Sensory and metabolic signals involved in eating behaviour

Fig. 1 illustrates how eating behaviour is guided by sensory and metabolic signals. In addition, sensory signals of foods are linked in the brain to the metabolic consequences of eating. These (mostly unconscious) learning processes that take place within the context of our living/eating environment shape our food pattern. They guide us throughout the day as to what to eat and also what to expect about the consequences of eating and drinking⁽Reference Woods⁴⁾. For example, people know what to consume for lunch in order to stay satiated until dinner. People know what to expect from a cup of coffee in the morning or a glass of wine in the evening.

Fig. 1. Factors effecting eating behaviour. Sensory processes determine what we eat, and are also responsible for variety in the diet. Metabolic processes determine how much we eat. In the brain, sensory signals during eating are linked to the metabolic consequences. These (largely unconscious) learning processes shape our nutrition pattern. The soft-background of foods represents our current-day food environment (Source: de Graaf and Kok⁽Reference de Graaf and Kok⁷⁵⁾).

Sensory signals play an important role in food choice; in general, we choose the foods we like, and we avoid the foods we dislike⁽Reference de Graaf, Kramer and Meiselman⁵⁾. Sensory signals are also involved in the drive for variety in the diet. Repeated exposure to a particular sensory signal (e.g. a meat flavour; a fruit flavour) during a meal leads to a decline in the reward value of that food, and a shift in interest towards other foods. This phenomenon is called sensory-specific satiation and/or satiety, and has been demonstrated in a large number of studies (e.g.Reference Rolls, Rowe and Rolls^6–Reference Griffioen-Roose, Finalyson and Mars¹²).

Metabolic signals are involved in the regulation of energy intake. Meal initiation is guided by learned associations (time on a clock; a dressed table), which are modulated, for example, by an empty stomach, high ghrelin levels and/or short transient declines in glucose levels in the blood⁽Reference Strubbe and Woods^13–Reference Drazen, Vahl and D'Alessio¹⁵⁾. During eating, hormonal and neural signals from the gastro-intestinal tract to the brain increase our feelings of satiation. Stretch receptors in the stomach and various hormones such as cholecystokinin, glucagon-like peptide-1 and insulin are involved in this process⁽Reference Strubbe and Woods^13–Reference Drazen, Vahl and D'Alessio¹⁵⁾. After the meal we stay satiated for some time depending on the amount and the composition of the foods that we ingest. After a couple of hours the cycle repeats itself. Apart from the short-term regulation of food intake, there is also a longer term regulation of energy balance based on signals coming from the adipose and lean tissue mass⁽Reference Strubbe and Woods^13–Reference Drazen, Vahl and D'Alessio¹⁵⁾. This is, however, outside the scope of this paper.

Various learning processes in food choice and food intake have been well documented. One example of such a learning process is flavour-nutrient learning, which leads to a preference for foods that are high in carbohydrates⁽Reference Birch, McPhee and Steinberg¹⁶⁾, fat⁽Reference Johnson, McPhee and Birch^17, Reference Kern, McPhee and Fisher¹⁸⁾ and/or energy density⁽Reference Appleton, Gentry and Sheperd¹⁹⁾. The absence of energy taste conditioning may also explain the low preference for the pure taste vegetables in children and adults⁽Reference Zeinstra, Koelen and Kok²⁰⁾.

Another example of learning in relation to hunger and satiety is the notion that our meal patterns are highly adjustable to external constraints/environmental circumstances. We usually get hungry just before the regular dinner time, and many people get an appetite for something sweet just before a coffee break. Environmental circumstances provide external cues (e.g. the time on a clock, the emptying of a soup bowl and the smell of food) related to eating behaviour⁽Reference Woods⁴⁾. These patterns and the accompanying anticipatory physiological signals (e.g. ghrelin levels; see e.g. Drazen et al.⁽Reference Drazen, Vahl and D'Alessio¹⁵⁾), can be easily changed with varying environmental circumstances. This flexibility and adaptability are reflected in the wide diversity of eating cultures in the world⁽Reference Woods^4, Reference Menzel and D'Aluisio²¹⁾, and also in the continuous change of eating patterns.

