- Email: [email protected]
Expectancies, dietary restraint, and test meal intake among undergraduate women
ARTICLE IN PRESS
Appetite 49 (2007) 30–37 www.elsevier.com/locate/appet
Research report
Expectancies, dietary restraint, and test meal intake among undergraduate women Robyn Syskoa,,1, B. Timothy Walshb, G. Terence Wilsona a
Department of Psychology, Rutgers, The State University of New Jersey, Eating Disorders Clinic, 41 C Gordon Road, Piscataway, NJ 08854, USA Department of Psychiatry, College of Physicians and Surgeons of Columbia University, New York State Psychiatric Institute, New York, NY, USA
b
Received 24 July 2006; received in revised form 1 November 2006; accepted 1 November 2006
Abstract This study investigated the relationship between self-reported dietary restraint and expectancies about caloric content on test meal consumption among undergraduate women. Participants completed two test meal sessions during which they were asked to consume as much milkshake from a covered opaque container as they wished. In one session, participants were instructed that the milkshake was made with high-calorie ingredients, and in the other that the milkshake was made with low-calorie ingredients. The milkshakes in both sessions were actually made with the same ingredients. Participants’ mean consumption was less on the low-calorie instruction day (402 g) than on the high-calorie instruction day (382 g), but the difference was not statistically significant. In addition, few significant relationships were observed between dietary restraint measures and total intake on either the low- or high-calorie instruction days. Thus, this study supports a growing body of literature indicating that scores on measures of dietary restraint are not related to the actual restriction of food intake. r 2006 Elsevier Ltd. All rights reserved. Keywords: Dietary restraint; Behavioral restriction; Eating behavior; Restraint theory
Introduction The term ‘‘dietary restraint’’ has been used to refer to a range of attitudes and behaviors, including food avoidance, ongoing attempts to restrict caloric intake to lose weight that are associated with unsuccessful dieting (Lowe, 1993), and a cognitive control over eating, which makes dieters vulnerable to episodes of uncontrolled eating when that control is disrupted (Polivy & Herman, 1985). Higher scores on measures of dietary restraint have been shown to predict the future development of binge eating (e.g., Stice & Agras, 1998), and dietary restraint is often suggested as a risk factor for the development of eating disorders (e.g., Polivy & Herman, 1985). A number of studies have evaluated the relationship between eating behavior, typically in an experimental Corresponding author.
E-mail address: [email protected] (R. Sysko). Current address: VA Connecticut Healthcare, 116B/Psychology, West Haven Campus, 950 Campbell Avenue, West Haven, CT 06516, USA. 1
0195-6663/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.appet.2006.11.002
paradigm, and self-report measures of dietary restraint. Much of this research has examined a theory described by Polivy, Herman, and colleagues. This theory, often termed the restraint hypothesis, postulates that restrained eaters, those who exert cognitive control over food consumption and score highly on the Restraint Scale (Herman & Polivy, 1975), will overeat after being exposed to a disinhibitor, such as breaking a dietary rule or consuming a forbidden food. The effect of disinhibitors like pre-loads (i.e., the consumption of a food prior to an experimental meal or taste test) on restrained eaters has been supported in numerous studies (for a review see Ruderman, 1986 or Lowe, 1993). These studies typically classify participants based on the median score of the Restraint Scale into low- and highrestrained groups. Participants consume a pre-load (e.g., milkshake) and subsequently complete a taste test (e.g., ice cream, cookies; Ruderman, 1986). The amount of food consumed during the taste test portion of the experiment is measured to determine the effect of the disinhibitor (the pre-load). High-restrained participants tend to eat
ARTICLE IN PRESS R. Sysko et al. / Appetite 49 (2007) 30–37
significantly more food during the taste test after consuming a high-calorie pre-load than low-restrained participants, suggesting that after breaking a dietary rule (i.e., consuming a high-calorie milkshake), the eating of highly restrained participants is disinhibited (Ruderman, 1986). Similar patterns of behavior are believed to play a role in the occurrence of binge eating (e.g. ‘‘I’ve blown my diet;’’ Fairburn, Marcus, & Wilson, 1993). Recent research has questioned whether the model described by Polivy, Herman, and colleagues accurately describes the relationship between dietary restraint and eating behavior. In a series of four experiments, Stice, Fisher, and Lowe (2004) examined whether commonly used self-report measures of dietary restraint were negatively correlated with food consumption. The experiments were conducted in diverse settings, including both laboratory and naturalistic eating environments, and participants included undergraduate women, women with bulimia nervosa, women with binge eating disorder, women without an eating disorder, and female patrons of a fast food restaurant (Stice et al., 2004). In all four studies, no relationship was observed between the measures of dietary restraint and caloric consumption, with one exception. Significant negative correlations were observed between the Dietary Intent Scale (DIS; Stice, 1998) and caloric intake and fat-gram intake for participants consuming a meal in a fast food restaurant. However, the correlation, while statistically significant, was ‘‘rather meager’’ (r ¼ 0.24, po0.05), and, as the authors noted, ‘‘there is still much room for improvement’’ (Stice et al., 2004, p. 55). Stice et al. (2004) concluded that the measures of dietary restraint utilized in the study were not valid assessments of true short-term food restriction. Martin et al. (2005) used a similar experimental design to examine the relationship between dietary restraint and meal consumption during a series of four meals. Individuals without an eating disorder were asked to consume sandwiches instead of their usual lunch on four different days. Across the four sessions, there was no effect of dietary restraint, as measured by the Three Factor Eating Questionnaire (TFEQ; Stunkard & Messick, 1988), on total intake during lunch (Martin et al., 2005). In addition, a recent study of eating behavior among patients with anorexia nervosa and individuals without an eating disorder also found that several widely used scales of dietary restraint did not correlate with total intake during a laboratory lunch meal for either participant group (Sysko, Walsh, Schebendach, & Wilson, 2005). Thus, while restrained individuals appear to increase food consumption in response to disinhibition in experiments involving a pre-load and a taste test (e.g., Ruderman, 1986), there does not appear to be a relationship between dietary restraint and food consumption during a standardized breakfast meal, meals at a university cafeteria (Stice et al., 2004), or lunch meals (Martin et al., 2005; Sysko et al., 2005). Some of the discrepancies observed between studies may be related to differences in experi-
31
