Appetite, 1992, 18, 167-183
Eating Style and Eating Behaviour
in Adolescents
JANE WARDLE, LOUISE MARSIAND, YASMIN SHEIKH, MARIE QUINN, INGRID FEDOROFF and JANE OGDEN Institute of Psychiatry University of London
The relationship between eating style, attitudes towards food and food intake was investigated in 846 British adolescent schoolchildren. Eating style was assessed with the Dutch Eating Behaviour Questionnaire, attitudes towards food with a series of specially prepared questions, and food intake with a diet history taken by a dietitian. The results showed that restraint, external and emotional eating were associated with very different profiles of attitudes and behaviour. Restrained subjects had a higher body weight, more negative attitudes towards food, a lower likelihood of overeating and a lower overall energy intake. External
eaters had a lower body weight, positive attitudes to food, and reported a higher energy intake. Emotional eaters fell in between in some ways, with some signs of situational loss of control combined with a negative attitude towards overeating. While external eating appeared to be attenuated by restraint, emotional eating was enhanced by it. The implications of these eating styles for later patterns of eating and weight are discussed.
INTRODUCTION
Psychological research on obesity has focused on the range of internal and external environmental cues which stimulate eating in the absence of a physiological need for food and might, therefore, contribute to a positive energy balance. Emotional eating, i.e. eating which is cued by emotional arousal, has featured in several theoretical accounts of obesity (Kaplan & Kaplan, 1957; Bruch, 1974; Slochower, 1983) and has consistently been found to discriminate obese from normal-weight groups (Slochower, 1983; Lowe & Fisher, 1983). Eating style also attracted attention when the work of Schachter and his colleagues suggested that there were differences between obese and normal-weight subjects in their responsiveness to food cues, a characteristic that was termed externality (cf. Schachter & Rodin, 1974). Experimental studies revealed evidence for over-responsiveness to a range of external food cues combined with under-responsiveness to internal (bodily) cues in obese subjects. Negative affect was also found to influence the food intake of obese and normal-weight subjects differentially. As more data have accrued, the linkage between external or emotional eating and obesity has become less clear-cut (cf. Rodin, 1980), and likewise the idea that the eating behaviour of normal-weight people is entirely regulated by internal need states has been questioned (Wooley et al., 1972). To account for the varying results, A grant from the Health Promotion Research Trust is gratefully acknowledged. Address correspondence to: Dr Jane Wardle, Health Behaviour Unit, Institute of Psychiatry, London SE5 8AF, UK. 0195~6663/92/030167+ 17 $03.00/O
0 1992 Academic Press Limited
168
J. WARDLE ET AL.
Nisbett proposed that the regulatory disturbances that had been observed were not fundamental to obesity, but were the consequences of the steps which many obese people take to control their weight (Nisbett, 1972). This idea was supported by Hibscher & Herman (1977) who demonstrated in a laboratory study that level of dietary restraint predicted disinhibition of food intake better than did body weight. Their work has led to the formulation of Restraint Theory, which proposes that a disposition to weight concern, and attempts to regulate food intake, lead to cognitive and physiological adaptations which can cause failures of regulation and episodes of excessive eating (cf. Herman & Polivy, 1980). Dietary restraint has now largely superseded other aspects of eating style as the critical individual-difference variable in experimental studies of eating behaviour. The eating behaviour of restrained eaters has often been characterized in terms of cognitive control contrasted with disinhibition: when restrained eaters are in control, they undereat, but under certain conditions there is a paradoxical loss of control (disinhibition). In the laboratory, both negative mood states and preloading with high-energy meals have been found to provoke disinhibition (Herman & Mack, 1975; Ruderman, 1985; Wardle & Beales, 1988; Jansen et aE., 1988). The disinhibitors of restraint theory are therefore very similar to the cues for overeating originally identified in the obese. Recently, psychometric tools for the assessment of externally and emotionally cued eating have been developed. The Dutch Eating Behaviour Questionnaire has scales for emotional and external eating as well as restrained eating (Van Strien et al., 1986a). Stunkard & Messick(1985) developed the Three Factor Eating Questionnaire which includes subscales for “Restraint”, “Hunger” and “Disinhibition of Cognitive Restraint”, the latter incorporating items relating to a variety of external and emotional cues which might trigger eating. The relationships between restraint and these various measures of disinhibition have not proved entirely straightforward. Van Strien and her colleagues found positive correlations between their scales of restrained and emotional eating in one study (Van Strien et al., 