Baker-Goering, MM et al. (2016) Applying behavioral economics to public health policy: illustrative examples and promising directions. Am J Preventive Med 50, S13S19.
3. Sunstein, CS & Thaler, RH (2008) Nudge: Improving Decisions About Health, Wealth, and Happiness, p. 312. New Haven and London: Yale University Press.
4. Bucher, T, Collins, C, Rollo, ME et al. (2016) Nudging consumers towards healthier choices: a systematic review of positional influences on food choice. Br J Nutr 115, 22522263.
5. Wilson, AL, Buckley, E, Buckley, JD et al. (2016) Nudging healthier food and beverage choices through salience and priming. Evidence from a systematic review. Food Quality and Preference 51, 4764.
6. Ly, K & Soman, D (2013) Nudging Around the World (Research Report Series). Retrieved from the Rotman School of Management, University of Toronto. http://inside.rotman.utoronto.ca/behaviouraleconomicsinaction/files/2013/12/Nudging-Around-The-World_Sep2013.pdf.
7. Sunstein, CR (2016 b) The council of psychological advisers. Ann Rev Psychol 67, 713737.
8. Reisch, LA & Sunstein, CR (2016) Do Europeans like nudges? J Judgment Decision Making 11, 310325.
9. Wardle, J, Carnell, S, Haworth, CM et al. (2008) Evidence for a strong genetic influence on childhood adiposity despite the force of the obesogenic environment. Am J Clin Nutr 87, 398404.
10. Locke, AE, Kahali, B, Berndt, SI et al. (2015) Genetic studies of body mass index yield new insights for obesity biology. Nature 518, 197206.
11. Benabou, R & Tirole, J (2005) Incentives and prosocial behavior. Am Economic Rev 96, 16521678.
12. Maas, J, de Ridder, DT, de Vet, E et al. (2012) Do distant foods decrease intake? The effect of food accessibility on consumption. Psychol Health 27, Suppl. 2, 5973.
13. Reisch, LA & Gwozdz, W (2013) Smart defaults and soft nudges. How insights from behavioral economics can inform effective nutrition policy. In Marketing, food, and the consumer. Festschrift in Honour of Klaus Grunert, pp. 189200 [Brunsø, K & Scholderer, J, editors]. New Jersey: Pearson Publisher.
14. Sunstein, CR & Reisch, LA (2014) Automatically green: behavioral economics and environmental protection. Harvard Environ Law Rev 38, 127158.
15. Bogenschneider, K & Corbett, TJ (2010) Evidence-Based Policymaking: Insights from Policy-Minded Researchers and rEsearch-Minded Policymakers, p. 368. Oxford: Routledge.
16. Joint Research Center of the EC (2016). Behavioural insights applied to policy. European Report Brussels. JRC.
17. Tobi, EW, Goeman, JJ, Monajemi, R et al. (2014) DNA methylation signatures link prenatal famine exposure to growth and metabolism. Nat Commun 5, 5592.
18. Linder, K, Schleger, F, Kiefer-Schmidt, I et al. (2015) Gestational diabetes impairs human fetal postprandial brain activity. J Clin Endocrinol Metab 100, 40294036.
19. Wang, Y & Lim, H (2012) The global childhood obesity epidemic and the association between socio-economic status and childhood obesity. Int Rev Psychiatry 24, 176188.
20. Belot, M & James, J (2011) Healthy school meals and educational outcomes. J Health Econ 30, 489504.
21. Brunton, PJ (2015) Programming the brain and behaviour by early-life stress: a focus on neuroactive steroids. J Neuroendocrinol 27, 468480.
22. Brunner, EJ, Chandola, T & Marmot, MG (2007) Prospective effect of job strain on general and central obesity in the Whitehall II study. Am J Epidemiol 165, 828837.
23. Rabasa, C & Dickson, SL (2016) Impact of stress on metabolism and energy balance. Curr Opin Behavl Sci 9, 7177.
24. Maier, SU, Makwana, AB & Hare, TA (2015) Acute stress impairs self-control in goal-directed choice by altering multiple functional connections within the brain's decision circuits. Neuron 87, 621631.
25. Dallman, MF, Pecoraro, N, Akana, SF et al. (2003) Chronic stress and obesity: a new view of ‘comfort food’. Proc Natl Acad Sci USA 100, 1169611701.
26. Stunkard, AJ, Faith, MS & Allison, KC (2003) Depression and obesity. Biol Psychiatry 54, 330337.
27. Kelley, AE, Baldo, BA, Pratt, WE et al. (2005) Corticostriatal-hypothalamic circuitry and food motivation: integration of energy, action and reward. Physiol Behav 86, 773795.
