Adolescent Food Choice Criteria: Role of Weight and Dieting Status

app p064

14-07-95 08:07:10

Appetite, 1995, 25, 51–76

Adolescent Food Choice Criteria: Role of Weight and Dieting Status

ISOBEL R. CONTENTO Department of Health and Nutrition Education, Columbia University Teachers College

JOHN L. MICHELA Department of Psychology, University of Waterloo

SUNYNA S. WILLIAMS Department of Psychology, State University of New York College at Buffalo

The purpose of this study was to examine the relation of weight status, dieting status and several associated variables to the criteria for everyday food choice used by adolescents. Study participants were 411 students between the ages of 11 and 18, drawn from 15 schools. The adolescents rated 20 foods in terms of nine food attributes (how tasty or healthful specific foods were, whether the foods were eaten by friends, and so forth). Within-person correlation coefficients were then calculated between these ratings and actual food choices as measured by a food frequency scale. The relation of weight and dieting status, as predictors of each of these correlational indices of the importance of potential food choice criteria, was then analysed using hierarchical multiple regression. In similar fashion, the relation was examined between weight and dieting status and: evaluations of food attributes (choice criteria); dietary quality; calorie, sugar and fat intake; body image; and physical activity. For a majority of food choice criteria and other variables, there was an apparent influence of weight as an independent variable. However, when dieting status was analysed simultaneously with weight, similar and stronger effects were now seen for dieting status and the effects of weight disappeared. Although some of the differences as a function of dieting status resembled differences shown previously in relation to dietary restraint, it is noteworthy that the simpler dieting variable yielded these associations. Overall, a “psychology of dieting” seems more relevant than “psychology of being fat versus being thin”. This psychology appears to involve cognitive self-regulation processes. It is thus crucial that intervention programs and research studies take into account both the dieting status and the weight status of participants.  1995 Academic Press Limited

This research was supported by National Institute of Child Health and Human Development grant no. ND 16559 awarded to the first two authors. We thank the students, teachers, and principals in the participating schools for making this study possible. We also thank a dedicated team of graduate research assistants for their help in this study. Correspondence concerning this article should be sent to Isobel R. Contento, Box 137, Teachers College, Columbia University, 525 West 120 Street, New York, NY 10027, U.S.A. 0195–6663/95/040051+26 $12.00/0

 1995 Academic Press Limited

52

J. R. CONTENTO ET AL.

Several lines of research have investigated adolescents’ food intake. Some studies have been descriptive of adolescent eating patterns or behaviors and have examined the nutritional quality of adolescent diets (Bull, 1988; Court, 1988; Farthing, 1991; Kuczmarski, Brewer & Cronin, 1986; Truswell & Darnton-Hill, 1981). Other studies have examined adolescents’ attitudes and beliefs with respect to food and foodrelated practices (George & Krondl, 1983; Story & Resnick, 1986; Worsley et al., 1984). Our research has focused on food choice in children and adolescents, using decision-making or choice models from contemporary social psychological theories that derive from the expectancy-value motivational model of Lewin (Lewin et al., 1994). When applied to food choice, these models suggest that people choose foods on the basis that eating these foods will bring about consequences they desire—for example, the foods will taste good, will be convenient to prepare or will be good for their health. Taste, or palatability, seems to be especially important as an influence on food choice because it is not only an anticipated consequence of eating but elicits sensory-affective responses as well (American Food Preferences, 1983; Beidler, 1982; Grinker, 1990; Lau et al., 1979; Rozin & Fallon, 1981; Tuorila-Ollikainen, Lahteenmaki & Salovaara, 1986). In addition, many models have included social and environmental influences as motivators of behavior in choice situations (Ajzen & Fishbein, 1980; Triandis, 1979; Wallston & Wallston, 1984). We have shown in prior research that such cognitive-motivational and socialenvironmental factors serve as important criteria in food choice in children, adolescents and mothers of young children (Contento, Michela & Goldberg, 1988; Contento et al., 1993; Michela & Contento, 1986). In these studies, individual differences in the importance of various criteria for food choice were also used as the basis for the identification of subgroups of the population having distinct orientations to food choice, ranging from the “hedonistic” to the “parent-supported health conscious” in adolescents. Other researchers have also increasingly used social psychological models as a basis for theorizing about factors influencing food choice (e.g. Axelson & Brinberg, 1989; Krondl, 1990; Shepherd, 1989; Worsley, 1980) and for conducting studies on the issue (e.g. Brinberg & Durand, 1983; Sapp, 1991; Saunders & Rahilly, 1990; Shepherd & Stockley, 1987). Given Western society’s concern about obesity and cultural pressures for thinness (Garner et al., 1980), it is likely that adolescents’ concerns about weight and body image and the consequent interest in dieting also influence everyday food choice. Studies on adolescents in this area to date have provided descriptive data on obesity status, concerns about weight, perceived body size, body image satisfaction, food attitudes and perceptions and dieting status as well as examinations of the complex interrelationships among these factors (Allen et al., 1993; Davies & Furnham, 1986; Drewnowski & Yee, 1987; Wardle et al., 1992; Nylander, 1971; Rosen & Gross, 1987; Wardle & Marsland, 1990). The construct of dietary restraint, which has been studied intensively in those in their late teens and twenties, is receiving increasing attention in studies of adolescents as well, since concern about weight, and subsequent dieting, seems to be occurring at an ever younger age, particularly in young women (e.g. Wardle & Beales, 1986). Dietary restraint is defined as “the tendency to control food intake consciously in order to prevent weight gain or to promote weight loss” (Tuschl, 1990, p. 105), but also carries the notion that this excessive concern over weight leads to overeating under conditions that are antagonistic to self-control (Polivy & Herman, 1985; Tuschl, 1990; Wardle, 1980a). The construct of dietary restraint is constantly evolving as a result of ongoing research (Herman & Polivy,

ADOLESCENT WEIGHT AND DIETING STATUS

53

1975; Ruderman, 1986; Weber, Klesges & Klesges, 1988). Researchers have debated the nature of the construct (e.g. Westenhoeffer, 1991) as well as how it should be measured (Heatherton et al., 1988; Johnson, Lake & Mahan, 1983; Ruderman, 1986; Stunkard & Messick, 1985; Wardle, 1980b). Several studies have attempted to link these variables to the quality of actual food intake in everyday life, not just quantity. Tuschl and colleagues (Laessle et al., 1989; Tuschl et al., 1990a; Tuschl et al., 1990b), for example, compared the food intakes of restrained to those of unrestrained eaters in adult samples in terms of the macronutrient content of their diets in some studies and in terms of frequencies of consumption of specified items in others. Wardle et al. (1992) examined, in adolescents, the interrelationships among restrained, emotional and external eating styles, attitudes toward food and intake of macronutrients as measured by 24-h dietary recalls. These studies did not, however, report on the relationship of these variables to the importance of the criteria used for everyday food choice by adolescents. Thus in the present study, the role of personal and social-environmental variables in everyday food choice was examined in relation to the weight status and dieting status of adolescents. Dieting status was used rather than dietary restraint because, although the two variables are similar in some respects, dieting is the more clearly defined of the two phenomena and is more straightforward to measure. While the dieting variable used in this study may be seen as one component of restraint, straightforward dieting, without other behavioral eating disorders, is now so commonly practiced that it has been suggested that a separate designation be assigned to it: “chronic dieting syndrome” (Grodner, 1991). Admittedly, even “dieting” covers a wide range of food restriction practices, but this range seems likely to be borne in mind when researchers or lay people discuss dieting. In sum, the study here was designed to examine the relationship of weight status and dieting status to what adolescents valued or wanted in the food they ate and the importance of various personal and social-environmental criteria they used to choose foods. The study also examined several other variables relevant to food choice and weight control: the quality and quantity of the food the adolescents actually ate, their physical activity levels and their body image.

M Overview of Study Design Preadolescent and adolescent students answered questions about food intake and some possible influences on food intake in one-to-one interviews at school and by telephone. During several class meetings, various questionnaires concerning food choice were also completed. Data analyses focused on group differences, between higher versus lower weight (body mass index) students, and more versus less frequent dieters. For example, one set of comparisons involved questionnaire ratings of how desirable (how “good” to “bad” on a 7-point scale) it is to eat foods that are “tasty”, “healthful”, “eaten by friends” and so forth. Some of the other scores involved in group comparisons were produced by calculating within-person correlations (Michela, 1990). A within-persons, acrossfoods correlational approach permits us to take into account the fact that, in everyday