1986a), but others have failed to find consistent positive associations between restraint and either emotional or external eating (Wardle, 1987; Van Strien et al., 1985; Ganley, 1988). Stunkard & Messick (1985) found a positive correlation between restraint and disinhibition in their full sample, but a negative correlation in a subsample of dieters. These results have left the nature of the inter-relationship between restraint and disinhibition unclear. Energy intake in everyday life has been studied extensively in relation to restraint, with the general finding being that restrained eaters report a lower food intake than non-restrained eaters. Van Strien et al. (1986b) evaluated energy intake from three 24-h recalls of food consumption in a sample of 110 women. They found a significant negative correlation between scores on restraint and “deviation from required caloric intake”, with restrained eaters being more likely to show underconsumption. Their method of assessing “required intake” was based on the subject’s body weight, not on lean body mass. This might have resulted in an overestimation of the energy requirement for the restrained subjects since restraint and weight are known to be correlated. Thus, the negative correlation between restraint and deviation from requirement could have been an artifact of the assessment of energy requirement. However, studies using other methods have reported similar results. Wardle & Beales (1987) in a smaller-scale study using a
EATING
STYLE AND EATING
BEHAVIOUR
IN ADOLESCENTS
169
single 24-h recall, also found a negative correlation between restraint and food intake. Laessle et at, (1989) reported on data from a 7-day dietary record in 60 women, and found a lower overall food intake in the restrainers, and Hill & Robinson (1991) obtained similar results in a small sample of nine-year-old schoolgirls. Women who diet more or less continuously and have low scores on disinhibition, have also been shown to have a lower than average energy expenditure, and by presumption a lower energy intake (Tuschl et al., 1990). Only one study failed to find lower food intake in restrained eaters; Kirkley et al. (1988) found no differences in caloric intake between high and low scorers on the restraint scale although when they excluded “binge days” from the averages the restrainers ate less. This study differed from those described above both in the measure of restraint used (the Herman and Polivy scale), the cultural background of the subjects (North American rather than European) and the type of subject (in this case recruited through the offer of treatment for binge eating). The recruitment method and the measure of restraint would have tended to generate a sample who were high not only on restraint but also on disinhibition of eating. Few studies have investigated the relationship between measures of external or emotional eating and food intake. Westenhoefer et al. (1990) showed data from obese dieters that indicated that disinhibiton is associated with higher energy intake, although there was also an interaction with restraint. In the laboratory, Jansen et al. (1988) found that high scores on either external or emotional eating were associated with failures of regulation after a preload. Two studies have assessed the relationship between externality and weight gain, which might be used as a surrogate for food intake: Rodin & Slochower (1976) found that children who were high on externality gained more weight during a week at summer camp and Weissenberger et al. (1986) found that disinhibition predicted weight gain in depressed people. Similarly, Blair et al. (1990) found that women whose scores on emotional eating increased across two measurement times, were less successful at approaching their target weight. None of these studies assesses the inter-relationship between restraint and disinhibition in relation to food intake. In the present study, eating style, food intake and subjective accounts of control of eating were assessed as part of an investigation of weight, body image and eating patterns in a large sample of adolescents. The use of an adolescent sample permitted the investigation of the development of eating patterns over the teenage years when weight concerns are becoming more widespread. The trends in the relationship between body image, eating and weight over the different age groups give an indication of the processes which may be at work. In previous work, restraint and disinhibition have not been highly related, although restraint and situational loss of control have been consistently related. In the present study the relationship between restraint, disinhibitory eating styles and episodes of loss of control over eating were investigate. Subjects who were higher on disinhibition were predicted to report more episodes of overeating, but only the restrained eaters among them were predicted to report distress after dietary violations. Restrained eaters were predicted to report lower energy intake, while external or emotional (disinhibited) eaters were predicted to report higher energy intake. An interesting issue, and one about which no specific predictions could be made, was how restraint and disinhibition would combine to influence energy intake.