mental design, which may have influenced the subsequent relationship between dietary restraint and eating behavior. For example, the expectations of participants during an eating behavior study may be particularly important in influencing total intake. Two previous studies examining the restraint hypothesis (Polivy, 1976; Spencer & Fremouw, 1979) illustrate the potential impact of cognitive factors in the study of dietary restraint. Polivy (1976) manipulated the expectations of restrained and unrestrained male undergraduates by instructing participants that they were receiving a ‘‘rich, high-calorie, gourmet, pudding-type dessert,’’ or a ‘‘low-calorie pudding’’ (Polivy, 1976, p. 239). After consuming the pudding, participants were presented with sandwich quarters, and the number of sandwiches eaten was measured. Expectations about the pudding were manipulated using both the instructions and the true caloric content of the food, including the following four conditions: (1) true high calorie (high-calorie pudding and high-calorie instructions); (2) true low calorie (low-calorie pudding and lowcalorie instructions); (3) fake high calorie (low-calorie pudding, high-calorie instructions); and (4) fake low calorie (high-calorie pudding, low-calorie instructions). Polivy (1976) found an interaction between dietary restraint and perceived calories (what participants believed they ate), suggesting that cognitive factors significantly affected meal consumption; however, the overall effect of the manipulation was not very large. Spencer and Fremouw (1979) conducted a similar study, but the cognitive manipulation was limited to the perception of whether the food consumed was high or low calorie. High- and low-restrained female undergraduates were given a liquid pre-load and told that the pre-load was either high or low calorie. In both the ‘‘told high-calorie’’ and ‘‘told low-calorie’’ conditions, the shake had an equivalent caloric density. After consuming the shake pre-load, participants completed a taste test with three ice creams. High-restrained participants consumed significantly more of the ice cream after the told high-calorie condition than the told low-calorie condition, indicating that the cognitive manipulation was successful. For restrained eaters, cognitive expectations about the caloric content of a pre-load produced disinhibited eating in the laboratory, regardless of the true caloric content of the pre-load. The current study was designed to examine the effect of expectations about the caloric content of a food on subsequent intake. Unlike both the Polivy (1976) and Spencer and Fremouw (1979) studies, this experiment did not use a pre-load paradigm. Studies utilizing a pre-load paradigm investigate whether the violation of a dietary rule (e.g., consuming a pre-load) by restrained eaters overrides their self-imposed dietary restriction during a subsequent taste test. The current study instead measured the effect of the cognitive manipulation on the consumption of a meal (lunch), which more closely approximates eating outside of the laboratory. The second aim of the current study was to
ARTICLE IN PRESS 32
R. Sysko et al. / Appetite 49 (2007) 30–37
examine the relationship between self-report measures of dietary restraint and total intake during the test meals, which would determine whether measures of dietary restraint are in fact predictive of behavioral restraint. The experimental design of the current study also allowed for an evaluation of a hypothesis suggested by Stice, Presnell, Lowe, and Burton (2006) that the use of a cognitive manipulation can lead to a more pronounced relationship between dietary restraint and eating behavior. We hypothesized that in the current study test meal consumption would be affected by a cognitive manipulation of participants’ expectations about the caloric content of the food, and that the effect of the cognitive manipulation would vary based on participants’ self-reported amount of dietary restraint. Specifically, the two main hypotheses were that (1) all participants would eat less when they believed they were consuming a high-calorie shake, and (2) that there would be a relationship between higher scores on measures of dietary restraint and lower amounts of total intake on the high-calorie instruction day. Methods Participants Forty-four female undergraduates participated in two test meal sessions in exchange for course credit. All participants provided informed consent. This study was approved by the Institutional Review Board of Rutgers University. Procedures Participants indicated their interest in the experiment on the psychology department’s website, and subsequently received an email with an attachment containing the consent form for the study and research questionnaires (described below). Eligibility was determined after the consent and questionnaires were completed. Six of the 50 students who signed up for the study on the website decided not to participate in the experiment by not returning a completed consent and questionnaires. The two test meal sessions occurred on non-consecutive days, and participants were told to fast (consume nothing but water) for the 4 h prior to their sessions. Before consuming the test meal, participants completed 15 cm Visual Analog Scales (VAS) assessing the constructs of fullness, hunger, sickness, anxiety, depression, and desire to eat, with the anchors of ‘‘not at all’’ and ‘‘extremely.’’ Participants also recorded when they last ate, how many hours of sleep they had in the previous evening, and whether they were on a diet to lose weight, as Lowe (1993) suggested that this single item may be a more valid measure of dietary restraint than other measures of restraint. The participants listened to one of two sets of taperecorded instructions. Before both test meals, participants were told to eat as much of the shake as they liked and that
the shake was their lunch for the day. In one meal session, participants were instructed that they were consuming a milkshake made with high-calorie ingredients (i.e., premium ice cream and whole milk) and for the other meal session, participants were instructed that they were consuming a shake made with low-calorie ingredients (i.e., fat-free frozen yogurt and skim milk). The shakes in both sessions were made with the same ingredients, which included all the low- and high-calorie ingredients mentioned in the instructions. Thus, the instructions were intended to elicit an expectation about the caloric content of the shakes; however, there was not a true difference in caloric content. Each participant received two shake flavors (e.g., chocolate and vanilla) in an attempt to conceal the similarity of the shakes. The high- and lowcalorie instructions, as well as the flavors of the shake, were counterbalanced to ensure that there would not be unequal pairings of each flavor of shake with the high- or lowcalorie instructions. Prior to the start of the study, three shake recipes were pilot tested with the staff of the Eating Disorders Research Unit (EDRU) of the New York State Psychiatric Institute/Columbia University Medical Center. The VAS ratings of the EDRU staff were reviewed, and we chose a shake rated as being moderately ‘‘rich,’’ such that it would be possible that the shake could be made with highcalorie ingredients. Participants were provided with approximately 1075 g of chocolate, vanilla, or strawberry shake (1.12 kcal/g, or approximately 1205.5 kcal; range of 988–1190 g or 1107.9–1334.5 kcal). When participants finished consuming the shake, they rang a doorbell, at which point they completed a final VAS assessment of hunger, fullness, sickness, anxiety, depression, desire