Liquid calories and the regulation of food intake and body weight

One important difference between our current food supply compared to the food supply of our ancestors is that many processed foods in our current diet can be eaten at a high rate. The Neolithic diet compared to our current diet consisted of a relatively high-protein diet (meat and fish) with many foods with a high-fibre content (fruit, vegetables and root plant foods)⁽Reference Cordain, Eaton and Sebastian²²⁾. Meat products and high-fibre products are typically foods that require chewing and are eaten at a relatively low rate. The Neolithic diet did also not contain any liquid calories, except for milk for infants. Infancy is a period of rapid growth of body weight, where body weight at 1 year after birth is usually more than two times the weight at birth⁽Reference Kuczmarski, Ogden and Grummer-Strawn²³⁾.

Liquids are consumed at much faster rates than solid foods⁽Reference Viskaal-Van Dongen, Kok and de Graaf²⁴⁾. This can be seen in Fig. 2, which shows that liquid foods stand out compared to other foods in terms of eating rate. They are consumed at rates higher than 200 g/min, with the exception of soup, which has an eating rate that is comparable to many solid foods (<100 g/min)⁽Reference Viskaal-Van Dongen, Kok and de Graaf²⁴⁾. In this sense, liquid foods can be considered as fast foods, foods that can be eaten quickly. One typical example of the effect of processing food on eating rate is illustrated by a study of Haber et al.⁽Reference Haber²⁵⁾. This showed that it takes about 17 min to consume 500 g apples, whereas 500 g apple juice is consumed in just 1·5 min⁽Reference Haber²⁵⁾. Eating 8368 kJ (2000 kcal), i.e. the average energy requirement for a sedentary woman in the industrialised world, would take more than 2 h for eating apples, whereas for the apple juice the required time would be about 18 min.

Fig. 2. Eating rate (g/min) of forty-five food products across the domain of our food supply. Eating rate was assessed by asking subjects to consume 50 g of each product and recording the time (seconds) necessary to ingest this 50 g (Source: Viskaal-van Dongen M, Kok FJ & de Graaf C (2011) Appetite (In the Press)).

From the appetite/satiety literature it has become clear that liquid foods have a lower satiating capacity than solid foods, with the exception of soup. Some early studies have shown that hunger, measured through subjective scales⁽Reference Blundell, de Graaf and Hulshof²⁶⁾ is more suppressed after consumption of solid fruits compared to an equivalent amount of fruit juices with similar energy and macronutrient content⁽Reference Haber^25, Reference Bolton, Heaton and Burroughs²⁷⁾. The results of Hulshof et al.⁽Reference Hulshof, de Graaf and Weststrate²⁸⁾ reinforced this idea by comparing the effects of hunger on liquid fat preloads with the effects of preloads made solid with gelatin or locust bean gum. The weak effects of liquid calories on energy intake compensation were recently confirmed for liquid and solid proteins, fats and carbohydrates⁽Reference Mourao, Bressan and Campbell^29, Reference Martens, Lemmens and Born³⁰⁾. The exceptional role of soup is illustrated by a study which showed that soups in a variety of forms reduce energy intake at a test meal⁽Reference Flood and Rolls³¹⁾. This issue is worked out in the next section on oro-sensory exposure time and satiety.

The low satiating effect of liquids is not only clear from short preload test meal studies but also from longer term intervention studies. For example, a ten-week study showed that the energy from sugar-sweetened beverages added up to the total energy content of a regular diet⁽Reference Raben, Vasilaras and Moller³²⁾. In a four-week cross-over study, fifteen subjects consumed 1·9 MJ of carbohydrates in either a solid or liquid form. The results showed that subjects ate less throughout the remainder of the day when the solid calories were consumed, but subjects did not lower their intake when the calories were consumed in the form of liquids⁽Reference DiMeglio and Mattes³³⁾. Body-weight changes varied accordingly; subjects gained weight in the liquid calorie condition, but not in the solid calorie condition. Also with respect to long-term weight change soup is different from caloric beverages. In a one-year weight loss/maintenance study, the provision of two soups led to higher weight loss after one year than providing one soup per day or two snacks⁽Reference Rolls, Roe and Beach³⁴⁾.