54

J. R. CONTENTO ET AL.

eating, individuals normally choose from among a variety of alternative foods. The correspondences between what a person eats and the perceived attributes of the foods chosen can serve as indicators of the importance of those attributes in an individual’s food choices. For example, a high correlation between what a person perceives to taste good, and what he or she eats, indicates that taste is an important criterion in the person’s food choices. We carried out a series of calculations upon questionnaire responses in order to produce within-person correlations usable as scores on which groups could be compared. Ratings by each of the students (n=411) of each of the foods (n=20) were organized into a computer file in which the various ratings of each food by each person constituted a separate “case” for analysis. Every person had rated the same 20 foods on the same rating scales, thus every case encoded the same variables, and the total number of cases was 20 times the number of respondents. This organization of data made it possible to instruct the computer to work through the data for each person separately, producing several correlations for each person, all calculated across the ratings of the 20 foods. In all instances, one of the variables involved in the correlation was a rating of how frequently the specific foods were eaten. In some instances, the other variable was a rating of a food attribute such as tastiness or healthfulness; in these instances, the resulting within-person correlations were considered to be correlational indices of importance of food choice criteria. In other instances, the other variable was a rating of nutritional quality, supplied by nutritionists; here, the within-person correlations provided food consumption quality indices because they described whether the foods most often eaten were ones rated most favorably by nutritionists. Analyses of 24-h recalls provided information about calorie and nutrient intakes; recall measures of physical activity were analysed to provide an annual total index; and interview measures of body image were scored to produce an index of desired change toward thinness. Like the other outcome variables, these were examined in relation to weight and dieting. Table 1 provides a scheme of the relationships among variables. Participants Participants in the study were 411 students between the ages of 11 and 18, with bimodial peaks, by design, in the age distribution at 11–12 and 16–18 years of age. They were drawn from 15 schools in the New York metropolitan area, selected to represent suburban, small town and urban settings, as well as a wide socioeconomic range, from lower class to upper-middle class. Intact classes were used. For example, usually the entire sixth grade or the entire junior or senior required science or health education classes in a given school were selected. Students in these classes who completed most of the instruments were included in data analyses, for an overall response rate of about 80%. Approximately 47% of the participants were white, 21% were African-Americans, 21% were Latino-Americans, 9% were Asian-Americans and 2% were “other”; 38% were male and 62% were female. Weight Status Weight status was expressed in terms of Weight Percentile. To obtain the Weight Percentile variable, body mass index (BMI) was computed for each subject using

ADOLESCENT WEIGHT AND DIETING STATUS

55

T 1 Scheme of relationships among variables Creation of dependent variables for use in multiple regression

Dependent variables in multiple regression: name and scoring

Independent variables in multiple regression

Evaluation of food attributes Students’ evaluation of eight food attributes

Taste Evaluation, Weight percentile Health Evaluation, etc. Dieting status (1 to 7)

Importance of food choice criteria Students’ rating of 20 foods on nine food attributes correlated withinperson with: student frequency of consumption of same 20 foods

Taste CI, Healthful CI, etc. (−1 to +1)

Weight percentile Dieting status

Food consumption quality Vit/Min Quality, Experts’ rating of 20 foods on four Sugar Quality, etc. objective food attributes correlated (−1 to +1) within-person with: student frequency of consumption of same 20 foods

Weight percentile Dieting status

Nutrient intakes Three random repeat 24 hr dietary recalls

Nutrient Intakes kCal, mg, etc.

Weight percentile Dieting status

Physical activity index Frequency×duration of 47 physical activities

PA Index calories/body weight/ year

Weight percentile Dieting status

Body image From five body image drawings “Self now” minus “self desired”

Body Me (1 to 5) Weight percentile Body Wish (−4 to +4) Dieting status

the following formula: BMI=weight (kg)/height2(m). Weight was measured using a portable scale, and height, with a measuring tape. Then each subject’s BMI was compared with a table of BMI percentiles (Cronk & Roche, 1982). The table contains BMI values for the 5th, 10th, 25th, 50th, 75th, 90th, and 95th percentiles, separately by sex, age and race (Black vs. not Black). This procedure was necessary because the meaning of a given BMI value varies considerably in this sample on account of age, sex and ethnic differences, and we wanted to work with a variable that would have a consistent meaning across the sample. Precise Weight Percentile values were obtained by interpolating between the two nearest percentiles on the table. Those subjects with BMI values in the top five or bottom five percentiles were assigned Weight Percentile values of 2·50 and 97·50, respectively. The body mass index values of the sample ranged from 16 to 41, with a mean of 21. Dieting Status The Dieting Status variable was created by combining responses from an oral dieting indicator (ODI) and a written dieting indicator (WDI). An average of these

56

J. R. CONTENTO ET AL.

two administrations was used to provide a more valid and reliable measure. The ODI was obtained from an oral administration of the Restraint Scale (Herman & Polivy, 1975). Subjects were asked “Have you ever tried to lose weight?” Those who answered “No” were given a score of 0 for the ODI. Those who answered “Yes” were then asked additional questions, one of which was “How often are you dieting?” Possible responses were Rarely (scored as 1), Sometimes (2), Usually (3) and Always (4). The responses to this item were then combined with responses to the first question to form the ODI, which was thus scaled from 0 to 4. The WDI was a single item from the Three-Factor Eating Questionnaire (Stunkard & Messick, 1985) that asked “How often are you dieting in a conscious effort to control your weight?” Possible responses again were Rarely, Sometimes, Usually and Always, scored from 1 to 4. When neither item was missing, the Dieting Status variable was the mean of the z-scores of the ODI and the WDI. When one indicator was missing, the Dieting Status variable was the z-score of the other indicator. The standardized item Cronbach’s alpha for this two-item scale was 0·79; r(353)=0·65, p<0·01, onetailed. Evaluations of Food Attributes In a questionnaire administered to study participants during regularly-scheduled class sessions in various subject areas, students were asked to provide an evaluative rating on each of eight food attributes. That is, students were asked to rate each of eight attributes on a seven-point semantic differential scale: “If a food is X[e.g. tasty], then it is . . .” with the line anchors being “good” and “bad”. Other ratings concerned the anticipated health and weight control consequences of eating the food—that is, how desirable it was for the food to be healthful, fattening, high in sugar; and social environmental factors—that is, how desirable it was for the food to be easy to get, served by parents, eaten by friends, or to cost a lot. These variables are labelled: Healthful Evaluation, Fattening Evaluation, and so forth. Measures of Food Choice Criteria On another questionnaire, students were asked to provide ratings of their beliefs about 20 foods, using five-point scales. The first rating concerned the degrees to which each of these 20 foods was regarded as tasty, using a scale with end points labeled “tastes very bad” and “tastes very good”. Other ratings concerned the degree to which each food was healthful, contained sugar, was fattening, was easy to get and so forth. This list of food attributes was the same as for the “evaluations of food attributes” task described above, except for an additional attribute, “contains a lot of vitamins and minerals”. Each questionnaire page contained the list of 20 foods as the left-hand column, the numbers 1 through 5 to the right of each food, and, at the top right of the page, the endpoints of the five-point scale for one of the attributes. The 20 foods listed on the questionnaire were selected so as to include foods liked and disliked, and foods eaten frequently and infrequently by adolescents (USDA, 1978). Foods were also chosen to represent nutritionists’ food groups as well as children’s own food classification schemes (Michela & Contento, 1984). These 20 foods were whole wheat bread, noodles/other pasta, non-sugared cereals (more

ADOLESCENT WEIGHT AND DIETING STATUS

57

nutritious grains and cereals); cakes/snack cakes and sugared cereals (less nutritious grains and cereals); chicken and beans (lower-fat meat/alternative); hamburger and eggs (higher-fat meat/alternative); milk, hard cheese and ice cream (dairy); green salad (more frequently eaten vegetable); apples (more nutritious snacks); candy, potato chips, French fries (less nutritious snacks); orange juice (more nutritious beverage) and soda (less nutritious beverage). This variation in foods provided a reasonable opportunity for detecting the associations between food attributes and food consumption that we indexed with within-person correlations. It should be noted that changes to the list of 20 foods can be expected to yield changes in the sample or population means of within-person correlations. However, there is no reason to expect that the relative standings of dieting- and weight-based groups on these indices would be affected by changes in the food list within the general selection guidelines.

Food Frequency Scale The same format was used to assess participants’ food consumption. Students were asked to rate how often they ate each of the 20 foods on a five-point scale, from “I never eat this food” to “I eat this food very often”. This measure of food intake was validated by comparing it with the Health and Nutrition Examination Survey food frequency measure modified for children (Michela & Contento, 1985, 1986).

Calculation of Correlational Indices of Importance of Food Choice Criteria To examine whether the criteria for everyday food choice vary depending on weight status and dieting status, indicators of the importance of potential criteria needed to be first obtained for each member of the sample. Questionnaire data were analysed by within-person correlations for this purpose. Specifically, each of the nine measures of food choice criteria (reflecting the categories of taste, social environmental factors and anticipated health and weight control consequences) was correlated with the food-frequency scale. For example, each study participant’s ratings of the 20 foods on the scale “is healthy for me” were correlated with his or her ratings of the same foods on the food frequency scale. In this example, a high positive correlation would indicate that healthfulness of foods is a substantial criterion for food choice for this person. This analysis provided nine within-person correlations for each member of the sample (i.e. nine “scores” usable as dependent variables in analyses involving weight and dieting status as independent variables). The correlations could range from −1 to +1. These indices of the importance of nine potential food choice criteria are referred to as the correlational indices (CIs) of food choice criteria. The nine variables are abbreviated as Taste CI, Healthful CI, Fattening CI and so forth. Our previous studies (Michela & Contento, 1986; Contento et al., 1988) have examined the validity of within-person correlations as indicators of the importance of such criteria in food choice. The evidence for validity lies in various convergences found between this and other methods for identifying individuals with relatively greater motivation to choose foods that are healthful, tasty, convenient and so forth. Nevertheless, the limitations inherent in correlations relationships should be kept in mind when interpreting the findings.