to eat, liking of the shake, pleasantness of the shake, and taste perceptions (sweet, rich, bitter, novelty), and estimated the amount of food eaten during the test meal session (in calories and ounces). A pictorial example of volume was provided to assist in this estimation. Some ratings after the second test meal measured the success of the intervention in altering expectations about the caloric content of the shake (see the description of the manipulation check below). The shake was weighed before and after the meal to determine the total amount consumed during the test meal session. After the second test meal, the participant’s height and weight was measured. Manipulation check A manipulation check was used to determine whether participants believed the cognitive manipulation regarding the caloric density of the shake. To be included in the analyses, participants were required to meet two of the following three conditions: (1) participants had to rate the shake on the high-calorie instruction day to be richer than the shake on the low-calorie instruction day on the post-meal VAS ratings for richness (e.g., VAS measurement of 10.4 cm on the high-calorie instruction day and
ARTICLE IN PRESS R. Sysko et al. / Appetite 49 (2007) 30–37
6.8 cm on the low-calorie instruction day); and after the second meal both (2) correctly indicating the instructions played at both test meals (i.e., circling the correct response to both: ‘‘Today’s meal was made with high/low-calorie ingredients’’ and ‘‘Last session, the shake was made with high/low-calorie ingredients’’), and (3) when asked to compare the shakes consumed in both sessions, participants needed to indicate that the high-calorie shake was richer than the low-calorie shake (e.g., compared to last session, the shake today was: much less rich/extremely rich on a scale of 0–7). Measures The following self-report measures were given to participants to assess dietary restraint, concerns about shape and weight, and positive and negative affects. Self-report measures of dietary restraint Four measures (DIS; TFEQ; Dutch Eating Behavior Questionnaire, DEBQ; and Eating Disorder Examination Questionnaire, EDE-Q) were used to assess dietary restraint. The DIS (Stice, 1998) is a nine-item scale assessing dietary restraint and measures attempts to control weight or prevent weight gain. The TFEQ (Stunkard & Messick, 1985) assesses behaviors to produce weight loss, and includes three subscales: disinhibition, restraint, and hunger. Westenhoefer (1991) derived two additional subscales (flexible and rigid control) from a factor analysis of the TFEQ. The DEBQ (van Strien, Frijters, Bergers, & Defares, 1986) is a popular method of assessing dietary restraint, focusing on weight loss and maintenance, with three subscales (restrained eating, emotional eating, and external eating). Finally, the EDEQ (Fairburn & Beglin, 1994) is a self-report questionnaire that measures dietary restraint, shape and weight concern, and eating disordered behaviors. Assessments of shape and weight concerns and affect The Body Shape Questionnaire (BSQ; Cooper, Taylor, Cooper, & Fairburn, 1987) and the Positive and Negative Affect Scales (PANAS; Watson, Clark, & Tellegen, 1988) were interspersed in the computerized assessment package with the four measures of dietary restraint in an attempt to distract participants from our focus on dietary restraint. As these measures were not administered to test specific hypotheses, the data are not presented here.
33
(calories and ounces) on the high- and low-calorie instruction days were compared using paired t-tests. Effect sizes (d) were calculated as the mean difference between the test meal days (e.g., low- and high-calorie instruction days) for a given variable divided by the mean standard deviation of that variable over the test meal sessions. To test the second main hypothesis, and to analyze all of the selfreported measures of dietary restraint (DIS, TFEQ, DEBQ, and EDE-Q) as continuous variables, Pearson correlation coefficients (r) were calculated between the restraint measures and total intake on the high-calorie instruction day. The TFEQ scoring (Stunkard & Messick, 1985) was used to derive three subscales (restraint, disinhibition, and hunger) and an additional two subscales (rigid control, flexible control) described by Westenhoefer (1991). Secondary data analysis Similar to Polivy (1976) and Spencer and Fremouw (1979), we divided participants into high- and low-restraint groups using a median split on each of the measures of dietary restraint to allow for comparisons with previous research on the restraint hypothesis. One-way ANOVAs were calculated between the dependent variables of intake on the low- or high-calorie instruction day, or the change in intake between the low- and high-calorie instruction days, and the independent variable of the median split (low-scoring vs. highscoring) for each of the measures of dietary restraint (DIS, TFEQ disinhibition, TFEQ restraint subscale, TFEQ hunger subscale, TFEQ flexible control subscale, TFEQ rigid control subscale, DEBQ restrained eating subscale, or EDE-Q restraint subscale). Finally, Pearson correlation coefficients were calculated between each of the self-report measures of dietary restraint and total intake on the low-calorie instruction day to determine whether there was a relationship between degree of dietary restraint and intake when participants believed the shake was low calorie. A power analysis conducted before recruitment began indicated that using a within-subjects design and a total sample size of 33, an a level of 0.05, and assuming a population standard distribution of 1.0, there was an 80% chance of finding a medium effect (Cohen’s d ¼ 0.5), suggesting adequate power to detect a clinically meaningful difference (Murphy & Myors, 2004). All statistical calculations were performed using SPSS for Windows, version 11.5. All means are reported7standard deviations. Results
Statistical analysis Demographic characteristics Primary statistical analyses Means and standard deviations were calculated for the participants’ demographic characteristics. To test the first main hypothesis, average amounts consumed on the highand low-calorie instruction days were compared using paired t-tests. Similarly, pre- and post-meal VAS ratings and the estimates of intake during the test meal sessions
Of the 44 participants who completed both test meal sessions, 38 (86.3%) met the criteria for the manipulation check and are included in the analyses described below. The meals were, on average, 13.4 days apart (range of 2–38 days). The mean age of the participants was 19 years and the mean body mass index (BMI; kg/m2) was 22.4973.44
ARTICLE IN PRESS 34
R. Sysko et al. / Appetite 49 (2007) 30–37