170
J. WARDLE ET AL, METHOD
Subjects
The sample consisted of 846 adolescents recruited from secondary schools whose pupils came from three different social backgrounds. Of the sample, 63% were white-Caucasian (predominantly from U.K. families), 16% were black (predominantly second-generation Afro-Caribbean immigrants) and 20% were Asian and Oriental (again predominantly second-generation immigrants). The data collection was organized through the school7 who had obtained parental approval. The questionnaires were completed during lesson times and thus participation rates were close to 100%. The researchers were only aware of five adolescents who failed to participate (one anorexic and one obese). The characteristics of the sample are shown in Table 1. The relationships between socio-demographic and psychological variables are discussed elsewhere (Warate & MarsIand, 1990). Measures
The subjects were given the measurement instruments to complete in the classroom setting. Each child then had a brief interview with one of the research psychologists to identify any problems with completion, and a longer interview with one of the dietitians to establish details of food intake. Eating style.
All subjects completed the Restrained, External and Emotional Rating scales of the Dutch Eating Behaviour Questionnaire (Van Strien et d, 1986; Wardle, 1987a). Attitudes tofood.
Subjects were presented with four sets of ratings relating to each of 12 food items. The four constructs ‘“like”,“good for you”, “makes you fat” and “feel guilty about eating it” were presented individually at the top of a page, and the 12 foods were presented down the page, each followed by five choices labelled “not at all”, “hardly”, “moderately”, “ quite a lot” and “very much”. They were asked to tick the box that indicated how much they thought the statement at the top applied to each food. Food items were selected from three groups: “slimming” foods (cottage
Schoolyear
Number of girls
Number of boys
Mean age (years)
67
62 69 71
If-8 12-9 13-8 14*8 15.9 16.9 18*0 24-5
;i 63 67 56 43 439
68 65 39 33 407
EATING STYLE AND EATING BEHAVIOUR IN ADOLESCENTS
cheese, Ryvita (low energy crispbread), apples, and (milk, bread, eggs, and cheese) and “fattening” foods chips). These categorizations had been validated in Beales, 1986) but were checked in the present sample you fat”.
171
yoghurt), intermediate foods (chocolate, cake, biscuits and a previous study (Wardle & against the construct “makes
Response to dietary violations.
A set of items was developed to assess the subject’s response to a dietary violation. They were asked “What happens after you have eaten some food that you have been trying to avoid? Do you . . . a) feel upset, b) feel better, c) find it hard to stop eating, d) miss the next meal, e) cut out food for the rest of the day, f) carry on and have a binge, g) make yourself sick. Responses were given as Yes or No. Other data had shown that 45% of the respondents reported that they had at some time tried to avoid one of a list of four foods (sweets, chocolate, chips, biscuits). On the basis of the relative frequency of attempts at avoidance, the subjects were not given the option “not applicable” for these response questions because experience suggested that too many took this “easy” option. Food intake.
Food intake in the previous day was assessed by asking subjects to complete a 24-h food record. This was then used as the basis of a brief (5-10min) interview with a dietitian, with the Cameron & Van Staveren (1988) guidelines being followed as closely as was possible. Previous data on the validity of children’s reports of food intake had led us to believe that the adolescents in our sample would be able to carry out the task. The nutritional evaluation of the food records was done using the “Microdiet” program. Food intake data were analysed for the intake of carbohydrate, fats and protein, and for overall energy intake. These were then assessed in relation to sex, age, restraint, external eating, and emotional eating. Body size.
Height and weight were measured in the classroom and Body Mass Index (BMI) calculated. Perceived size.
Subjective appraisal of body size was made by asking subjects to select, from a list of descriptions from “very fat” to “very thin”, the description that they felt best applied to them.
&XJLTS
Eating Style
Scores on the restrained, emotional and external eating scales of the DEBQ, for each age/sex group are shown in Table 2. Restrained
eating.