58

J. R. CONTENTO ET AL.

Expert Ratings To provide a standard against which to judge adolescents’ responses on the food choice criteria questionnaire, the same questionnaire (concerning beliefs about 20 foods) was given to ten nutritionists and graduate students in nutrition. Their responses will be referred to as the “expert ratings”. Responses were solicited on only the four scales for which objective assessments could be made, namely, the degree to which the 20 foods were healthful, contained vitamins and minerals, contained sugar and were fattening. The expert ratings were used to construct the measure of food consumption quality described immediately below. Food-Consumption Quality Scores Food-consumption quality scores were calculated by correlating experts’ ratings of food attributes with study participants’ self-reported food consumption across the 20 foods on a within-person basis. Thus, a study participant’s score on the healthful variable approached +1·0 to the extent that he or she reported eating the same foods that the panel of experts rated as healthful. This score approached −1·0 if the foods the study participant reported eating were rated as relatively unhealthful by the experts. Ratings near 0·0 reflected little relation between experts’ ratings of the attributes of the 20 foods and the study participant’s reported consumption of those foods. The four variables in this category are abbreviated as Healthful Quality, Vitamins/Mineral Quality, Sugar Quality and Fattening Quality. Twenty-Four-Hour Recall and Nutrient Intake Analysis There is ongoing debate on appropriate methods for obtaining information about people’s dietary intakes (Block, 1982; Mertz, 1992). For the purposes of this study and the accuracy desired, the use of several random repeat 24-h dietary recalls (Balogh, Kahn & Medalie, 1971) was judged to be the most suitable method for obtaining food intake information from these teenagers. Consequently, three recalls were obtained from each participant, one in person and two by telephone. The inperson recall was obtained by means of an individual interview, using food models and standard measures such as cups and spoons to help the participant estimate portion size. The interview was conducted in an available room in the participant’s school. The two telephone recalls were obtained 2–4 weeks later, on days that would, overall, provide data on two weekdays and one weekend day for each person. These interviews yielded information about food items consumed, methods of preparation, and portion sizes, which were entered into a microcomputer program for nutrient analysis based on the USDA data base (Nutri-Calc, 1983). Each 24-h recall was analysed separately, and results were averaged separately for each nutrient across the 3 days for each person. Averaging the data provides more valid and reliable estimates of intake. If a member of the sample could not be reached for one of the telephone interviews, averages were calculated for nutrients across the days for which data were available. Data were available for all 3 days for 295, 2 days for 65, and 1 day for 51 participants. Physical Activity The physical activity of the adolescents was assessed using the Minnesota Leisure Time Activity Questionnaire (Taylor, Jacobs & Schuker, 1978) modified after pilot-

ADOLESCENT WEIGHT AND DIETING STATUS

59

testing to be more appropriate to this adolescent population (e.g. horseback riding was omitted, and disco-dancing was added). The instrument was given orally in an individual interview. For each of 47 activities (e.g. basketball, disco-dancing), the participants were asked to estimate the following frequencies: (a) the weeks per year engaged in the activity; (b) the times per week engaged in the activity during those weeks; and (c) the minutes per occasion engaged in the activity. The participants were asked to list other activities not included in the questionnaire in which they were engaged in the past year and to make the same frequency estimates as above for these. The Physical Activity Index represents the vigorousness of the overall level of leisure time activity of individuals and was measured by the total number of calories expended per pound body weight per year. Body Image Body image assessments were obtained from adolescent females by using a series of five outline drawings of the female figure described by Storz and Greene (1983). Male figures modelled on those of the female figures were used with male study participants. In an individual interview, the adolescent was presented with the set of the five outline figure drawings of his or her sex and was asked two questions: (a) “Which looks the most like you?” and (b) “Which would you most want to look like?” The answer to (a) constituted the Body Me variable. The difference between the answers to (a) and (b) resulted in the Body Wish variable, with higher scores in the negative direction indicating a wish to be thinner, and higher scores in the positive direction, fatter.

R Multiple Regression Analysis Design The relationships of Weight Percentile and Dieting Status with each of several dependent variables (Evaluations of Food Choice Criteria, Correlational Indices of Food Choice Criteria, Food Consumption Quality Scores, Nutrient Intakes, Physical Activity and Body Image) were studied using hierarchical multiple regression (MR) analyses. The MR analyses were carried out in two steps, always controlling for Age and Sex. Sex was dummy coded as 0=male, 1=female. In the first step, the dependent variable was regressed on Weight Percentile (and Age and Sex, simultaneously). Because Weight Percentile and Dieting Status were found to be correlated (r=0·51, p<0·01), in the second step Dieting Status was added to the MR equation, to examine whether apparent effects of weight might be attributable to dieting instead. A third step was also run for each analysis in which we examined the two-way interaction of weight and dieting status, but these analyses did not yield a greater number of significant results than would be expected by chance. An alternative third step was then run for each analysis, in which we addressed whether age and sex moderated the effects of Weight Percentile and Dieting Status. Corresponding multiplicative interaction terms were formed. These terms were entered in various combinations (along with main effects of the earlier steps) to ensure that significant effects would not be masked by correlations among interaction vectors. Again, we did not find any more significant effects than would

60

J. R. CONTENTO ET AL.

be expected by chance, and none was compelling. Consequently, we will report findings only for main effects. Because the correlational indices of the importance of various criteria in food choice and the food consumption quality scores are within-person correlations, they were transformed from r to Fisher’s z′ before being used in the multiple regression analyses. Such a transformation is desirable when correlations are used as scores in statistical analyses for which normality is assumed (Michela, 1990). For ease of comprehension, however, the group means that are presented for these variables have been transformed back from z′ to r. Dependent variable group means (which we will report for different levels of Weight Percentile and Dieting Status) are based on a procedure described by Kenny (1985) for use with continuous predictor variables employed in multiple regression analysis. Each mean is the expected dependent variable value at plus-or-minus one standard deviation from the mean of the continuous variable, where the expectation is given by unstandardized coefficients of the regression equation (multiplied by the appropriate mean, plus-or-minus one standard deviation). This way of presenting means ensures that factors such as Age and Sex were not confounding the group differences under examination. That is, no “groups” of participants are actually created in the data analysis. Instead, two points along the regression line are presented for ease of interpretation of the results. In the presentation of results, low weight individuals are those one standard deviation below the Weight Percentile mean and high weight individuals are those one standard deviation above the mean, which correspond to values of about 22 and 82 (on a weight-for-height percentile scale from 0 to 99), respectively. These values in our sample correspond reasonably closely with the corresponding values of 16 and 84 expected with a completely normal distribution. According to oral (ODI) and written (WDI) self-report ratings, 53·3% of those in this sample (64% of the males, 47% of the females) “never” or “rarely” diet, 25·3% of the sample (21% males, 28% females) “sometimes” diet, and 21·4% (15% males, 25% females) “often” or “always” diet. The differences between males and females are statistically significant: females diet more frequently, v2(2, n=409)=12·03, p<0·01. Low dieters are those one standard deviation below the Dieting Status mean and high dieters are those one standard deviation above the mean, corresponding approximately to self-ratings of “never” diet and “often” diet, respectively. Evaluations of Food Attributes or Choice Criteria MR analysis was used to examine the effects of Weight Percentile and Dieting Status on the adolescents’ evaluative ratings of each of eight food choice criteria on a seven-point good–bad semantic differential scale (1=bad and 7=good). Table 2 presents the scale means and Table 3 presents the beta weights and multiple regression coefficients for the two hierarchical regression steps for the Evaluation Variables. No differences were found for Taste Evaluation in the first step of the regression analysis with both high weight and low weight individuals valuing taste highly as a food choice criterion. A significant effect was found, however, for Dieting Status on the second regression step, such that low dieters valued tastiness of foods more than did high dieters though both ratings are highly favorable. Among social-environmental factors, no significant effects were found for either Weight Percentile or Dieting Status on Friends Evaluation and Cost Evaluation,

61

ADOLESCENT WEIGHT AND DIETING STATUS

T 2 Means of evaluations of food choice attributes by weight percentile and dieting status on the first and second hierarchical steps of multiple regression analysis Evaluation of food attribute∗

Multiple regression step 1 Sample mean

Taste Evaluation Parents Evaluation Friends Evaluation Easy Evaluation Cost Evaluation Healthy Evaluation Fattening Evaluation Sugar Evaluation

Multiple regression step 2

Low† weight

High† weight

Low† weight

High† weight

Low‡ dieter

High‡ dieter

5·98

6·01

5·96

5·93

6·04

6·14§

5·83§

5·49

5·58

5·40

5·58

5·40

5·52

5·46

4·32

4·38

4·26

4·29

4·35

4·45

4·18

5·34

5·31

5·37

5·16∀

5·52∀

5·62§

5·05§

3·62

3·77

3·47

3·68

3·56

3·77

3·46

6·56

6·50

6·62

6·50

6·62

6·59

6·53

2·81

3·14∀

2·48∀

3·02∀

2·60∀

3·01§

2·60§

3·91

3·97

3·85

3·88

3·94

4·08

3·74

ns for dependent variables ranged from 377 to 379. ∗ In the evaluation of food attributes, 1=“bad” and 7=“good”. † “Low weight” and “high weight” individuals are those at one standard deviation below and above the mean for weight-for-height, respectively (see Results section). ‡ “Low dieter” and “high dieter” are individuals at one standard deviation below and above the dieting status mean, respectively (see Results section). § Values for high dieters and low dieters differ significantly (see Table 3). ∀ Values for high weight and low weight individuals differ significantly (see Table 3).