(range 18.43–37.42). Nineteen participants (50.0%) were Caucasian, 13 (34.2%) were Asian, 3 (7.9%) were Hispanic, 2 (5.3%) were African-American, and 1 (2.6%) was Indo-Caribbean and Afro-Caribbean. The mean scores7standard deviations and ranges, in parentheses, for the self-report measures, including restraint, shape and weight concerns, and other psychological symptoms are as follows: DIS, 19.3076.06 (9–31); TFEQ restraint subscale, 9.8472.15 (6–14); TFEQ disinhibition subscale, 8.9172.08 (3–13); TFEQ hunger subscale, 7.4172.05 (2–11); TFEQ flexible control subscale, 0.5270.63 (0–2); TFEQ rigid control subscale, 2.1671.63 (0–6); DEBQ restrained eating subscale, 23.5078.02 (9–38); EDE-Q restraint subscale, 1.4071.21 (0–4.2); EDE-Q eating concern subscale, 0.677 0.83 (0–3.4); EDE-Q shape concern subscale, 2.0371.39 (0–4.5); and EDE-Q weight concern subscale, 1.4771.13 (0–4.2). Only two of 38 (5.3%) participants reported being on a diet to lose weight during either of the test meal sessions. On the EDE-Q, two individuals reported self-inducing vomiting in the month (28 days) prior to completing the self-report assessments (1 episode each), and one of the individuals had also used laxatives once in the prior month to control her shape or weight. Sixteen participants reported objective bulimic episodes (OBEs), or consuming what other people would regard as an unusually large amount of food (given the circumstances) with a sense of having lost control over eating (at the time the eating occurred), in the month prior to completing the self-report assessments. Five participants reported 1 OBE, three reported 2 OBEs, one reported 3 OBEs, two reported 4 OBEs, two reported 5 OBEs, one reported 6 OBEs, one reported 14 OBEs, and one reported 15 OBEs on the EDE-Q. Two participants endorsed both binge eating and purging (one participant with 1 OBE and 1 episode of vomiting, one participant with 6 OBEs and 1 episode each of vomiting and laxative abuse). Comparison of the low- and high-calorie instruction test sessions Participants consumed an average of 402.677166.87 g (range 105.0–819.0) on the low-calorie instruction day, and an average of 382.127155.74 g (range 107.0–634.0) on the high-calorie instruction day, t(37) ¼ 0.97, p ¼ 0.3, d ¼ 0.13. Participants’ estimates of caloric consumption on the high-calorie instruction day were significantly higher than on the low-calorie instruction day (low-calorie instruction day: 270.447160.04 calories, high-calorie instruction day: 387.707206.54 calories, t(37) ¼ 3.47, po0.001, d ¼ 0.61); however, estimates of the number of ounces of shake consumed were not different between the high- and low-calorie instruction sessions (low-calorie instruction day: 10.1276.84 ounces, high-calorie instruction day: 10.6277.08 ounces, t(37) ¼ 0.49, p ¼ 0.6, d ¼ 0.07). The means, standard deviations, and results of paired t-tests of the pre-meal and post-meal VAS ratings on the low- and high-calorie instruction days for the participants
are presented in Table 1. On the high-calorie instruction day, participants rated the shake as significantly more sweet, t(37) ¼ 2.49, p ¼ 0.02, d ¼ 0.21, and rich, t(37) ¼ 5.42, po0.001, d ¼ 0.65, than on the lowcalorie instruction day. There were no significant differences in pre- or post-meal VAS ratings of hunger, fullness, sickness, anxiety, depression, or desire to eat, or in postmeal VAS ratings of bitterness or novelty. In addition, there were no significant differences observed for total intake on the low-and high-calorie instruction days between ethnic groups or between shake flavors. Correlations between dietary restraint and intake on the high-calorie instruction day There was a significant correlation between the total intake on the low- and high-calorie instruction days (r ¼ 0.64, po0.001). Correlation coefficients were calculated between total intake on the high-calorie instruction day and measures of dietary restraint (DIS, TFEQ restraint subscale, TFEQ disinhibition subscale, TFEQ hunger subscale, TFEQ flexible control and rigid control subscales (Westenhoefer, 1991), DEBQ restrained eating subscale, and EDE-Q restraint subscale). Other correlations for the restraint scores and total intake on the high-calorie instruction day ranged from 0.005 to 0.26, but none of the correlations was statistically significant. Secondary analyses Median split based on dietary restraint measures Similar to previous studies of the restraint hypothesis, we analyzed the data using a median split. The a values for this Table 1 Means, standard deviations (SD), and results of paired t-tests for the preand post-meal Visual Analog Scale (VAS) line ratings of participants (n ¼ 38) on the low- and high-calorie instruction days
Hunger Fullness Sickness Anxiety Depression Desire to eat Like what eating Pleasant Sweetness Richnessa Bitternessa Novelty
Low-calorie instruction day
High-calorie instruction day
Pre-meal
Post-meal
Pre-meal
Post-meal
M
SD
M
SD
M
SD
M
SD
8.6 3.7 2.3 3.4 2.0 9.3
3.1 3.1 2.9 3.2 3.8 3.3
2.3 11.2 3.5 2.3 1.2 2.5 10.5 10.6 10.2 8.6 1.3 5.8
2.7 2.7 3.7 2.7 2.1 2.7 3.1 2.8 2.9 3.6 2.1 3.6
8.8 3.4 2.3 2.9 1.9 8.7
2.7 2.6 3.3 2.9 2.6 3.7
2.2 11.0 3.5 1.9 1.7 2.1 9.7 9.8 11.4 11.6 1.2 5.9
2.5 2.7 3.3 2.6 2.7 2.5 3.7 3.9 3.2 2.5 2.4 3.8
Note: maximum VAS rating was 15.0 cm. a Post-meal paired t-tests between low- and high-calorie instruction days, po0.05.
ARTICLE IN PRESS R. Sysko et al. / Appetite 49 (2007) 30–37
analysis were divided by 14 in accord with the Bonferroni correction for multiple comparisons (p ¼ 0.05/14, or pp0.004). With this a level, none of the comparisons of high- and low-scoring individuals on the restraint measures (DIS, TFEQ disinhibition, TFEQ restraint subscale, TFEQ hunger subscale, TFEQ rigid control subscale, TFEQ flexible control subscale, DEBQ restrained eating subscale, or EDE-Q restraint subscale) were significant for total meal intake. Correlations between dietary restraint and intake on the low-calorie instruction day The correlations between intake on the low-calorie instruction day and the TFEQ rigid control subscale (r ¼ 0.36, p ¼ 0.03) and the DEBQ restrained eating subscale (r ¼ 0.33, p ¼ 0.04) were statistically significant. However, when one outlier with a TFEQ rigid control score of 6 was removed from the analysis, the correlation was no longer significant (r ¼ 0.20, p ¼ 0.2). While there was a trend toward significance of the correlation between intake on the low-calorie instruction day and the DIS (r ¼ 0.30, p ¼ 0.07), none of the other correlations between measures of dietary restraint and intake on the low-calorie instruction day was significant (range of r ¼ 0.04 to 0.28). Discussion The purpose of the current study was twofold: first, to examine the effect of a cognitive manipulation on total shake consumption during two experimental meal sessions, and second, to evaluate the relationship between selfreported measures of dietary restraint and intake during the meals. The data indicated that the cognitive manipulation was successful, with 86% (n ¼ 38 of 44) participants meeting the criteria for the manipulation check. In addition, on the high-calorie instruction day, participants rated the shake as significantly more sweet by VAS and reported consuming significantly more calories during the session. However, despite the success of the cognitive manipulation and the differences in ratings of sweetness and calories consumed on the high-calorie instruction day, the instructions had little effect on total food intake, as no significant differences in shake consumption were observed between the low- and high-calorie instruction days. We, therefore, did not find support for the hypothesis that participants would consume less when they believed they were consuming a shake made with high-calorie ingredients. In addition, since none of the correlations between the measures of dietary restraint and intake were significant on the high-calorie instruction day, we found no support for our second main hypothesis. When participants with higher scores on measures of dietary restraint were presented with a food believed to be higher in calories, their self-reported attempts at restraint, dieting, or weight loss did not translate into reduced food intake.