In a recently published long-term (18 months) randomised controlled behavioural trial among 810 subjects, it was shown that a reduction in liquid calorie intake had a stronger effect on weight loss than did a reduction in calorie intake⁽Reference Chen, Appel and Loria³⁵⁾. Apparently, people detected the solid calories, but not the liquid calories, even after repeated long-term exposure. The notion that liquid calories are not well detected is in line with the results of a number of large-scale prospective epidemiological studies that show a positive association between sugar-sweetened beverage consumption and weight gain and obesity⁽Reference Schulze, Manson and Ludwig^36–Reference Hu and Malik³⁸⁾.

The effect of oro-sensory exposure time on appetite and food intake

One of the fundamental differences between the apple and the apple juice in the studies of Haber et al.⁽Reference Haber²⁵⁾, Mattes⁽Reference Mattes³⁹⁾ and Mattes and Campbell⁽Reference Mattes and Campbell⁴⁰⁾ is the oral sensory-exposure time. With the apple juice there is only a brief moment of sensory stimulation; the vast majority of the sugar and aroma molecules pass the oral/pharyngeal cavity without contact with the senses of smell and taste. The big difference in the duration of aroma (=olfactory) stimulation between various liquid and solid foods was recently confirmed by Ruijschop et al.⁽Reference Ruijschop, Burgering and Jacobs⁴¹⁾, who measured aroma molecules in the exhaled air during food consumption. The duration of aroma release was about ten times longer for mature cheese (solid) than for a strawberry-flavoured dairy liquid⁽Reference Ruijschop, Burgering and Jacobs⁴¹⁾.

As noted earlier, the exception to the rule concerning the low satiating capacity of liquids is soup. Soup has consistently shown to be a high satiating liquid. Mattes compared the hunger response after whole apples, apple juice and apple ‘soup’, which was the apple juice consumed using a spoon⁽Reference Mattes³⁹⁾. Consuming apple juice using a spoon enhances the sensory exposure time to the taste system. The apple ‘soup’ yielded a comparable satiating effect to the apples; the whole apples and the apple soup were much more satiating than the apple juice. One possible explanation for this effect is the observation that oro-sensory exposure with soup is much higher than oro-sensory exposure with a drink. This was confirmed by a study where vegetable soup and tomato soup had eating rate well below 100 g/min⁽Reference Viskaal-Van Dongen, Kok and de Graaf²⁴⁾ (see Fig. 2).

When the effects of sensory signals on food intake are considered, it may be good to realise that sensory signals are mainly operational during eating. This makes it probable that the sensory signals predominantly affect satiation (meal termination) and to a lesser degree satiety (absence of hunger in between meal, involved in meal initiation). This idea is also clear from the satiety cascade from Blundell⁽Reference Blundell, de Graaf and Hulshof²⁶⁾, which distinguishes between sensory, cognitive, post-ingestive and post-absorptive factors in appetite.

The role of eating rate and oro-sensory exposure in satiation was recently studied in a series of experiments in which eating rate and oro-sensory exposure were systematically varied. In a first within-subjects cross-over study in a cinema setting 108 subjects consumed ad libitum a liquid (milk) or semi-solid (custard) chocolate dairy product which were equal in palatability, energy density and macronutrient composition⁽Reference Zijlstra, Mars and de Wijk⁴²⁾. Subjects consumed the products with a broad straw (1·2 cm diameter) from 1·5 litre carton boxes which prevented weight cues and visual cues from playing an important role in intake. The ad libitum consumption of the chocolate-flavoured liquid was about 30% higher compared to a similarly flavoured semi-solid. In a subsequent study under laboratory conditions with about fifty subjects, we investigated whether the eating effort and/or eating rate were responsible for the 30% difference in intake. This was done with the help of peristaltic pumps, which eliminated the effort to get the milk/custard into the mouth, and which could be set at standardised rates. Eliminating the effort had little effect on the difference between ad libitum intakes between the liquid and semi-solid (29%). However, when we matched the eating rate for both products, the difference in ad libitum intakes disappeared to a large extent⁽Reference Zijlstra, Mars and de Wijk⁴²⁾ (see Fig. 3).