both factors being given a more-or-less neutral evaluation. For Easy Evaluation, although no significant effect was found for Weight Percentile on the first regression step, significant effects were found for both Weight Percentile and Dieting Status on the second regression step, in opposite directions, such that high-weight individuals valued easy access as a food choice criterion more than did low-weight individuals, while high dieters valued easy access less than did low dieters. For Parents Evaluation, there were no significant effects for either Weight Percentile or Dieting Status in either of the regression steps. The adolescents generally valued their parents’ serving them food. Among anticipated health consequences and weight control factors, no significant effects were found for either Weight Percentile or Dieting Status on Healthful Evaluation and Sugar Evaluation with all adolescents highly valuing healthfulness of a food as a food choice criterion, and somewhat disapproving of sugary foods. For Fattening Evaluation, a significant effect was found for Weight Percentile on the first regression step. In addition, on the second step, significant effects were found for both Weight Percentile and Dieting Status, such that high-weight individuals or high dieters disapproved more of fatteningness as a food choice criterion than did low-weight individuals or low dieters.

62

J. R. CONTENTO ET AL.

T 3 Beta weights and multiple correlation coefficients for the hierarchical regressions of evaluations of food choice criteria on age, sex, weight and dieting∗† Variable

Step

Age beta

Sex beta

Weight beta

Dieting Multiple beta R

Taste Evaluation

1 2

−0·17∗∗ −0·03 −0·17∗∗ −0·00

−0·02 0·05

−0·12∗

0·17∗ 0·20∗∗

Parents Evaluation

1 2

−0·08 −0·08

0·02 0·03

−0·07 −0·06

−0·02

0·11 0·12

Friends Evaluation

1 2

−0·07 −0·07

−0·01 0·01

−0·05 0·02

−0·12

0·09 0·13

Easy Evaluation

1 2

−0·05 −0·05

0·08 0·13∗

−0·20∗∗

0·10 0·19∗∗

Cost Evaluation

1 2

−0·03 −0·03

0·02 0·05

−0·10 −0·04

−0·10

0·10 0·13

Healthful Evaluation

1 2

−0·02 −0·02

0·13∗ 0·13∗∗

0·05 0·07

−0·04

0·13 0·14

Fattening Evaluation

1 2

0·03 0·03

Sugar Evaluation

1 2

0·10 0·10

0·03 0·13∗

−0·15∗∗ −0·19∗∗ −0·12∗ −0·13∗ −0·12∗ 0·03 0·06

−0·03 0·02

−0·11

0·23∗∗ 0·25∗∗ 0·11 0·14

† Dieting was not entered in the first hierarchical step. ∗ p<0·05, two-tailed. ∗∗ p<0·01, two tailed.

Table 3 also reveals a few significant main effects of sex and age for several of the evaluation variables. In a procedure similar to the one used for Table 2, means on these outcome variables were calculated from the results of steps 1 and 2 of the regression analysis. Taste Evaluation was found to have mean ratings of 6·19 for younger and 5·78 for older individuals in both steps 1 and 2. Healthful Evaluation was found to have a mean rating of 6·40 for males and 6·65 for females in both steps 1 and 2. The corresponding means for Fattening Evaluation were 3·14 for males and 2·62 for females in step 1 and 3·07 and 2·65, respectively, in step 2. As noted in the description of the design of the regression analysis, there were, however, no interactions between sex, age, Weight Percentile and Dieting Status for these variables. Correlational Indices of Importance of Food Choice Criteria Also examined were the relationships of Weight Percentile and Dieting Status to indices of the influences of various food choice criteria. Table 4 presents the mean scores and Table 5 presents the beta weights and multiple regression coefficients for the two regression steps for the correlational indices (CIs). While there was a main

63

ADOLESCENT WEIGHT AND DIETING STATUS

T 4 Mean correlational indices (CI) of importance of food choice criteria by weight percentile and dieting status on the first and second hierarchical steps of multiple regression analysis Food choice criteria (CIs)∗

Multiple regression step 1 Sample mean

Taste CI 0·63 Parents Serve CI 0·63 Friends Eat CI 0·36 Easy to Get CI 0·51 Costs a Lot CI 0·07 Healthful CI 0·19 Has Vitamins/Minerals CI 0·10 Fattening CI 0·00 Has Sugar CI −0·02

Multiple regression step 2

Low weight

High weight

Low High weight weight

Low dieter

High dieter

0·65 0·63 0·41† 0·53 0·07 0·13† 0·04† 0·03 0·04†

0·61 0·63 0·30† 0·49 0·07 0·25† 0·16† −0·03 −0·08†

0·63 0·63 0·63 0·63 0·36 0·36 0·51 0·51 0·07 0·07 0·19 0·19 0·10 0·10 −0·03 0·03 −0·02 −0·02

0·65 0·62 0·46‡ 0·52 0·10 0·09‡ 0·01‡ 0·13‡ 0·08‡

0·61 0·63 0·25‡ 0·50 0·04 0·29‡ 0·18‡ −0·12‡ −0·13‡

ns for dependent variables ranged from 399 to 405. ∗ Correlational Indices may range from −1 to +1. † CIs for high weight and low weight individuals differ significantly (see Table 5). ‡ CIs for high dieters and low dieters differ significantly (see Table 5).

effect of age for Taste CI such that the CI for older respondents was higher than that for younger (0·68 vs. 0·58), again, there were no interactions among age, sex, Weight Percentile and Dieting Status. As we can see from Tables 4 and 5, there was no significant effect of either Weight Percentile or Dieting Status on Taste CI. Given these results and sample mean for Taste CI score of 0·63, it seems clear that, to a substantial degree, the adolescents typically ate what tasted good to them, regardless of weight or dieting status. Among social-environmental factors, no significant effects were found for either Weight Percentile or Dieting Status on Parents CI, Easy CI, and Cost CI, with the first two CIs being quite high and the last being minimal. If “parents serve” is one measure of availability for these adolescents, and “easy to get” a measure of convenience, then these results suggest that availability and convenience are major determinants of the food choices while cost is not very important for adolescents, regardless of weight or dieting status. For Friends CI, a significant effect was found for Weight Percentile on the first regression step. However, on the second regression step, Dieting Status, but not Weight Percentile, had a significant effect, such that low dieters were more likely to eat foods they believed their friends ate than were high dieters. This result illustrates the importance of analysing Weight Percentile and Dieting Status simultaneously, because when Dieting Status was not in the regression analysis, the beta for Weight Percentile was significant, but when Dieting Status was included in the analysis, the effect for Weight Percentile disappeared. Among anticipated health consequences and weight control factors, for Fattening CI, a significant effect was found for Dieting Status on the second regression step,

64

J. R. CONTENTO ET AL.

T 5 Beta weights and multiple correlation coefficients for the hierarchical regressions of correlational indices of importance of food choice criteria on age, sex, weight and dieting† Age beta

Sex beta

Weight beta

Dieting beta

Multiple R

1 2

0·19∗∗ 0·19∗∗

0·02 0·04

−0·07 −0·02

−0·08

0·20∗∗ 0·22∗∗

1 2

0·08 0·08

0·06 0·05

0·02 0·01

0·02

Friends CI

1 2

−0·04 −0·04

−0·06 0·00

−0·13∗∗ 0·01

−0·27∗∗

0·14∗ 0·26∗∗

Easy CI

1 2

0·02 0·02

−0·12∗ −0·11

−0·05 −0·03

−0·04

0·13 0·13

Cost CI

1 2

0·09 0·10

−0·05 −0·03

−0·02 0·02

−0·09 −0·09

0·13 0·13

Healthful CI

1 2

0·04 0·04

−0·02 −0·07

0·13∗∗ 0·01

0·23∗∗

0·14 0·24∗∗

Vitamins/ Minerals CI

1 2

0·06 0·06

−0·01 −0·06

0·14∗∗ 0·03

0·21∗∗

0·15 0·23∗∗

Fattening CI

1 2

−0·02 −0·01

−0·10 −0·02

−0·09 0·09

−0·33∗∗

0·12 0·30∗∗

Sugar CI

1 2

−0·00 −0·00

−0·03 0·03

−0·16∗∗ −0·02

−0·26∗∗

0·16∗ 0·27∗∗

Variable

Step

Taste CI Parents CI

0·10 0·10

† Dieting was not entered in the first hierarchical step. ∗ p<0·05, two-tailed. ∗∗ p<0·01, two-tailed.

such that low dieters tended to eat foods that they believed were fattening, while high dieters tended to avoid such foods. For Healthful CI, Vitamins/Minerals CI and Sugar CI, significant effects were found for Weight Percentile on the first regression step, such that high weight individuals, compared to low weight individuals, ate foods they judged to be more healthful, higher in vitamins and minerals, and lower in sugar. On the second step, however, Dieting Status but not Weight Percentile had significant effects, such that high dieters (a) were more likely to eat foods they believed were healthful than were low dieters, (b) were more likely to eat foods they believed contained vitamins and minerals than were low dieters and (c) tended to avoid food high in sugar, while low dieters had a slight tendency to eat such foods. Again, these results emphasize the importance of analysing weight and dieting status simultaneously.