35
As many studies of dietary restraint and eating behavior divide participants into low- and high-scoring groups using a median split, we conducted a secondary analysis to determine whether similar results would also be obtained in the current study. No significant relationships were observed between intake and dietary restraint for highscoring participants. Thus, when analyzing the self-report measures of dietary restraint as continuous variables (described above) or using a median split, the results were the same. Another secondary analysis, comparing measures of dietary restraint and intake on the low-calorie instruction day, found that scores on the TFEQ rigid control subscale and the DEBQ restraint subscale were significantly correlated with intake, with higher scores associated with greater consumption. However, this correlation was no longer significant when an outlier was removed from the analysis. As both subscales include items measuring efforts at behavioral restraint to avoid weight gain, this finding suggests that high-restrained participants on the TFEQ and DEBQ exhibited disinhibited eating when they believed that the shake included low-calorie ingredients. The disinhibited eating may have occurred because participants thought they could drink more of the low-calorie shake without consuming a large number of calories or gaining weight. Thus, with the recent proliferation of foods marketed as ‘‘low-calorie,’’ restrained consumers could be triggered to overeat when eating food believed to be low calorie, which could result in weight gain, but the finding needs to be replicated. Taken together, the two secondary analyses found only small effects. Given the size of the current study, the analyses should have been adequately powered to detect a clinically meaningful difference, but the interpretation of these analyses do not support a strong relationship between measures of dietary restraint and eating behavior on either the low- or high-calorie instruction days. The restraint hypothesis posits that restrained eaters are triggered to eat after breaking a dietary rule, or consuming a forbidden food. This theory was not specifically tested in this study. The current study provided participants with a lunch meal instead of using a pre-load paradigm, or the consumption of a food, like a shake, prior to an experimental meal or taste test. A direct comparison of the results of the current study and studies testing the restraint hypothesis is therefore not possible. However, the findings of the current study and two previous studies (Polivy, 1976; Spencer & Fremouw, 1979) can be compared. Although these studies were designed to test the restraint hypothesis, their focus on cognition provides potential explanations for the discrepancies between studies testing the theory of Polivy, Herman, and colleagues and other eating behavior studies. Both Polivy (1976) and Spencer and Fremouw (1979) demonstrated that restrained subjects increased their tastetest intake when they believed they had consumed a highcalorie preload. Therefore, among restrained individuals, the effect of dietary restraint on eating behavior may be
ARTICLE IN PRESS 36
R. Sysko et al. / Appetite 49 (2007) 30–37
primarily related to cognitive factors (Polivy, 1976; Spencer & Fremouw, 1979). The current study also supports the importance of a cognitive control over eating, as a relationship was observed between dietary restraint and total intake when participants received instructions intended to produce expectations about the caloric density of a shake, although only on the low-calorie instruction day. The findings of these studies support Stice et al.’s (2006) suggestion that the use of a cognitive manipulation may lead to a more pronounced relationship between dietary restraint and eating behavior. Other cognitive factors may be important in testing the restraint hypothesis, where participants are asked to consume a pre-load and complete a subsequent ‘‘taste test.’’ This type of food consumption is not typical of naturalistic eating, as it requires the breaking of a dietary rule, and focuses participants’ attention on the pre-load and taste test, with foods (e.g., shake, ice cream, cookies) and a meal structure that are not usual for meals outside of the laboratory. Although studies testing the restraint hypothesis usually observe disinhibited eating among high-restrained participants after consuming a pre-load, these experiments involve procedures that are relatively artificial. Therefore, the relationship between dietary restraint and eating behavior may only be apparent under special circumstances that are elicited in a laboratory meal paradigm under atypical eating conditions. As such, it is important to examine the role the experimental design in influencing the relationship between dietary restraint and eating behavior. The ability to predict eating outside of the laboratory from scores on measures of dietary restraint is an important issue. In three studies, using more naturalistic laboratory paradigms (Martin et al., 2005; Stice et al., 2004; Sysko et al., 2005), no relationship was observed between dietary restraint and eating behavior. In addition, a study assessing reported energy intake among restrained and unrestrained participants (Bathalon et al., 2000) found no effect of dietary restraint and total energy expenditure over an 18-day period. These studies indicate that scores on self-report measures of dietary restraint do not measure acute (Martin et al., 2005; Stice et al., 2004; Sysko et al., 2005) or chronic caloric restriction (Bathalon et al., 2000), and may be assessing the perception of efforts to lose weight as opposed to true weight loss dieting, or the behavioral restriction of food intake (Stice et al., 2006). Self-report measures of dietary restraint may not provide information about behavior during laboratory meals that lack a pre-load, or in studies without instructions designed to draw attention to a specific aspect of the experimental manipulation. Thus, there is only limited evidence that selfreport measures of dietary restraint are valid assessments of food restriction. The clinical significance of the findings from the current study is not clear. Most participants included in the study did not report clinically significant eating disturbances. In addition, only six participants had a BMI in the overweight
or obese range (BMIX25 kg/m2), and only two participants reported that they were dieting to lose weight at either test meal session. In addition, prior to participating in the study, all interested undergraduates were told that the experiment involved consuming a shake made with either high- and low-calorie ingredients. As a result, the sample likely included individuals who had fewer concerns with eating and weight than participants who would not volunteer for a study of eating behavior, as evidenced by the limited range of scores on self-report measures of dietary restraint among participants in the study. Therefore, the results from this study cannot be generalized to other populations, as the sample was primarily normalweight undergraduate women without high levels of dietary restraint or significant eating disorder symptoms. There are also several limitations to the design of the current study. Similar to other studies of eating behavior in a laboratory, it is not known whether the findings of the current study generalize to a naturalistic eating environment. The use of the single-item test meal (ice-cream shake) for lunch, the presentation of the test meal in a large opaque container with the amount and type of food unknown to the participant, is also inconsistent with a typical lunch. Future research should include additional experimental studies to help further elucidate the relationship between dietary restraint and eating behavior, such as a study with a within-subjects design investigating the effect of meal instructions on test meal intake. Participants could be asked in one session to complete a taste test like those used to test the restraint hypothesis (e.g., pre-load shake followed by tasting cookies) and in another to consume a more normal meal (e.g., breakfast as in the Stice et al. (2004) study). In this way, the explanation proposed by the current study for the discrepancy between the studies testing restraint theory and more recent studies of dietary restraint, and the importance of the instructions during the experiment, could be tested. In summary, this study focused on the relationship between a cognitive manipulation, self-reported dietary restraint, and total intake during two experimental meal sessions. Although the cognitive manipulation was successful in producing an expectation about the caloric content of the food used in the study, there was no significant difference in intake between the low- and high-calorie instruction days. There were also few significant relationships between self-report measures of dietary restraint and intake during either test meal. The effects of dietary restraint on food intake in this study and others (e.g., Bathalon et al., 2000; Martin et al., 2005; Stice et al., 2004; Sysko et al., 2005) are not robust, and are not supportive of the restraint hypothesis (e.g., Ruderman, 1986). Measures of dietary restraint may therefore be evaluating another construct (e.g., cognitive restraint over eating) instead of behavioral restraint. To resolve the discrepancies between the studies of dietary restraint and eating behavior, and to better understand the construct assessed by self-report dietary restraint measures, additional studies are needed.