Fig. 3. (a) Ad libitum intake (g, with standard deviation) and energy intake (kJ) of the liquid, semi-liquid and semi-solid chocolate dairy products in a cinema setting (n 108; within subjects). Test products were similar in palatability, energy and macronutrient content, and consumed through a straw from 1·5 litre carton boxes over a period of about 90 min. Intake from the semi-solid was 30% less than from the liquid test-food. (b) Ad libitum intake (g, with standard deviation) and energy intake (kJ) of the liquid and semi-solid chocolate dairy product in a laboratory setting (sensory cabin) (n 49; within subjects) within a time frame of about 15 min. In the free eating rate, different effort, subjects consumed the products in the same way as in the cinema setting. In the free eating rate, no effort, subjects consumed the product from a tube making use of a peristaltic pump; in this condition subjects could adjust the rate of delivery. In the fixed eating rate, no effort condition, investigators set the rate of delivery of the test-products into the mouth of the subjects. (Source: Zijlstra et al.⁽Reference Zijlstra, Mars and de Wijk⁴²⁾).

In a subsequent study, it was shown that changing the oro-sensory exposure of chocolate custard from 3 to 9 s/bite led to a 10% lower ad libitum intake⁽Reference Zijlstra, de Wijk and Mars⁴³⁾. The suppressive effect of oro-sensory exposure time on ad libitum intake was confirmed in another study with orangeade, where we held the eating rate constant (150 g/min) but only varied the relative duration of sensory exposure (long: 30 s/min of consumption; short 15 s/min of consumption)⁽Reference Weijzen, Smeets and de Graaf⁴⁴⁾. In a later experimental study, harder and softer versions of solid foods were compared with respect to their effect on ad libitum intake⁽Reference Zijlstra, Mars and Stafleu⁴⁵⁾. Again, foods that were eaten more slowly resulted in lower food and energy intakes⁽Reference Zijlstra, Mars and Stafleu⁴⁵⁾.

In a recent study, subjects (n 105) were repeatedly (ten times) exposed to liquid and semi-solid yoghurt (drinks) for breakfast. Subjects consumed the breakfast either in a liquid form with a straw (n 35), a liquid form with a spoon (n 35) or in a semi-solid form with a spoon (n 35). Eating rate and energy/food intake were higher in the straw condition compared to the spoon conditions⁽Reference Hogenkamp, Mars and Stafleu⁴⁶⁾. This result suggests, as the results of the other studies, that it is not the texture per se, but that the effect of texture on food intake works through oro-sensory exposure time.

In the studies described, the eating rate was varied within one particular food, and the foods used were mostly liquids and semi-solids. Viskaal-van Dongen et al.⁽Reference Viskaal-Van Dongen, Kok and de Graaf²⁴⁾ studied the relationship between eating rate and ad libitum food and energy intake across a range of forty-five different food products, representative of the current food supply. One of the most striking results was the magnitude of the differences in eating rate between various foods. Peanuts were consumed at a rate of about 14 g/min, whereas apple juice and diet coke were consumed at rates of more than 600 g/min. In this study, we observed a positive association between eating rate and ad libitum intake in g (R ²=0·37). With respect to peanuts, it is worth mentioning that peanuts and other nuts have a high energy density, and therefore they are expected to have a low satiating efficiency. However, they have a high satiating capacity⁽Reference Alper and Mattes^47, Reference Mattes, Kris-Etherton and Foster⁴⁸⁾ which is attributable to their low consumption rate.