65

ADOLESCENT WEIGHT AND DIETING STATUS

T 6 Means of food consumption quality, nutrient intakes, physical activity, and body image satisfaction by weight percentile and dieting status on the first and second hierarchical steps of multiple regression analysis Relevant variable∗

Multiple regression step 1 Sample mean

Low weight

0·00 −0·00

−0·06† −0·06†

0·02 0·02

0·05† 0·08†

High weight

Multiple regression step 2 Low weight

High weight

Low dieter

0·06† 0·06†

0·03 −0·00

−0·03 −0·00

−0·12‡ −0·12‡

−0·01† −0·04†

−0·01 0·02

0·05 0·02

High dieter

Food consumption quality Healthful Quality Vitamins/Minerals Quality Fattening Quality Sugar Quality

0·13‡ 0·11‡

0·14‡ −0·10‡ 0·10‡ −0·07‡

Nutrient intakes Calorie intake (kcal) (MJ) Total fat (g) Sugar (g) Vitamin A (IU) Vitamin C (mg) Iron (mg) Calcium (mg)

2081 (8·7) 78 72 5093 141 14 883

2236† (9·3) 83† 75 5455 147 15 925

1926† (8·1) 74† 68 4731 125 12 842

2117 (8·8) 78 66 5470 144 14 903

2045 (8·5) 78 77 4715 138 13 864

2300‡ (9·6) 87‡ 89‡ 5064 147 14 923

1861‡ (7·8) 70‡ 54‡ 5122 135 13 843

1395

1369

1421

1306

1484

1513

1277

Physical activity Physical activity index Body image Body Me Body Wish

2·80 −0·37

2·11† 0·35†

3·48† −1·08†

2·17† 3·42† 2·73 2·87 0·20† −0·93† −0·12‡ −0·62‡

ns for dependent variables ranged from 403 to 411. ∗ See text and Table 1 for description of scoring. † Values for high weight and low weight individuals differ at p<0·01 for variable (see Table 7). ‡ Values for high dieter and low dieter differ at p<0·01 for variable (see Table 7).

Food Consumption Quality Scores MR analysis was used also to examine the relationships of Weight Percentile and Dieting Status with the Food Consumption Quality Scores—i.e. scores representing extent of consumption of foods judged by experts to have particular health-related qualities. The findings are shown in Tables 6 and 7. For all four of the Food Consumption Quality Scores (Healthful Quality, Vitamins/Minerals Quality, Fattening Quality and Sugar Quality), significant effects were found for Weight Percentile on the first regression step: those of high weight, compared to those of low weight, were more likely to eat foods that were healthful (score=0·06 vs.−0·05) and higher in vitamin/mineral content (0·06 vs.−0·06), and were less likely to eat foods that were fattening (−0·01 vs. 0·05) and high in sugar

66

J. R. CONTENTO ET AL.

T 7 Beta weights and multiple correlation coefficients for the two step hierarchical regressions of food consumption quality scores, nutrient intakes, physical activity, and body image on age, sex, weight and dieting† Step

Age beta

Sex beta

1 2

0·06 0·06

0·06 −0·02

0·13∗∗ −0·04

0·31∗∗

0·15∗ 0·30∗∗

Vitamins/ Minerals Quality

1 2

0·05 0·05

−0·03 −0·09

0·14∗∗ −0·01

0·28∗∗

0·15∗ 0·28∗∗

Fattening Quality

1 2

−0·05 −0·05

−0·07 0·00

−0·11∗∗ 0·07

−0·34∗∗

0·14 0·31∗∗

Sugar Quality

1 2

−0·04 −0·04

−0·06 0·00

−0·14∗∗ 0·00

−0·27∗∗

0·16 0·47∗∗

Calorie Intake

1 2

0·08 0·08

−0·38∗∗ −0·31∗∗

−0·20∗∗ −0·05

−0·28∗∗

0·41∗∗ 0·47∗∗

Fat Intake

1 2

0·13∗∗ 0·13∗∗

−0·32∗∗ −0·26∗∗

−0·13∗∗ −0·00

−0·24∗∗

0·35∗∗ 0·40∗∗

Sugar Intake

1 2

−0·13 −0·06

−0·07 0·11

−0·34∗∗

0·16∗ 0·32∗∗

Physical Activity

1 2

0·27∗∗ 0·27∗∗

−0·23∗∗ −0·21∗∗

0·02 0·08

−0·10

0·36∗∗ 0·37∗∗

Body Me

1 2

0·12∗∗ 0·12∗∗

−0·08∗ −0·10∗

0·67∗∗ 0·63∗∗

0·07

0·69∗∗ 0·69∗∗

−0·21∗∗

0·62∗∗ 0·64∗∗

Variable Healthful Quality

Body Wish

1 2

−0·05 −0·05

−0·01 −0·01

−0·30∗∗ −0·26∗∗

Weight beta

−0·58∗∗ −0·47∗∗

Dieting beta

Multiple R

† Dieting was not entered in the first hierarchical step. ∗ p<0·05, two tailed. ∗∗ p<0·01, two tailed.

content (−0·04 vs. 0·08). On the second step, however, Dieting Status had significant effects while Weight Percentile did not, such that high dieters (a) tended to eat foods that were healthful and that contained vitamins and minerals, while low dieters tended to eat foods that were not healthful and foods that did not contain vitamins and minerals and (b) tended to avoid fattening foods and foods that contained sugar, while low dieters tended to eat foods that were fattening and contained sugar. Nutrient Intakes MR analysis was used next to examine the relationships of Weight Percentile and Dieting Status to the intakes of nutrients as obtained from the 24-h dietary recalls. (See Table 6 for mean intakes of selected nutrients and Table 7 for regression

ADOLESCENT WEIGHT AND DIETING STATUS

67

results.) For calories and grams of fat consumed per day, significant effects were found for Weight Percentile on the first regression step such that those in the high weight group consumed fewer calories and less fat than those in the low weight group. On the second step, however, Weight Percentile no longer had significant effects. Dieting Status now had a significant effect, such that high dieters consumed fewer calories than did low dieters and less fat. For grams of sugar consumed per day, a significant effect was found for Dieting Status on the second regression step, such that high dieters consumed less sugar than did low dieters. Diets in this sample were adequate in nutrients, averaging above the Recommended Dietary Allowances, for protein, vitamins and minerals, except with regard to calcium and iron, for which intakes for all averaged about two-thirds of the Recommended Dietary Allowances. Given the significant main effects of sex for fat and calorie intakes in Table 7, mean intakes were calculated separately for male and female respondents from steps 1 and 2 of the regression analysis. The corresponding means were approximately 2400 calories and 90 g of fat for males and 1875 calories and 70 g of fat for females. There was a significant age-by-dieting interaction for calories and fat such that the dieting effect was more pronounced for older subjects, mostly due to high levels of consumption by older low dieters: younger low dieters consumed 2135 calories and 78 g of fat and high dieters, 1922 calories and 70 g of fat, while older low dieters consumed 2435 calories and 93 g of fat and high dieters, 1183 calories and 71 g of fat. Physical Activity Index and Body Image Finally, the relationships of Weight Percentile and Dieting Status with Physical Activity Index and Body Image were examined using MR. As shown in Tables 6 and 7, there was no significant effect of either Weight Percentile or Dieting Status on Physical Activity Index. The Physical Activity Index sample mean is 1395. Thus, a 100-lb person would be expected to expend 139 500 total calories per year in the leisure time activities surveyed. Males (PA Index=1711), however, expended more energy in leisure time activities than did females (1207), and older adolescents (1711) expended more than younger adolescents (1079). Table 7 shows further that Weight Percentile in the first step of the MR was significantly related to both Body Me and Body Wish. In the case of Body Me, this result shows that students’ body images, overall, were at least somewhat accurate. This is consistent with the finding also of main effects for both sex and age on steps 1 and 2: males judged themselves to be larger (2·92 units) than females (2·74) and older respondents larger (2·91) than younger (2·69). In the case of Body Wish, this effect of Weight Percentile may reflect the culturally-shared desirability of thinness. That is, when most chose a drawing of one of the thinner bodies as the way they wanted to look, and when a drawing of a fatter body was more often chosen by students of higher weight, as the preceding result for Body Me indicates, Body Wish (the difference between these two) subsequently yielded the result seen for Weight Percentile in Table 7. This interpretation seems to be corroborated by the finding that girls, regardless of weight and dieting status, wished to be thinner by approximately 0·65 of a unit, while boys did not. For the second regression step for Body Me, there was no effect of Dieting Status. This finding is of some interest in relation to the alternative: a positive, significant beta would have shown that when actual body shape is controlled, by

68

J. R. CONTENTO ET AL.

inclusion of Weight Percentile, dieters see themselves as fatter than non-dieters. In this sense we did not see body image distortion among dieters across the sample as a whole. For Body Wish, a significant effect of Dieting Status was found, such that low dieters as a group wanted to be only slightly thinner than they were, while high dieters wanted to be quite a bit thinner.