ARTICLE IN PRESS R. Sysko et al. / Appetite 49 (2007) 30–37
References Bathalon, G. P., Tucker, K. L., Hays, N. P., Vinken, A. G., Greenberg, A. S., McCrory, M. A., et al. (2000). Psychological measures of eating behavior and the accuracy of 3 common dietary assessment methods in healthy postmenopausal women. American Journal of Clinical Nutrition, 71, 739–745. Cooper, P. J., Taylor, M. J., Cooper, Z., & Fairburn, C. G. (1987). The development and validation of the Body Shape Questionnaire. International Journal of Eating Disorders, 6, 485–494. Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders: Interview or self-report questionnaire? International Journal of Eating Disorders, 16, 363–370. Fairburn, C. G., Marcus, M. D., & Wilson, G. T. (1993). Cognitivebehavioral therapy for binge eating and bulimia nervosa: A comprehensive treatment manual. In C. G. Fairburn, & G. T. Wilson (Eds.), Binge eating: Nature, assessment and treatment (pp. 361–404). New York: Guilford Press. Herman, C., & Polivy, J. (1975). Anxiety, restraint, and eating. Journal of Abnormal Psychology, 84, 666–672. Lowe, M. R. (1993). The effects of dieting on eating behavior: A three factor model. Psychological Bulletin, 114, 100–121. Martin, C. K., Williamson, D. A., Geiselman, P. J., Walden, H., Smeets, M., Morales, S., et al. (2005). Consistency of food intake over four eating sessions in the laboratory. Eating Behaviors, 6, 365–372. Murphy, K. R., & Myors, B. (2004). Statistical power analysis: A simple and general model for traditional and modern hypothesis tests. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers. Polivy, J. (1976). Perception of calories and regulation of intake in restrained and unrestrained subjects. Addictive Behaviors, 1, 237–243. Polivy, J., & Herman, C. P. (1985). Dieting and bingeing: A causal analysis. American Psychologist, 40, 193–201. Ruderman, A. J. (1986). Dietary restraint: A theoretical and empirical review. Psychological Bulletin, 99, 247–262.
37
Spencer, J. A., & Fremouw, W. J. (1979). Binge eating as a function of restraint and weight classification. Journal of Abnormal Psychology, 88, 262–267. Stice, E. (1998). Prospective relation of dieting behaviors to weight change in a community sample of adolescents. Behavior Therapy, 29, 277–291. Stice, E., & Agras, W. S. (1998). Predicting onset and cessation of bulimic behaviors during adolescence: A longitudinal grouping analyses. Behavior Therapy, 29, 257–276. Stice, E., Fisher, M., & Lowe, M. R. (2004). Are dietary restraint scales valid measures of acute dietary restriction? Unobtrusive observational data suggest not. Psychological Assessment, 16, 51–59. Stice, E., Presnell, K., Lowe, M. R., & Burton, E. (2006). Validity of dietary restraint scales: Reply to van Strien et al. Psychological Assessment, 18, 95–99. 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. Stunkard, A. J., & Messick, S. (1988). Eating inventory manual. San Antonio: The Psychological Corporation. Sysko, R., Walsh, B. T., Schebendach, J., & Wilson, G. T. (2005). Eating behavior among women with anorexia nervosa. American Journal of Clinical Nutrition, 82, 296–301. van Strien, T., Frijters, J. E. R., Bergers, G. P. A., & Defares, P. B. (1986). The Dutch Eating Behavior Questionnaire (DEBQ) for assessment of restrained, emotional, and external eating behavior. International Journal of Eating Disorders, 5, 295–315. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Westenhoefer, J. (1991). Dietary restraint and disinhibition: Is restraint a homogeneous construct? Appetite, 16, 45–55.
Appetite 49 (2007) 30–37 www.elsevier.com/locate/appet
Research report
Expectancies, dietary restraint, and test meal intake among undergraduate women Robyn Syskoa,,1, B. Timothy Walshb, G. Terence Wilsona a
Department of Psychology, Rutgers, The State University of New Jersey, Eating Disorders Clinic, 41 C Gordon Road, Piscataway, NJ 08854, USA Department of Psychiatry, College of Physicians and Surgeons of Columbia University, New York State Psychiatric Institute, New York, NY, USA
b
Received 24 July 2006; received in revised form 1 November 2006; accepted 1 November 2006
Abstract This study investigated the relationship between self-reported dietary restraint and expectancies about caloric content on test meal consumption among undergraduate women. Participants completed two test meal sessions during which they were asked to consume as much milkshake from a covered opaque container as they wished. In one session, participants were instructed that the milkshake was made with high-calorie ingredients, and in the other that the milkshake was made with low-calorie ingredients. The milkshakes in both sessions were actually made with the same ingredients. Participants’ mean consumption was less on the low-calorie instruction day (402 g) than on the high-calorie instruction day (382 g), but the difference was not statistically significant. In addition, few significant relationships were observed between dietary restraint measures and total intake on either the low- or high-calorie instruction days. Thus, this study supports a growing body of literature indicating that scores on measures of dietary restraint are not related to the actual restriction of food intake. r 2006 Elsevier Ltd. All rights reserved. Keywords: Dietary restraint; Behavioral restriction; Eating behavior; Restraint theory
Introduction The term ‘‘dietary restraint’’ has been used to refer to a range of attitudes and behaviors, including food avoidance, ongoing attempts to restrict caloric intake to lose weight that are associated with unsuccessful dieting (Lowe, 1993), and a cognitive control over eating, which makes dieters vulnerable to episodes of uncontrolled eating when that control is disrupted (Polivy & Herman, 1985). Higher scores on measures of dietary restraint have been shown to predict the future development of binge eating (e.g., Stice & Agras, 1998), and dietary restraint is often suggested as a risk factor for the development of eating disorders (e.g., Polivy & Herman, 1985). A number of studies have evaluated the relationship between eating behavior, typically in an experimental Corresponding author.