Altogether, these studies show that foods that can be eaten quickly lead to higher food intakes and lower satiating effects. With fast foods, it may be more difficult to learn to associate the sensory signals during eating with the metabolic consequences after eating (see Fig. 1). This idea was recently partly confirmed in a study in which people were capable of adjusting their intake after repeated exposure to high energy dense semi-solids, but not after repeated exposure to high energy dense liquids⁽Reference Mars, Hogenkamp and Gosses⁴⁹⁾.

To summarise, the effects of texture on appetite and food intake mainly work through oro-sensory exposure time, and thus oro-sensory stimulation. This finding raises the question: how do these oro-sensory signals exert their effect? The responses to sensory cues, such as the taste and smell of food, include a cascade of pre-absorptive physiological responses, which are collectively referred to as CPR⁽Reference Zafra, Moilia and Puerto^50, Reference Smeets, Erkner and de Graaf⁵¹⁾. CPR are a possible way by which we may explain the different satiating capacities of foods that are eaten quickly and foods that are eaten slowly.

The role of cephalic phase responses in the regulation of food intake

CPR are a set of physiological responses, which are conceived to prepare the digestive system for the incoming flow of nutrients after ingestion, with the aim of maintaining homeostasis⁽Reference Zafra, Moilia and Puerto⁵⁰⁾. One of the clearest examples is the salivation response after the sight of or the mere thinking about food. The famous Russian scientist Pavlov was one of the first investigators to study this response, showing that dogs started to salivate when they were exposed to meat⁽⁵²⁾. CPR are also thought to be learned, implying that they are adjustable to the type and size of the meal that is eaten. The responses are transient and often last for only a few minutes; however, their effects on post-prandrial metabolism can be of a longer duration⁽Reference Robertson⁵³⁾.

The homeostatic function of the CPR is clear from a study which showed that the glucose and insulin responses after an intragastric glucose load were lower when subjects chewed on a peanut butter sandwich compared to a no-chewing control condition⁽Reference Teff and Engelman⁵⁴⁾. On average plasma glucose levels returned to baseline earlier after oral sensory stimulation (106 min) than when oral stimulation did not occur. The results of this study indicate that the oral sensory stimulation helped in maintaining glucose homeostasis.

Cephalic stimulation works through the mere thought of food, and through the five senses involved in food perception, i.e. the senses of vision, taste, smell, touch and hearing. Cephalic stimulation ultimately activates the vagus nerve, which results in a myriad of autonomic responses such as the release of various physiologically active compounds, including saliva, gastric acid, pancreatic exocrine enzymes, as well as hormones from the endocrine pancreas⁽Reference Zafra, Moilia and Puerto⁵⁰⁾. Early CPR include saliva and later CPR include gastric acid secretion and the release of pancreatic polypeptide⁽Reference Zafra, Moilia and Puerto⁵⁰⁾. The pancreatic polypeptide response is usually considered as the biological marker for vagal stimulation⁽Reference Robertson⁵³⁾. Most CPR are instrumental for digestion such as hydrochloric acid, gastrin and lipase release in the stomach and digestive enzymes from the pancreas and the duodenum. They have little to do with appetite⁽Reference Smeets, Erkner and de Graaf⁵¹⁾.

The CPR in human subjects is usually assessed through a modified sham-feeding paradigm, where subjects chew on foods and spit the food out at the time where they would normally swallow a food. For a detailed explanation of this methodology, see papers by Teff and co-workers⁽Reference Teff and Engelman^54–Reference Crystal and Teff⁵⁷⁾ and Robertson and co-workers⁽Reference Robertson^53, Reference Heath, Jones and Frayn⁵⁸⁾. Recovery rates of the chewed foods should be close to 100%⁽Reference Robertson⁵³⁾.