D Our study with 411 adolescents examined the relation of Weight Percentile and Dieting Status to what adolescents valued or wanted in the food they ate; the criteria they used to choose foods; the quality and quantity of foods they actually ate; their physical activity levels and body image. The sample was deliberately selected to include both males and females, and adolescents who differed in age (i.e. predominantly 11to-12- and 16-to-18-year-olds) so as to be able to examine whether there were gender and age effects for the variables of interest. However, there were overall no interactions between sex, age, Weight Percentile and Dieting Status. This means that although boys dieted less than girls, when boys did diet, they did not behave differently from girls who dieted on the psychological and behavioral variables we examined; nor did higher-weight boys differ from higher-weight girls on these variables. Nor did age make much of a difference except that the dieting effect was more pronounced for older subjects in terms of calories and fat intake. This is despite the expected findings of main effects such that regardless of weight and dieting status, boys valued the healthfulness and disapproved of the fatteningness of foods less than girls, ate more calories and fat, were more active, and had less of a desire to be thin; and older adolescents used taste as a criterion for food choice more than younger adolescents, ate more fat and were more active. Consequently, the discussion below will focus on the ways in which fatter adolescents differed from thinner adolescents and high dieters differed from low dieters from analyses in which age and sex were controlled. First, however, it should be noted that there were some commonalities. In terms of what adolescents wanted from their foods (i.e. evaluations of food attributes), essentially all wanted their foods to be healthful, tasty, convenient (available) and not fattening. They seemed neutral about the sugar content and cost. In terms of the food choice criteria by which they selected the foods they ate (i.e. correlational indices), taste, availability and convenience were important for all and cost was not. Weight Status The multiple regression analyses indicate that Weight Percentile (a variable based on weight-for-height) was uniquely associated with relatively few variables. In Tables 2 and 3, we saw that fatter adolescents were more inclined than thinner ones to want their foods not to be fattening. Subsequent tables examined whether these motivational tendencies were translated into action. Tables 4 and 5 capture this translation from the adolescent’s point of view, by describing whether the foods most often consumed were ones believed by the adolescent to be less fattening. When dieting was controlled, high weight adolescents were not seen to choose foods they believed to be less fattening. Tables 6 and 7 use an objective point of view in the sense that experts’ beliefs about properties of foods were used to characterize aspects

ADOLESCENT WEIGHT AND DIETING STATUS

69

of dietary quality (including the “fattening” aspect). Again it was seen that there was no behavioral counterpart to the difference, as a function of weight percentile, in evaluations of fatteningness. By design, when dieting is controlled, the weight variable concerns weight isolated from dieting. Thus it makes sense that there is no avoidance of actual choice of fattening foods or foods higher in calories, fat or sugar by higher-weight people when weight is isolated from dieting; to a large extent, dieting is avoidance of fattening foods. However, the fact that higher-weight persons are still more likely to believe that a food is “bad” if it is fattening even with dieting status controlled, combined with the approximately 25% overlap in variance between the weight and dieting variables may mean simply that some but not all persons of higher weight act on their motivation to avoid fattening food by dieting. The findings for the desire for convenience (Easy Evaluation) would appear to explain some of the cases in which such action is not taken: higher-weight adolescents appeared to have a countervailing desire for convenient foods—which tend to be snacks that are high in sugar or fat. However, because higher-weight adolescents were not more likely than others to actually eat foods they believed to be convenient (from the Easy CI results), this aspect of higher-weight adolescents’ food choice motivations is less clear from our study. No comparison was attempted with an objective standard here (as for Fattening CI results) because convenience is so highly subjective. Generally, the predominant pattern of the results was for Weight Percentile to be replaced by Dieting Status as a significant predictor of food choice. For ten out of 13 variables where Weight Percentile seemed to have a significant association, the association disappeared when Dieting Status was controlled. Weight Percentile effects survived entry of Dieting Status into the analyses only for Fattening Evaluation (as discussed earlier in this section), for Body Me and Body Wish. The latter two findings merely confirmed that adolescents’ body images were generally accurate overall. The frequent displacement of Weight Percentile by Dieting Status strongly indicates that these two variables should be analysed simultaneously in research on psychological and behavioral aspects of food choice and obesity. If the present study had been conceptualized only as concerning weight status or obesity, and if our analyses had not included a measure of dieting, many associations properly attributable to Dieting Status would have been credited to Weight Percentile instead. For example, if dieting had not been taken into consideration, we would have concluded that weight had a significant effect on many criteria used in everyday food choice—particularly criteria related to health consequences and weight control factors. We would also have concluded that weight status is a predictor of the quality and quantity of diets consumed. Indeed, given numerous attempts to relate food consumption to obesity, it is especially striking that if quantity of food consumed had been studied only in terms of weight status, it would have been concluded that fatter adolescents ate less than thinner ones. Yet entry of Dieting Status into regression analyses along with Weight Percentile indicated clearly that Dieting Status accounted for the differences initially seen in our sample for Weight Percentile. Results concerning convenience illustrate another potential benefit of analysing these two variables simultaneously. Specifically, here we found that the substantial, positive correlation of weight and dieting served to obscure or suppress relations of the form seen with Easy Evaluation, where one of the two status variables (when controlled for the other) has a positive relation with a dependent variable and the

70

J. R. CONTENTO ET AL.

other has a negative relation. Such a finding is entirely plausible; it is logically possible that the two variables have opposite influences when each variable is considered in isolation, and that these influences cancel each other out when considered together. The Sugar Consumption dependent variable yielded essentially this same form of results, except that the beta weight for Weight Percentile just missed reaching statistical significance. Again, these results indicate that it is crucial that studies involving weight status should also take into account the dieting status of participants. Dieting Status What, then, is the relation of Dieting Status to the psychological and behavioral variables examined? First, high dieters (i.e. those who dieted “often”) were significantly more likely than low dieters (i.e. predominantly those who “never” dieted) to want their foods not to be fattening. Presumably this finding connects a behavioral characterization of adolescents, Dieting Status, with its motivational underpinnings. One of several illustrations of this motivational-behavioral connection was in the fact that for high dieters, the within-person correlations between foods they ate and foods they believed to be fattening was negative. Compared to their low-dieting counterparts, they also ate fewer of the foods judged by experts to be fattening, and to have fewer calories, less fat and less sugar in their diets. Next, high dieters valued taste and convenience less than low dieters. High dieters may thus be saying that they will forego taste and convenience to some degree in order to obtain less-fattening food. These differential valuations of taste and convenience would be expected to lead to stronger associations of perceived taste and convenience with actual food choices among low dieters as compared with high dieters—as was seen with fatteningness of foods. However, the results did not go this way; taste and convenience were equally—and highly—important in both groups for actual food choice. Thus it is possible that high dieters are, in effect, saying that although they believe they should not choose foods according to taste or convenience, that is what they actually do. However, for the choice criterion of taste, a different possibility is suggested by the tendency of people to come to like what they eat (e.g. Birch & Marlin, 1982; Pliner, 1982; Rozin, 1988)—as opposed to eating what they like. The reasoning here begins with the premise that dieters are people whose food consumption behavior is aligned (at least to some degree) with their wish to be thin. The present study gives evidence of this alignment, and of the distinctiveness of dieters’ food consumption patterns, respectively, in the results from the Fattening CI variable and the Food Quality and Nutrient Intake variables. Initially dieters may put aside motivations of taste and convenience in the service of their weight-reduction goals. Having arrived at their particular food choices, dieters then may, through familiarity, come to have greater liking of the foods they eat. Although we have no direct evidence of this sequence, we do find the various cross-sectional findings to coincide rather neatly with the notion that liking for foods is a consequence as well as a cause of food choices. Such an interpretation of these results is also consistent with findings that familiarity or exposure is a major determinant of food acceptance patterns in young children and that adults eating low-fat or reduced-salt diets come to judge them as more pleasant than before (Bertino, Beauchamp & Engelman, 1982; Mattes, 1993).