E-mail address: [email protected] (R. Sysko). Current address: VA Connecticut Healthcare, 116B/Psychology, West Haven Campus, 950 Campbell Avenue, West Haven, CT 06516, USA. 1
0195-6663/$ - see front matter r 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.appet.2006.11.002
paradigm, and self-report measures of dietary restraint. Much of this research has examined a theory described by Polivy, Herman, and colleagues. This theory, often termed the restraint hypothesis, postulates that restrained eaters, those who exert cognitive control over food consumption and score highly on the Restraint Scale (Herman & Polivy, 1975), will overeat after being exposed to a disinhibitor, such as breaking a dietary rule or consuming a forbidden food. The effect of disinhibitors like pre-loads (i.e., the consumption of a food prior to an experimental meal or taste test) on restrained eaters has been supported in numerous studies (for a review see Ruderman, 1986 or Lowe, 1993). These studies typically classify participants based on the median score of the Restraint Scale into low- and highrestrained groups. Participants consume a pre-load (e.g., milkshake) and subsequently complete a taste test (e.g., ice cream, cookies; Ruderman, 1986). The amount of food consumed during the taste test portion of the experiment is measured to determine the effect of the disinhibitor (the pre-load). High-restrained participants tend to eat
ARTICLE IN PRESS R. Sysko et al. / Appetite 49 (2007) 30–37
significantly more food during the taste test after consuming a high-calorie pre-load than low-restrained participants, suggesting that after breaking a dietary rule (i.e., consuming a high-calorie milkshake), the eating of highly restrained participants is disinhibited (Ruderman, 1986). Similar patterns of behavior are believed to play a role in the occurrence of binge eating (e.g. ‘‘I’ve blown my diet;’’ Fairburn, Marcus, & Wilson, 1993). Recent research has questioned whether the model described by Polivy, Herman, and colleagues accurately describes the relationship between dietary restraint and eating behavior. In a series of four experiments, Stice, Fisher, and Lowe (2004) examined whether commonly used self-report measures of dietary restraint were negatively correlated with food consumption. The experiments were conducted in diverse settings, including both laboratory and naturalistic eating environments, and participants included undergraduate women, women with bulimia nervosa, women with binge eating disorder, women without an eating disorder, and female patrons of a fast food restaurant (Stice et al., 2004). In all four studies, no relationship was observed between the measures of dietary restraint and caloric consumption, with one exception. Significant negative correlations were observed between the Dietary Intent Scale (DIS; Stice, 1998) and caloric intake and fat-gram intake for participants consuming a meal in a fast food restaurant. However, the correlation, while statistically significant, was ‘‘rather meager’’ (r ¼ 0.24, po0.05), and, as the authors noted, ‘‘there is still much room for improvement’’ (Stice et al., 2004, p. 55). Stice et al. (2004) concluded that the measures of dietary restraint utilized in the study were not valid assessments of true short-term food restriction. Martin et al. (2005) used a similar experimental design to examine the relationship between dietary restraint and meal consumption during a series of four meals. Individuals without an eating disorder were asked to consume sandwiches instead of their usual lunch on four different days. Across the four sessions, there was no effect of dietary restraint, as measured by the Three Factor Eating Questionnaire (TFEQ; Stunkard & Messick, 1988), on total intake during lunch (Martin et al., 2005). In addition, a recent study of eating behavior among patients with anorexia nervosa and individuals without an eating disorder also found that several widely used scales of dietary restraint did not correlate with total intake during a laboratory lunch meal for either participant group (Sysko, Walsh, Schebendach, & Wilson, 2005). Thus, while restrained individuals appear to increase food consumption in response to disinhibition in experiments involving a pre-load and a taste test (e.g., Ruderman, 1986), there does not appear to be a relationship between dietary restraint and food consumption during a standardized breakfast meal, meals at a university cafeteria (Stice et al., 2004), or lunch meals (Martin et al., 2005; Sysko et al., 2005). Some of the discrepancies observed between studies may be related to differences in experi-
31
mental design, which may have influenced the subsequent relationship between dietary restraint and eating behavior. For example, the expectations of participants during an eating behavior study may be particularly important in influencing total intake. Two previous studies examining the restraint hypothesis (Polivy, 1976; Spencer & Fremouw, 1979) illustrate the potential impact of cognitive factors in the study of dietary restraint. Polivy (1976) manipulated the expectations of restrained and unrestrained male undergraduates by instructing participants that they were receiving a ‘‘rich, high-calorie, gourmet, pudding-type dessert,’’ or a ‘‘low-calorie pudding’’ (Polivy, 1976, p. 239). After consuming the pudding, participants were presented with sandwich quarters, and the number of sandwiches eaten was measured. Expectations about the pudding were manipulated using both the instructions and the true caloric content of the food, including the following four conditions: (1) true high calorie (high-calorie pudding and high-calorie instructions); (2) true low calorie (low-calorie pudding and lowcalorie instructions); (3) fake high calorie (low-calorie pudding, high-calorie instructions); and (4) fake low calorie (high-calorie pudding, low-calorie instructions). Polivy (1976) found an interaction between dietary restraint and perceived calories (what participants believed they ate), suggesting that cognitive factors significantly affected meal consumption; however, the overall effect of the manipulation was not very large. Spencer and Fremouw (1979) conducted a similar study, but the cognitive manipulation was limited to the perception of whether the food consumed was high or low calorie. High- and low-restrained female undergraduates were given a liquid pre-load and told that the pre-load was either high or low calorie. In both the ‘‘told high-calorie’’ and ‘‘told low-calorie’’ conditions, the shake had an equivalent caloric density. After consuming the shake pre-load, participants completed a taste test with three ice creams. High-restrained participants consumed significantly more of the ice cream after the told high-calorie condition than the told low-calorie condition, indicating that the cognitive manipulation was successful. For restrained eaters, cognitive expectations about the caloric content of a pre-load produced disinhibited eating in the laboratory, regardless of the true caloric content of the pre-load. The current study was designed to examine the effect of expectations about the caloric content of a food on subsequent intake. Unlike both the Polivy (1976) and Spencer and Fremouw (1979) studies, this experiment did not use a pre-load paradigm. Studies utilizing a pre-load paradigm investigate whether the violation of a dietary rule (e.g., consuming a pre-load) by restrained eaters overrides their self-imposed dietary restriction during a subsequent taste test. The current study instead measured the effect of the cognitive manipulation on the consumption of a meal (lunch), which more closely approximates eating outside of the laboratory. The second aim of the current study was to
ARTICLE IN PRESS 32
R. Sysko et al. / Appetite 49 (2007) 30–37