The role of cephalic responses in the regulation of food intake has hardly been studied (for a review see⁽Reference Smeets, Erkner and de Graaf⁵¹⁾). There are various possible ways in which CPR may affect appetite. There could be a direct centrally mediated effect on appetite. This direct effect is presumably a learned response based on the association between sensory signals and subsequent metabolic consequences. The notion that people can assess expected satiety from various foods immediately, i.e. within a few seconds after the first bite of food is in line with this idea⁽Reference Brunstrom, Shakeshaft and Scott-Samuel^59, Reference Wilkinson and Brunstrom⁶⁰⁾. The neuro-biological mechanism behind this effect is not yet clear. The findings of studies showing that tasting and ingesting glucose have an impact on the activity of the hypothalamus within 2 min may be related to this mechanism⁽Reference Smeets, de Graaf and Stafleu^61, Reference Smeets, de Graaf and Stafleu⁶²⁾.

CPR may also affect appetite through the release of gastro-intestinal hormones such as insulin, pancreatic polypeptide, leptin and ghrelin⁽Reference Smeets, Erkner and de Graaf⁵¹⁾. In a study where subjects were given a 50 g intragastric fat load, subjective fullness ratings were increased after 1 h modified sham feeding (MSF) of pizza⁽Reference Heath, Jones and Frayn⁵⁸⁾. In this study, a 25% suppression in ghrelin concentration was observed after the fat load, which was enhanced further in the MSF condition. The lower ghrelin levels in the MSF condition were accompanied by lower hunger ratings.

In two studies, Smeets and colleagues showed that the 15–20 min oro-sensory stimulation with fat increases satiety and a relative increase in glucose and insulin levels⁽Reference Smeets, Lejeune and Westerterp-Plantenga⁶³⁾, and may decrease ghrelin levels relative to a water control lunch⁽Reference Smeets and Westerterp-Plantenga⁶⁴⁾. Crystall and Teff⁽Reference Crystal and Teff⁵⁷⁾ showed that a 3 min MSF of high-fat cakes led to higher pancreatic polypeptide levels than MSF of non-fat cakes. The 3 min MSF is much shorter than the 15–20 min of Smeets and colleagues⁽Reference Smeets, Lejeune and Westerterp-Plantenga^63, Reference Smeets and Westerterp-Plantenga⁶⁴⁾ and the 60 min of Heath et al.⁽Reference Heath, Jones and Frayn⁵⁸⁾. Interestingly, the 3 min MSF feeding with high-fat cakes led to a higher intake of high-fat foods.

In one of her first studies on CPR, Teff et al.⁽Reference Teff, Devine and Engelman⁵⁶⁾ showed that sweet-tasting liquids are not adequate stimuli for the elicitation of the cephalic phase insulin response. However, in that same study, MSF apple pie did result in changes in glucose and insulin levels. In a very recent study, Teff⁽Reference Teff⁵⁵⁾ measured the pancreatic polypeptide response (as a measure of stimulation of the vagal nerve) to liquid stimuli, chewing gum and regular mixed nutrient food that required chewing. This study showed that the liquids and chewing gum were not good elicitors of a CPR, but the mixed nutrient foods on which people had to chew were more effective in eliciting a CPR. The results from another group⁽Reference Kokkinos, Le Roux and Alexiadou⁶⁵⁾ showed that consumption of a 2824·2 kJ (675 kcal) mixed meal during 30 min resulted in about 25% higher response of the satiety hormones PYY and glucagon-like peptide 1 compared to the consumption of the same meal within 5 min.

The role of oro-sensory stimulation in appetite regulation compared to the gastric contribution to satiety was also nicely shown in a series of elegant studies⁽Reference Cecil, Francis and Read^66, Reference Cecil, Francis and Read⁶⁷⁾. The main characteristic of these studies was the independent manipulation of oral and gastric stimulation. In one of the first studies, it was shown that oral administration of soup produced the greatest suppression of appetite compared to intragastric delivery and intraduodenal delivery of soup⁽Reference Cecil, Francis and Read⁶⁶⁾. Overt intragastric delivery led to a higher suppression of hunger than covert intragastric delivery⁽Reference Cecil, Francis and Read⁶⁶⁾. Oral administration also delayed gastric emptying. These findings were later confirmed in other studies with oral and gastric administration of fats and carbohydrates⁽Reference Cecil, Francis and Read⁶⁷⁾ and sucrose v. maltose⁽Reference Lavin, French and Read⁶⁸⁾.