28. Adan, RA, Vanderschuren, LJ & la Fleur, SE (2008) Anti-obesity drugs and neural circuits of feeding. Trends Pharmacol Sci 29, 208217.
29. Berridge, KC (2007) The debate over dopamine's role in reward. Psychopharmacology 191, 391431.
30. Benton, D & Young, HA (2016) A meta-analysis of the relationship between brain dopamine receptors and obesity: a matter of changes in behavior rather than food addiction? Int J Obes (Lond) 40, S12S21.
31. Stice, E, Spoor, S, Bohon, C et al. (2008) Relation between obesity and blunted striatal response to food is moderated by TaqIA A1 allele. Science 322, 449452.
32. Egecioglu, E, Jerlhag, E, Salomé, N et al. (2010) Ghrelin increases intake of rewarding food in rodents. Addict Biol 15, 304311.
33. Figlewicz, DP, MacDonald Naleid, A & Sipols, AJ (2007) Modulation of food reward by adiposity signals. Physiol Behav 91, 473478.
34. Kullmann, S, Heni, M, Veit, R et al. (2015) Selective insulin resistance in homeostatic and cognitive control brain areas in overweight and obese adults. Diabetes Care 38, 10441050.
35. Anderberg, RH, Hansson, C, Fenander, M et al. (2016) The stomach-derived hormone ghrelin increases impulsive behavior. Neuropsychopharmacology 41, 11991209.
36. Leng, G (2014) Gut instinct: body weight homeostasis in health and obesity. Exp Physiol 99, 11011103.
37. Frank, GK, Oberndorfer, TA, Simmons, AN et al. (2008) Sucrose activates human taste pathways differently from artificial sweetener. Neuroimage 39, 15591569.
38. Johnstone, LE, Fong, TM & Leng, G (2006) Neuronal activation in the hypothalamus and brainstem during feeding in rats. Cell Metab 4, 313321.
39. Lundström, JN, Boesveldt, S & Albrecht, J (2011) Central processing of the chemical senses: an overview. ACS Chem Neurosci 2, 516.
40. Van Meer, F, van der Laan, LN, Adan, RA et al. (2015) What you see is what you eat: an ALE meta-analysis of the neural correlates of food viewing in children and adolescents. Neuroimage 104, 3543.
41. van der Laan, LN, de Ridder, DT, Viergever, MA et al. (2011) The first taste is always with the eyes: a meta-analysis on the neural correlates of processing visual food cues. NeuroImage 55, 296303.
42. Hege, MA, Stingl, KT, Ketterer, C et al. (2013) Working memory-related brain activity is associated with outcome of lifestyle intervention. Obesity (Silver Spring) 21, 24882494.
43. Murphy, KG & Bloom, SR (2006) Gut hormones and the regulation of energy homeostasis. Nature 444, 854859.
44. Verhagen, LA, Egecioglu, E, Luijendijk, MC et al. (2011) Acute and chronic suppression of the central ghrelin signaling system reveals a role in food anticipatory activity. Eur Neuropsychopharmacol 21, 384392.
45. Perello, M & Dickson, SL (2015) Ghrelin signalling on food reward: a salient link between the gut and the mesolimbic system. J Neuroendocrinol 27, 424434.
46. Skibicka, KP, Hansson, C, Alvarez-Crespo, M et al. (2011) Ghrelin directly targets the ventral tegmental area to increase food motivation. Neuroscience 180, 129137.
47. Schéle, E, Bake, T, Rabasa, C et al. (2016) Centrally administered ghrelin acutely influences food choice in rodents. PLoS ONE 11, e0149456.
48. Skibicka, KP, Shirazi, RH, Rabasa-Papio, C et al. (2013) Divergent circuitry underlying food reward and intake effects of ghrelin: dopaminergic VTA-accumbens projection mediates ghrelin's effect on food reward but not food intake. Neuropharmacology 73, 274283.
49. van der Plasse, G, van Zessen, R, Luijendijk, MC et al. (2015) Modulation of cue-induced firing of ventral tegmental area dopamine neurons by leptin and ghrelin. Int J Obes (Lond) 39, 17421749.
50. Batterham, RL, Ffytche, DH, Rosenthal, JM et al. (2007) PYY modulation of cortical and hypothalamic brain areas predicts feeding behavior in humans. Nature 450, 106109.
51. Kullmann, S, Heni, M, Hallschmid, M et al. (2016) Brain insulin resistance at the crossroads of metabolic and cognitive disorders in humans. Physiol Rev 96, 11691209.