Dietary restraint was higher in girls than in boys (Table 2; F[1,838] = 91.4, p
Eating Style and Eating Behaviour
in Adolescents
JANE WARDLE, LOUISE MARSIAND, YASMIN SHEIKH, MARIE QUINN, INGRID FEDOROFF and JANE OGDEN Institute of Psychiatry University of London
The relationship between eating style, attitudes towards food and food intake was investigated in 846 British adolescent schoolchildren. Eating style was assessed with the Dutch Eating Behaviour Questionnaire, attitudes towards food with a series of specially prepared questions, and food intake with a diet history taken by a dietitian. The results showed that restraint, external and emotional eating were associated with very different profiles of attitudes and behaviour. Restrained subjects had a higher body weight, more negative attitudes towards food, a lower likelihood of overeating and a lower overall energy intake. External
eaters had a lower body weight, positive attitudes to food, and reported a higher energy intake. Emotional eaters fell in between in some ways, with some signs of situational loss of control combined with a negative attitude towards overeating. While external eating appeared to be attenuated by restraint, emotional eating was enhanced by it. The implications of these eating styles for later patterns of eating and weight are discussed.
INTRODUCTION
Psychological research on obesity has focused on the range of internal and external environmental cues which stimulate eating in the absence of a physiological need for food and might, therefore, contribute to a positive energy balance. Emotional eating, i.e. eating which is cued by emotional arousal, has featured in several theoretical accounts of obesity (Kaplan & Kaplan, 1957; Bruch, 1974; Slochower, 1983) and has consistently been found to discriminate obese from normal-weight groups (Slochower, 1983; Lowe & Fisher, 1983). Eating style also attracted attention when the work of Schachter and his colleagues suggested that there were differences between obese and normal-weight subjects in their responsiveness to food cues, a characteristic that was termed externality (cf. Schachter & Rodin, 1974). Experimental studies revealed evidence for over-responsiveness to a range of external food cues combined with under-responsiveness to internal (bodily) cues in obese subjects. Negative affect was also found to influence the food intake of obese and normal-weight subjects differentially. As more data have accrued, the linkage between external or emotional eating and obesity has become less clear-cut (cf. Rodin, 1980), and likewise the idea that the eating behaviour of normal-weight people is entirely regulated by internal need states has been questioned (Wooley et al., 1972). To account for the varying results, A grant from the Health Promotion Research Trust is gratefully acknowledged. Address correspondence to: Dr Jane Wardle, Health Behaviour Unit, Institute of Psychiatry, London SE5 8AF, UK. 0195~6663/92/030167+ 17 $03.00/O
0 1992 Academic Press Limited
168
J. WARDLE ET AL.
Nisbett proposed that the regulatory disturbances that had been observed were not fundamental to obesity, but were the consequences of the steps which many obese people take to control their weight (Nisbett, 1972). This idea was supported by Hibscher & Herman (1977) who demonstrated in a laboratory study that level of dietary restraint predicted disinhibition of food intake better than did body weight. Their work has led to the formulation of Restraint Theory, which proposes that a disposition to weight concern, and attempts to regulate food intake, lead to cognitive and physiological adaptations which can cause failures of regulation and episodes of excessive eating (cf. Herman & Polivy, 1980). Dietary restraint has now largely superseded other aspects of eating style as the critical individual-difference variable in experimental studies of eating behaviour. The eating behaviour of restrained eaters has often been characterized in terms of cognitive control contrasted with disinhibition: when restrained eaters are in control, they undereat, but under certain conditions there is a paradoxical loss of control (disinhibition). In the laboratory, both negative mood states and preloading with high-energy meals have been found to provoke disinhibition (Herman & Mack, 1975; Ruderman, 1985; Wardle & Beales, 1988; Jansen et aE., 1988). The disinhibitors of restraint theory are therefore very similar to the cues for overeating originally identified in the obese. Recently, psychometric tools for the assessment of externally and emotionally cued eating have been developed. The Dutch Eating Behaviour Questionnaire has scales for emotional and external eating as well as restrained eating (Van Strien et al., 1986a). Stunkard & Messick(1985) developed the Three Factor Eating Questionnaire which includes subscales for “Restraint”, “Hunger” and “Disinhibition of Cognitive Restraint”, the latter incorporating items relating to a variety of external and emotional cues which might trigger eating. The relationships between restraint and these various measures of disinhibition have not proved entirely straightforward. Van Strien and her colleagues found positive correlations between their scales of restrained and emotional eating in one study (Van Strien et al., 1986a), but others have failed to find consistent positive associations between restraint and either emotional or external eating (Wardle, 1987; Van Strien et al., 1985; Ganley, 1988). Stunkard & Messick (1985) found a positive correlation between restraint and disinhibition in their full sample, but a negative correlation in a subsample of dieters. These results have left the nature of the inter-relationship between restraint and disinhibition unclear. Energy intake in everyday life has been studied extensively in relation to restraint, with the general finding being that restrained eaters report a lower food intake than non-restrained eaters. Van Strien et al. (1986b) evaluated energy intake from three 24-h recalls of food consumption in a sample of 110 women. They found a significant negative correlation between scores on restraint and “deviation from required caloric intake”, with restrained eaters being more likely to show underconsumption. Their method of assessing “required intake” was based on the subject’s body weight, not on lean body mass. This might have resulted in an overestimation of the energy requirement for the restrained subjects since restraint and weight are known to be correlated. Thus, the negative correlation between restraint and deviation from requirement could have been an artifact of the assessment of energy requirement. However, studies using other methods have reported similar results. Wardle & Beales (1987) in a smaller-scale study using a
EATING
STYLE AND EATING
BEHAVIOUR
IN ADOLESCENTS
169
single 24-h recall, also found a negative correlation between restraint and food intake. Laessle et at, (1989) reported on data from a 7-day dietary record in 60 women, and found a lower overall food intake in the restrainers, and Hill & Robinson (1991) obtained similar results in a small sample of nine-year-old schoolgirls. Women who diet more or less continuously and have low scores on disinhibition, have also been shown to have a lower than average energy expenditure, and by presumption a lower energy intake (Tuschl et al., 1990). Only one study failed to find lower food intake in restrained eaters; Kirkley et al. (1988) found no differences in caloric intake between high and low scorers on the restraint scale although when they excluded “binge days” from the averages the restrainers ate less. This study differed from those described above both in the measure of restraint used (the Herman and Polivy scale), the cultural background of the subjects (North American rather than European) and the type of subject (in this case recruited through the offer of treatment for binge eating). The recruitment method and the measure of restraint would have tended to generate a sample who were high not only on restraint but also on disinhibition of eating. Few studies have investigated the relationship between measures of external or emotional eating and food intake. Westenhoefer et al. (1990) showed data from obese dieters that indicated that disinhibiton is associated with higher energy intake, although there was also an interaction with restraint. In the laboratory, Jansen et al. (1988) found that high scores on either external or emotional eating were associated with failures of regulation after a preload. Two studies have assessed the relationship between externality and weight gain, which might be used as a surrogate for food intake: Rodin & Slochower (1976) found that children who were high on externality gained more weight during a week at summer camp and Weissenberger et al. (1986) found that disinhibition predicted weight gain in depressed people. Similarly, Blair et al. (1990) found that women whose scores on emotional eating increased across two measurement times, were less successful at approaching their target weight. None of these studies assesses the inter-relationship between restraint and disinhibition in relation to food intake. In the present study, eating style, food intake and subjective accounts of control of eating were assessed as part of an investigation of weight, body image and eating patterns in a large sample of adolescents. The use of an adolescent sample permitted the investigation of the development of eating patterns over the teenage years when weight concerns are becoming more widespread. The trends in the relationship between body image, eating and weight over the different age groups give an indication of the processes which may be at work. In previous work, restraint and disinhibition have not been highly related, although restraint and situational loss of control have been consistently related. In the present study the relationship between restraint, disinhibitory eating styles and episodes of loss of control over eating were investigate. Subjects who were higher on disinhibition were predicted to report more episodes of overeating, but only the restrained eaters among them were predicted to report distress after dietary violations. Restrained eaters were predicted to report lower energy intake, while external or emotional (disinhibited) eaters were predicted to report higher energy intake. An interesting issue, and one about which no specific predictions could be made, was how restraint and disinhibition would combine to influence energy intake.