ADOLESCENT WEIGHT AND DIETING STATUS

71

Similar reasoning could be applied to explaining the divergence between the Easy Evaluation and Easy CI variables: dieting adolescents may, with direct experience, find less-fattening foods also to be less inconvenient than they previously thought these foods would be. If this is so, the divergence between the results for Easy Evaluation and Easy CI found for low weight vs. high weight adolescents also may not be problematic—contrary to our suggestion earlier in this discussion when considering the finding of differences by weight status in the importance of the criterion of convenience. Health Conscious or Weight Conscious? From the fact that high dieters eat higher quality diets, it is tempting to conclude that high dieters are more health conscious. However, our results indicate that this is not the most likely explanation. High dieters are indeed more likely than low dieters to eat foods that they believe to be healthful, that experts judge to be healthful, and that are similar in vitamin and mineral composition though lower in calories, fat and sugar. However, we did not see a low-high dieter difference in evaluation of how “good” it is to eat foods that are healthful. These findings suggest that the more healthful eating by high dieters is not a consequence of wanting positive health effects from their food choices; it is a consequence of wanting a weight control effect or the “physical” effect, thinness. It so happens that the particular foods and food consumption patterns that promote thinness also tend to be those that promote health in the sense of prevention of diseases such as heart disease and cancer. Researchers of health motivations in adolescent food choice thus have as much need to be cognizant of the dieting status of participants as researchers of adolescent obesity—because of the potential for misinterpretation of findings due to confounding from positive correlations among healthful eating, dieting and obesity. Questions About Physical Activity Given the concern of high dieters for thinness, we might expect that they would also be more physically active than low dieters. Yet such a difference was not found. This lack of difference might reflect a preference for dieting over exercise as a way to control weight. However, when thinking about this lack of difference it may be important to keep in mind that the dieting variable, because of its measurement, specifically isolates those who behave in particular ways with respect to food. The expectation of an association of dieting and physical activity is based on an implicit assumption that this behavior is a manifestation of a motivational orientation that might also drive exercise behavior. Evidently in this sample, dieting and exercise behavior are more independent than this logic implies. However, the result may also be due to limitations in the precision of our procedure for measuring physical activity. That is, it is possible that this measure is sensitive only to large differences in physical activity, as obtained with sex and age comparisons, and not to differences due to dieting or weight if these were smaller in magnitude. Even if this is so, it is of some interest that the latter magnitudes are smaller. Of course, it is also possible that self-presentation concerns among non-dieters or higher-weight persons attenuated differences that otherwise would have been seen. At same time, it is not clear on an a priori basis why this elimination of differences would occur for reports about exercise but not about food intake, beliefs or values.

72

J. R. CONTENTO ET AL.

Dieting or Dietary Restraint? A Cognitive Self-Regulation Integration Because dieting status shares some features with dietary restraint (and our indicators of dieting were components of measures of restraint), it is not surprising that some of the findings for dieting status in this study are similar to findings obtained for dietary restraint in terms of quantity of foods consumed and quality of diet. Several studies, for example, have reported that restrained eaters, like the high dieters in our study, ate fewer calories (Wardle et al., 1992) or had a lower overall food intake compared to unrestrained eaters (Laessle et al., 1989; Tuschl et al., 1990; Van Strien et al., 1986). Quality of diet has also been recently assessed in a few studies, using different measures. In the study of Wardle et al. (1992), quality of diet was measured in terms of the three categories of macronutrients— carbohydrates, fats, and proteins. The restrained adolescents ate less of all three categories. In a study of adults that examined the differences in reported consumption of certain food items by normal-weight restrained and unrestrained eaters (Tuschl et al., 1990), few differences were found in terms of basic meal items or snacks. Restrained eaters did not avoid animal products nor were they more likely to select items that might be termed healthy; nor were they more likely to avoid sweet snacks. However, they did avoid fats more—oils, mayonnaise, butter and high-fat dairy, and ate more low-fat dairy. The dieters in our study, on the other hand, were significantly more likely to eat foods judged by experts to be healthy and high in vitamins and minerals and they ate less sugar as well as less fat. We found no studies that examined restraint in relation to the criteria for everyday food choice with which to compare the findings reported here. The closest are studies conducted by Wardle and colleagues (Wardle & Beales, 1986; Wardle et al., 1992) in which restraint was examined in relation to the food attitudes “like”, “good for you”, “makes you fat” and “feel guilty about eating it”. It was found that highrestraint adolescents were more likely to regard all but the most slimming foods as “fattening” and as less “good for you” and felt more guilty about eating them. Food attitudes were not directly related to consumption of food items as was done in this study. Diet was separately assessed using the 24-h recall and quality was judged in terms of intake of the three major macronutrients as indicated above. From a social psychological perspective, the picture that emerges from this study as a whole is of adolescents who, to varying degrees, want particular consequences from their food choices, and who align their food choice behaviors with their goals to some extent. This alignment is a process of food choice that we have labelled cognitive self-regulation (Michela & Contento, 1986). In this study, a “psychology of dieting” appears to involve wanting to be thinner and then making food choices that are believed to accomplish this goal. Our study sheds little light on whether there is a distinct “psychology of being thinner versus fatter.” Evidently for this population and for the psychological and behavioral variables examined, lower weight and higher weight people do not differ very much when dieting status is taken into account. While dieting status would seem to be a determinant of food choice and weight status an outcome, it could also be that higher weight, and consequent body image dissatisfaction in a culture that values thinness may lead to dieting. The results presented here, as well as the results of other studies in this area, show that the interrelationships are quite complex. Our finding that a “psychology of dieting” seemed more important than a “psychology of being thinner versus being fatter” is consistent with the emerging view from the extensive research on obesity

ADOLESCENT WEIGHT AND DIETING STATUS

73

and restraint that “dietary restraint has now largely replaced other aspects of eating style as the critical individual difference variable in experimental studies of eating behavior” (Wardle et al., 1992, p. 168). Indeed, dietary restraint has received such wide attention partly because it seems to provide a way to account for varying, and sometimes conflicting, results on the regulation of intake by obese and normalweight individuals (Herman & Polivy, 1975; Ruderman, 1986). Our findings suggest that a psychology of dieting by itself, irrespective of other aspects of dietary restraint such as counter-regulation, appears to affect the criteria used for everyday food choice by aligning food choices with desired goals. The alignment of weight-reduction goals and behaviors is seen most clearly in the difference between high and low dieters’ within-person correlations between the foods they believe to be fattening and the food they eat. High dieters, who have a greater wish to be thin, are more likely to eat foods they believe to be less fattening. We have no way of assessing whether dieters’ weight reduction goal is actually being accomplished in the sense of whether dieters in our sample are thinner than they otherwise would be. Nevertheless, from results of the Food Consumption and Nutrient Intake variables, it is clear that the food choices reported by adolescent high dieters are, for weight reduction, superior to the food choices reported by low dieters. If adolescents do enact a cognitive self-regulating process for food choice, and if knowledge moderates the relation between nutrition-related goals and behaviors, implications follow for nutrition education practice. Specifically, nutrition education can influence food consumption behavior by providing people with knowledge of behavior-outcome linkages. Although adolescents appear to have definite knowledge about the extent to which particular foods are fattening, they probably know less about the consequences of particular foods or food components for risks of chronic disease. Adolescents’ wish to be thin provides a possible starting point for teaching about food components and their relation to health as well as to fatness. Such teaching need not buy into the culturally-promoted image of thinness and its potential consequences of anorexia and bulimia. Instead, adolescents could be taught that the choice of low-fat, low-sugar, high-fiber foods will not only help them achieve their weight-reduction goals, but is health-promoting as well.

R Allen, K. M., Thombs, D. L., Mahoney, C. A. & Daniel, E. L. (1993) Relationships between expectancies and adolescent dieting behaviors. Journal of School Health, 63, 176–81. American food preferences (1983) Camden, NJ: Campbell Soup Company. Axelson, M. L. & Brinberg, D. (1989) Determinants of food choice—a social psychological perspective. New York: Springer-Verlag. Ajzen, I. & Fishbein, M. (1980) Understanding attitudes and predicting social behavior. Englewood Cliffs, NJ: Prentice-Hall. Balogh, M., Kahn, H. A. & Medalie, J. H. (1971) Random repeat 24-hour dietary recalls. American Journal of Clinical Nutrition, 24, 304–10. Beidler, L. M. (1982) Biological basis of food selection. In Barker (Ed.), The psychobiology of human food selection Pp. 3–15. Westport, CT: AVI. Bertino, M., Beauchamp, G. K. & Engelman, K. (1982) Long-term reduction in dietary sodium alters the taste of salt. American Journal of Clinical Nutrition, 36. 1113–44. Birch, L. L. & Marlin, D. W. (1982) I don’t like it; I never tried it: effects of exposure on two year old children’s food preferences. Appetite, 3, 353–60.