52. Field, AE, Austin, SB, Taylor, CB et al. (2003) Relation between dieting and weight change among preadolescents and adolescents. Pediatrics 112, 900906.
53. Neumark-Sztainer, D, Wall, M, Story, M et al. (2012) Dieting and unhealthy weight control behaviors during adolescence: associations with 10-year changes in body mass index. J Adolescent Health 50, 8086.
54. van Meer, F, Charbonnier, L & Smeets, PA (2016) Food decision-making: effects of weight status and age. Curr Diab Rep 16, 84.
55. van der Laan, LN, Charbonnier, L, Griffioen-Roose, S et al. (2016) Supersize my brain: a cross-sectional voxel-based morphometry study on the association between self-reported dietary restraint and regional grey matter volumes. Biol Psychol 117, 108116.
56. van der Laan, LN & Smeets, PA (2015) You are what you eat: a neuroscience perspective on consumers’ personality characteristics as determinants of eating behavior. Curr Op Food Sci 3, 1118.
57. Bennett, CM & Miller, MB (2010) How reliable are the results from functional magnetic resonance imaging? Ann NY Acad. Sci 1191, 133155.
58. Smeets, PA, Charbonnier, L, van Meer, F et al. (2012) Food-induced brain responses and eating behavior. Proc Nutr Soc 71, 511520.
59. Charbonnier, L, van Meer, F, van der Laan, LN et al. (2016) Standardized food images: a photographing protocol and image database. Appetite 96, 166173.
60. Blechert, J, Meule, A, Busch, NA et al. (2014) Food-pics: an image database for experimental research on eating and appetite. Front Psychol 5, 617.
61. Nutritional Neuroscience Laboratory (2015). http://nutritionalneuroscience.eu/resources/forc-toolbox.
62. Witten, IB, Steinberg, EE, Lee, SY et al. (2011) Recombinase-driver rat lines: tools, techniques, and optogenetic application to dopamine-mediated reinforcement. Neuron 72, 721733.
63. de Backer, MW, Garner, KM, Luijendijk, MC et al. (2011) Recombinant adeno-associated viral vectors. Methods Mol Biol 789, 357376.
64. Boender, AJ, de Jong, JW, Boekhoudt, L et al. (2014) Combined use of the canine adenovirus-2 and DREADD-technology to activate specific neural pathways in vivo . PLoS ONE 9, e95392.
65. Brunstrom, JM (2007) Associative learning and the control of human dietary behavior. Appetite 49, 268271.
66. Hardman, CA, Ferriday, D, Kyle, L et al. (2015) So many brands and varieties to choose from: does this compromise the control of food intake in humans? PLoS ONE 10, e0125869.
67. Fay, SH, Ferriday, D, Hinton, EC et al. (2011) What determines real-world meal size? Evidence for pre-meal planning. Appetite 56, 284289.
68. Brunstrom, JM & Shakeshaft, NG (2009) Measuring affective (liking) and non-affective (expected satiety) determinants of portion size and food reward. Appetite 52, 108114.
69. Wijlens, AG, Erkner, A, Alexander, E et al. (2012) Effects of oral and gastric stimulation on appetite and energy intake. Obesity (Silver Spring) 20, 22262232.
70. Spetter, MS, Mars, M, Viergever, MA et al. (2014) Taste matters – effects of bypassing oral stimulation on hormone and appetite responses. Physiol Behav 137, 917.
71. Spetter, MS, de Graaf, C, Mars, M et al. (2014) The sum of its parts–effects of gastric distention, nutrient content and sensory stimulation on brain activation. PLoS ONE 9, e90872.
72. Brunstrom, JM, Burn, JF, Sell, NR et al. (2012) Episodic memory and appetite regulation in humans. PLoS ONE 7, e50707.
73. Cassady, BA, Considine, RV & Mattes, RD (2012) Beverage consumption, appetite, and energy intake: what did you expect? Am J Clin Nutr 95, 587593.
74. Brunstrom, JM & Mitchell, GL (2006) Effects of distraction on the development of satiety. Br J Nutr 96, 761769.
75. Higgs, S & Donohoe, JE (2011) Focusing on food during lunch enhances lunch memory and decreases later snack intake. Appetite 57, 202206.
76. Camps, G, Mars, M, de Graaf, C et al. (2016) Empty calories and phantom fullness: a randomized trial studying the relative effects of energy density and viscosity on gastric emptying determined by MRI and satiety. Am J Clin Nutr 104, 7380.
77. Herbert, BM, Blechert, J, Hautzinger, M et al. (2013) Intuitive eating is associated with interoceptive sensitivity. Effects on body mass index. Appetite 70, 2230.