170
J. WARDLE ET AL, METHOD
Subjects
The sample consisted of 846 adolescents recruited from secondary schools whose pupils came from three different social backgrounds. Of the sample, 63% were white-Caucasian (predominantly from U.K. families), 16% were black (predominantly second-generation Afro-Caribbean immigrants) and 20% were Asian and Oriental (again predominantly second-generation immigrants). The data collection was organized through the school7 who had obtained parental approval. The questionnaires were completed during lesson times and thus participation rates were close to 100%. The researchers were only aware of five adolescents who failed to participate (one anorexic and one obese). The characteristics of the sample are shown in Table 1. The relationships between socio-demographic and psychological variables are discussed elsewhere (Warate & MarsIand, 1990). Measures
The subjects were given the measurement instruments to complete in the classroom setting. Each child then had a brief interview with one of the research psychologists to identify any problems with completion, and a longer interview with one of the dietitians to establish details of food intake. Eating style.
All subjects completed the Restrained, External and Emotional Rating scales of the Dutch Eating Behaviour Questionnaire (Van Strien et d, 1986; Wardle, 1987a). Attitudes tofood.
Subjects were presented with four sets of ratings relating to each of 12 food items. The four constructs ‘“like”,“good for you”, “makes you fat” and “feel guilty about eating it” were presented individually at the top of a page, and the 12 foods were presented down the page, each followed by five choices labelled “not at all”, “hardly”, “moderately”, “ quite a lot” and “very much”. They were asked to tick the box that indicated how much they thought the statement at the top applied to each food. Food items were selected from three groups: “slimming” foods (cottage
Schoolyear
Number of girls
Number of boys
Mean age (years)
67
62 69 71
If-8 12-9 13-8 14*8 15.9 16.9 18*0 24-5
;i 63 67 56 43 439
68 65 39 33 407
EATING STYLE AND EATING BEHAVIOUR IN ADOLESCENTS
cheese, Ryvita (low energy crispbread), apples, and (milk, bread, eggs, and cheese) and “fattening” foods chips). These categorizations had been validated in Beales, 1986) but were checked in the present sample you fat”.
171
yoghurt), intermediate foods (chocolate, cake, biscuits and a previous study (Wardle & against the construct “makes
Response to dietary violations.
A set of items was developed to assess the subject’s response to a dietary violation. They were asked “What happens after you have eaten some food that you have been trying to avoid? Do you . . . a) feel upset, b) feel better, c) find it hard to stop eating, d) miss the next meal, e) cut out food for the rest of the day, f) carry on and have a binge, g) make yourself sick. Responses were given as Yes or No. Other data had shown that 45% of the respondents reported that they had at some time tried to avoid one of a list of four foods (sweets, chocolate, chips, biscuits). On the basis of the relative frequency of attempts at avoidance, the subjects were not given the option “not applicable” for these response questions because experience suggested that too many took this “easy” option. Food intake.
Food intake in the previous day was assessed by asking subjects to complete a 24-h food record. This was then used as the basis of a brief (5-10min) interview with a dietitian, with the Cameron & Van Staveren (1988) guidelines being followed as closely as was possible. Previous data on the validity of children’s reports of food intake had led us to believe that the adolescents in our sample would be able to carry out the task. The nutritional evaluation of the food records was done using the “Microdiet” program. Food intake data were analysed for the intake of carbohydrate, fats and protein, and for overall energy intake. These were then assessed in relation to sex, age, restraint, external eating, and emotional eating. Body size.
Height and weight were measured in the classroom and Body Mass Index (BMI) calculated. Perceived size.
Subjective appraisal of body size was made by asking subjects to select, from a list of descriptions from “very fat” to “very thin”, the description that they felt best applied to them.
&XJLTS
Eating Style
Scores on the restrained, emotional and external eating scales of the DEBQ, for each age/sex group are shown in Table 2. Restrained
eating.
Dietary restraint was higher in girls than in boys (Table 2; F[1,838] = 91.4, p