74

J. R. CONTENTO ET AL.

Block, G. (1982) A review of validations of dietary assessment methods. American Journal of Epidemiology, 115, 492–505. Brinberg, D. & Durand, J. H. (1983) Eating at fast food restaurants—an analysis using two behavioral intention models. Journal of Applied Social Psychology, 13, 459–72. Bull, N. L. (1988) Studies of the dietary habits, food consumption, and nutrient intakes of adolescents and young adults. World Review of Nutrition and Dietetics, 57, 24–74. Contento, I. R., Basch, C., Shea, S., Gutin, B., Zybert, P., Michela, J. L. & Rips, J. (1993) Relationship of mothers’ food choice criteria to food intake of preschool children: Identification of family subgroups. Health Education Quarterly, 20, 243–59. Contento, I. R., Michela, J. L. & Goldberg, C. J. (1988) Food choice among adolescents: population segmentation by motivations. Journal of Nutrition Education, 20, 289–98. Court, J. M. (1988) Nutrition and adolescents: an overview of concerns in Western society. The Medical Journal of Australia, 148, S2–7. Cronk, C. E. & Roche, A. F. (1982) Race- and sex-specific reference data for triceps and subscapular skinfolds and weight/stature2. American Journal of Clinical Nutrition, 35, 347–54. Davies, E. & Furnham, A. (1986) The dieting and body shape concerns of adolescent females. Journal of Child Psychology and Psychiatry, 27, 417–27. Drewnowski, A. & Yee, D. K. (1987) Men and body image: are males satisfied with their body weight? Psychosomatic Medicine, 49, 626–34. Farthing, M. C. (1991) Current eating patterns of adolescents in the United States. Nutrition Today, 35–9. Garner, D. M., Garfinkel, P. E., Schwartz, D. & Thompson, M. (1980) Cultural expectations of thinness in women. Psychological Reports, 47, 483–91. George, R. S. & Krondl, M. (1983) Perceptions and food use of adolescent boys and girls. Nutrition and Behavior, 1, 115–25. Grinker, J. (1990) Sensory factors in the control of food intake: taste and palatability. Appetite, 14, 69–72. Heatherton, T. F., Herman, C. P., Polivy, J., King, G. A. & McGee, S. T. (1988) The (mis)measurement of restraint: an analysis of conceptual and psychometric issues. Journal of Abnormal Psychology, 97, 19–28. Herman, C. P. & Polivy, J. (1975) Anxiety, restraint, and eating behavior. Journal of Abnormal Psychology, 84, 666–72. Johnson, W. G., Lake, L. & Mahan, J. M. (1983) Restrained eating: measuring an elusive construct. Addictive Behavior, 8, 413–18. Kenny, D. A. (1985) Quantitative methods for social psychology. In Lindzey, G. & Aronson, E. (Eds.), Handbook of social psychology, 3rd ed., Vol. 1, Pp. 487–508. New York: Random House. Krondl, M. (1990) Conceptual models. In G. H. Anderson (Ed.), Diet and behavior: multidisciplinary approaches Pp. 5–15. New York: Springer-Verlag. Krondl, M. & Lau, D. (1983) Social determinants in human food selection. In Barker (Ed.), The psychobiology of human food selection Pp. 139–51. Westport, CT: AVI. Kuczmarski, R. J., Brewer, E. R. & Cronin, F. J. (1986) Food choices among white adolescents: the Lipid Research Clinics prevalence study. Pediatric Research, 20, 309–15. Laessle, R. G., Tuschl, R. J., Kotthaaus, B. C. & Pirke, K. M. (1989) Behavioral and biological correlates of dietary restraint in normal life. Appetite, 12, 83–94. Lau, D., Hanada, L., Kaminsky, O. & Krondl, M. (1979) Predicting food use by measuring attitude and preference. Food Product Development, 13, 66. Lewin, K., Dembo, T., Festinger, L. & Sears, P. S. (1944) Level of aspiration. In L. M. Barker (Ed.), Personality and the behavior disorders Pp. 333–78. New York: Ronald. Mattes, D. (1993) Fat preference and adherence to a reduced fat diet. American Journal of Clinical Nutrition, 57, 373–81. Mertz, W. (1992) Food intake measurements: is there a gold standard? Journal of the American Dietetics Association, 92, 1463–5. Michela, J. L. (1990) Within-person correlational design and analysis. In Hendrick, C. & Clark, M. (Eds.), Review of personality and social psychology Vol. 11, Pp. 279–311. Beverly Hills, CA: Sage. Michela, J. L. & Contento, I. R. (1984) Spontaneous classification of foods by elementary school-aged children. Health Education Quarterly, 11, 57–76.

ADOLESCENT WEIGHT AND DIETING STATUS

75

Michela, J. L. & Contento, I. R. (1985) Evidence for the validity of new measures of food frequency and food choice motivations for social psychological research on determinants of children’s food choices. Health Promotion Reports no. 85–5. New York: Center for Health Promotion, Teachers College, Columbia University. Michela, J. L. & Contento, I. R. (1986) Cognitive, motivational, social, and environmental influences on children’s food choice. Health Psychology, 5, 209–30. National Center for Health Statistics. (1983) Dietary intake source data: United States 1976–80. In Vital and health statistics. DHSS Publication no. PHS 83-1681. Series 11, no. 231. Washington, DC: U.S. Government Printing Office. NUTRI-CALC. (1983) Disks, operating manual, [based on Nutritive Value of Foods, Home and Garden Bulletin no. 72. USDA (1981).] Penn Yan, NY: PCD Systems, Inc. Nylander, I. (1971) The feeling of being fat and dieting in a school population. Acta Sociomedica Scandinavica, 3, 17–26. Pliner, P. (1982) The effects of mere exposure on liking for edible substances. Appetite, 3, 283–90. Polivy, J. & Herman, C. P. (1985) Dieting and binging: a causal analysis. American Psychologist, 40, 193–201. Rosen, J. C. & Gross, J. (1987) Prevalence of weight reducing and weight gaining in adolescent girls and boys. Health Psychology, 6, 131–47. Rozin, J. P. (1988) Social learning about food by humans. In Zentall, T. R. & Galef, Jr., B. G. (Eds.), Social learning: psychological and biological perspectives Pp. 165–87. Hillsdale, NJ: Erlbaum. Rozin P. & Fallon, A. E. (1981) The acquisition of likes and dislikes for foods. In Solms, J. & Hall, R. L. (Eds.), Criteria of food acceptance Pp. 35–48. Zurich, Switzerland: Foster Verlag. Ruderman, A. J. (1986) Dietary restraint: a theoretical and empirical review. Psychological Bulletin, 99, 247–62. Shepherd, R. (1989) Factors affecting food preferences and choice. In R. Shepherd (Ed.), Handbook of the psychophysiology of human eating Pp. 3–24. New York: Wiley. Story, M. & Resnick, M. D. (1986) Adolescents’ views on food and nutrition. Journal of Nutrition Education, 18, 188–92. Storz, N. S. & Greene, W. H. (1983) Body weight, body image, and perception of fad diets in adolescent girls. Journal of Nutrition Education, 15, 15–18. Stunkard, A. J. & Messick, S. (1985) The three-factor eating questionnaire to measure dietary restraint, disinhibition, and hunger. Journal of Psychosomatic Research, 29, 71–83. Taylor, H., Jacobs, D. R. & Schucker, B. (1978) A questionnaire for the assessment of leisure time physical activities. Journal of Chronic Diseases, 31, 741–54. Towler, G. & Shepherd, R. (1992). Application of Fishbein and Ajzen’s expectancy-value model to understanding fat intake. Appetite, 18, 15–27. Triandis, H. C. (1979) Values, attitudes, and interpersonal behavior. Nebraska Symposium on Motivation, 195–259. Truswell, A. S. & Darnton-Hill, I. (1981) Food habits of adolescents. Nutrition Reviews, 39, 73–88. Tuorila-Ollikainen, H., Lahteenmaki, L. & Salovaara, H. (1986) Attitudes, norms, intentions and hedonic responses in the selection of low salt bread in a longitudinal choice experiment. Appetite, 7, 127–39. Tuschl, R. J. (1990) From dietary restraint to binge eating: some theoretical considerations. Appetite, 14, 105–9. Tuschl, R. J., Laessle, R. G., Platte, R. & Pirke, K. M. (1990a) Differences in food choicefrequencies between restrained and unrestrained eaters. Appetite, 14, 9–13. Tuschl, R. J., Platte, R., Laessle, R. G., Stichler, W. & Pirke, K. M. (1990b) Energy expenditure and everyday eating behaviour in healthy young women. American Journal of Clinical Nutrition, 52, 81–6. United States Department of Agriculture. (1978) Nationwide food consumption survey, 1977–78 (Preliminary Report no. 2). Food and nutrient intake of individuals in one day in the United States, Spring 1977. Washington, DC: Office of the Director, USDA Science and Education Administration, 1978.

76

J. R. CONTENTO ET AL.

Van Strien, T., Frijters, J. E., Van Staveren, W. A., Defanes, P. B. & Deurenberg, P. (1986) The predictive validity of the Dutch Restrained Eating Scale. International Journal of Eating Disorders, 5, 747–55. Wallston, B. S. & Wallston, K. A. (1984) Social psychological models of health behavior: An examination and integration. In Baum, A., Taylor, S. E. & Singer, J. E. (Eds.), Handbook of psychology and health: Vol. 4. Social psychological aspects of health Pp. 23–53. Hillsdale, NJ: Lawrence Erlbaum Associates, Inc. Wardle, J. (1980a). Compulsive eating and dietary restraint. British Journal of Clinical Psychology, 26, 47–55. Wardle, J. (1980b). The assessment of restrained eating. Behavior Research and Therapy, 24, 213–315. Wardle, J. & Beales, S. (1986) Restraint, body image and food attitudes in children from 12 to 18 years. Appetite, 7, 209–17. Wardle, J. & Marsland, L. (1990) Adolescent concerns about weight and eating: a socialdevelopmental perspective. Journal of Psychosomatic Research, 34, 377–91. Wardle, J., Marsland, L., Sheikh, Y., Quinn, M., Federoff, I. & Ogden, J. (1992) Eating style and eating behaviour in adolescents. Appetite, 18, 167–83. Westenhoeffer, J. (1991) Dietary restraint and disinhibition: is restraint a homogeneous construct? Appetite, 16, 45–55. Worsley, A. (1978) Looking for determinants of eating behavior. Proceedings of the Nutrition Society of Australia, 3, 35–45. Worsley, A., Baghurst, P., Worsley, A. J., Coonan, W. & Peters, M. (1984) Australian 10year-olds’ perceptions of food (1): sex differences. Ecology of Food and Nutrition, 15, 231–46. Received 9 February 1994, revision 10 November 1994