78. Griffioen-Roose, S, Smeets, PA, Weijzen, PL et al. (2013) Effect of replacing sugar with non-caloric sweeteners in beverages on the reward value after repeated exposure. PLoS ONE 8, e81924.
79. Smeets, PA, Weijzen, P, de Graaf, C et al. (2011) Consumption of caloric and non-caloric versions of a soft drink differentially affects brain activation during tasting. Neuroimage 54, 13671374.
80. van Rijn, I, de Graaf, C & Smeets, PA (2015) Tasting calories differentially affects brain activation during hunger and satiety. Behav Brain Res 279, 139147.
81. Frank, S, Veit, R, Sauer, H et al. (2016) Dopamine depletion reduces food-related reward activity independent of BMI. Neuropsychopharmacology 41, 15511559.
82. Lerner, JS, Small, DA & Loewenstein, G (2004) Heart strings and purse strings: carry-over effects of emotions on economic transactions. Psychol Sci 15, 337341.
83. Cryder, CE, Lerner, JS, Gross, JJ et al. (2008) Misery is not miserly: sad and self-focused individuals spend more. Psychol Sci 19, 525530.
84. Garg, N, Wansink, B & Inman, JJ (2007) The influence of incidental affect on consumers’ food intake. J Marketing 71, 194206.
85. Pogoda, L, Holzer, M, Mormann, F et al. (2016) Multivariate representation of food preferences in the human brain. Brain Cogn 110, 4352.
86. Val-Laillet, D, Aarts, E, Weber, B et al. (2015) Neuroimaging and neuromodulation approaches to study eating behavior and prevent and treat eating disorders and obesity. NeuroImage 8, 131.
87. van der Laan, LN, de Ridder, DT, Viergever, MA et al. (2014) Activation in inhibitory brain regions during food choice correlates with temptation strength and self-regulatory success in weight-concerned women. Front Neurosci 8, 308.
88. van der Laan, LN, Barendse, ME, Viergever, MA et al. (2016) Subtypes of trait impulsivity differentially correlate with neural responses to food choices. Behav Brain Res 296, 442450.
89. Hare, TA, Camerer, CF & Rangel, A (2009) Self-control in decision-making involves modulation of the vmPFC valuation system. Science 324, 646648.
90. Hare, TA, Malmaud, J & Rangel, A (2011) Focusing attention on the health aspects of foods changes value signals in vmPFC and improves dietary choice. J Neurosci 31, 1107711087.
91. Harris, A, Hare, T & Rangel, A (2013) Temporally dissociable mechanisms of self-control: early attentional filtering versus late value modulation. J Neurosci 33, 1891718931.
92. Lim, SL, Cherry, JBC, Davis, AM et al. (2016) The child brain computes and utilizes internalized maternal choices. Nature Comm 7, 1700.
93. Leng, G & MacGregor, DJ (2008) Mathematical modelling in neuroendocrinology. J Neuroendocrinol 20, 713718.
94. Leng, G, Onaka, T, Caquineau, C et al. (2008) Oxytocin and appetite. Prog Brain Res 170, 137151.
95. Blevins, JE & Ho, JM (2013) Role of oxytocin signaling in the regulation of body weight. Rev Endocr Metab Disord 14, 311329.
96. Olszewski, PK, Klockars, A & Levine, AS. (2016) Oxytocin: a conditional anorexigen whose effects on appetite depend on the physiological, behavioural and social contexts. J Neuroendocrinol 28. Available at https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4879516/
97. Olszewski, PK, Klockars, A, Olszewska, AM et al. (2010) Molecular, immunohistochemical, and pharmacological evidence of oxytocin's role as inhibitor of carbohydrate but not fat intake. Endocrinology 151, 47364744.
98. Maícas Royo, J, Brown, CH, Leng, G et al. (2016) Oxytocin neurones: intrinsic mechanisms governing the regularity of spiking activity. J Neuroendocrinol 28. Available at http://onlinelibrary.wiley.com/doi/10.1111/jne.12376/abstract
99. Deco, G, Jirsa, VK, Robinson, PA et al. (2008) The dynamic brain: from spiking neurons to neural masses and cortical fields. PLoS Comput Biol 4, e1000092.
100. Mavritsaki, E, Allen, HA & Humphreys, GW (2010) Decomposing the neural mechanisms of visual search through model-based analysis of fMRI: top-down excitation, active ignoring and the use of saliency by the right TPJ. Neuroimage 52, 934946.
101. Krajbich, I, Hare, T, Bartling, B et al. (2015) A common mechanism underlying food choice and social decisions. PLoS Comput Biol. 11, e1004371.