examine the relationship between self-report measures of dietary restraint and total intake during the test meals, which would determine whether measures of dietary restraint are in fact predictive of behavioral restraint. The experimental design of the current study also allowed for an evaluation of a hypothesis suggested by Stice, Presnell, Lowe, and Burton (2006) that the use of a cognitive manipulation can lead to a more pronounced relationship between dietary restraint and eating behavior. We hypothesized that in the current study test meal consumption would be affected by a cognitive manipulation of participants’ expectations about the caloric content of the food, and that the effect of the cognitive manipulation would vary based on participants’ self-reported amount of dietary restraint. Specifically, the two main hypotheses were that (1) all participants would eat less when they believed they were consuming a high-calorie shake, and (2) that there would be a relationship between higher scores on measures of dietary restraint and lower amounts of total intake on the high-calorie instruction day. Methods Participants Forty-four female undergraduates participated in two test meal sessions in exchange for course credit. All participants provided informed consent. This study was approved by the Institutional Review Board of Rutgers University. Procedures Participants indicated their interest in the experiment on the psychology department’s website, and subsequently received an email with an attachment containing the consent form for the study and research questionnaires (described below). Eligibility was determined after the consent and questionnaires were completed. Six of the 50 students who signed up for the study on the website decided not to participate in the experiment by not returning a completed consent and questionnaires. The two test meal sessions occurred on non-consecutive days, and participants were told to fast (consume nothing but water) for the 4 h prior to their sessions. Before consuming the test meal, participants completed 15 cm Visual Analog Scales (VAS) assessing the constructs of fullness, hunger, sickness, anxiety, depression, and desire to eat, with the anchors of ‘‘not at all’’ and ‘‘extremely.’’ Participants also recorded when they last ate, how many hours of sleep they had in the previous evening, and whether they were on a diet to lose weight, as Lowe (1993) suggested that this single item may be a more valid measure of dietary restraint than other measures of restraint. The participants listened to one of two sets of taperecorded instructions. Before both test meals, participants were told to eat as much of the shake as they liked and that
the shake was their lunch for the day. In one meal session, participants were instructed that they were consuming a milkshake made with high-calorie ingredients (i.e., premium ice cream and whole milk) and for the other meal session, participants were instructed that they were consuming a shake made with low-calorie ingredients (i.e., fat-free frozen yogurt and skim milk). The shakes in both sessions were made with the same ingredients, which included all the low- and high-calorie ingredients mentioned in the instructions. Thus, the instructions were intended to elicit an expectation about the caloric content of the shakes; however, there was not a true difference in caloric content. Each participant received two shake flavors (e.g., chocolate and vanilla) in an attempt to conceal the similarity of the shakes. The high- and lowcalorie instructions, as well as the flavors of the shake, were counterbalanced to ensure that there would not be unequal pairings of each flavor of shake with the high- or lowcalorie instructions. Prior to the start of the study, three shake recipes were pilot tested with the staff of the Eating Disorders Research Unit (EDRU) of the New York State Psychiatric Institute/Columbia University Medical Center. The VAS ratings of the EDRU staff were reviewed, and we chose a shake rated as being moderately ‘‘rich,’’ such that it would be possible that the shake could be made with highcalorie ingredients. Participants were provided with approximately 1075 g of chocolate, vanilla, or strawberry shake (1.12 kcal/g, or approximately 1205.5 kcal; range of 988–1190 g or 1107.9–1334.5 kcal). When participants finished consuming the shake, they rang a doorbell, at which point they completed a final VAS assessment of hunger, fullness, sickness, anxiety, depression, desire to eat, liking of the shake, pleasantness of the shake, and taste perceptions (sweet, rich, bitter, novelty), and estimated the amount of food eaten during the test meal session (in calories and ounces). A pictorial example of volume was provided to assist in this estimation. Some ratings after the second test meal measured the success of the intervention in altering expectations about the caloric content of the shake (see the description of the manipulation check below). The shake was weighed before and after the meal to determine the total amount consumed during the test meal session. After the second test meal, the participant’s height and weight was measured. Manipulation check A manipulation check was used to determine whether participants believed the cognitive manipulation regarding the caloric density of the shake. To be included in the analyses, participants were required to meet two of the following three conditions: (1) participants had to rate the shake on the high-calorie instruction day to be richer than the shake on the low-calorie instruction day on the post-meal VAS ratings for richness (e.g., VAS measurement of 10.4 cm on the high-calorie instruction day and
ARTICLE IN PRESS R. Sysko et al. / Appetite 49 (2007) 30–37
6.8 cm on the low-calorie instruction day); and after the second meal both (2) correctly indicating the instructions played at both test meals (i.e., circling the correct response to both: ‘‘Today’s meal was made with high/low-calorie ingredients’’ and ‘‘Last session, the shake was made with high/low-calorie ingredients’’), and (3) when asked to compare the shakes consumed in both sessions, participants needed to indicate that the high-calorie shake was richer than the low-calorie shake (e.g., compared to last session, the shake today was: much less rich/extremely rich on a scale of 0–7). Measures The following self-report measures were given to participants to assess dietary restraint, concerns about shape and weight, and positive and negative affects. Self-report measures of dietary restraint Four measures (DIS; TFEQ; Dutch Eating Behavior Questionnaire, DEBQ; and Eating Disorder Examination Questionnaire, EDE-Q) were used to assess dietary restraint. The DIS (Stice, 1998) is a nine-item scale assessing dietary restraint and measures attempts to control weight or prevent weight gain. The TFEQ (Stunkard & Messick, 1985) assesses behaviors to produce weight loss, and includes three subscales: disinhibition, restraint, and hunger. Westenhoefer (1991) derived two additional subscales (flexible and rigid control) from a factor analysis of the TFEQ. The DEBQ (van Strien, Frijters, Bergers, & Defares, 1986) is a popular method of assessing dietary restraint, focusing on weight loss and maintenance, with three subscales (restrained eating, emotional eating, and external eating). Finally, the EDEQ (Fairburn & Beglin, 1994) is a self-report questionnaire that measures dietary restraint, shape and weight concern, and eating disordered behaviors. Assessments of shape and weight concerns and affect The Body Shape Questionnaire (BSQ; Cooper, Taylor, Cooper, & Fairburn, 1987) and the Positive and Negative Affect Scales (PANAS; Watson, Clark, & Tellegen, 1988) were interspersed in the computerized assessment package with the four measures of dietary restraint in an attempt to distract participants from our focus on dietary restraint. As these measures were not administered to test specific hypotheses, the data are not presented here.
33
(calories and ounces) on the high- and low-calorie instruction days were compared using paired t-tests. Effect sizes (d) were calculated as the mean difference between the test meal days (e.g., low- and high-calorie instruction days) for a given variable divided by the mean standard deviation of that variable over the test meal sessions. To test the second main hypothesis, and to analyze all of the selfreported measures of dietary restraint (DIS, TFEQ, DEBQ, and EDE-Q) as continuous variables, Pearson correlation coefficients (r) were calculated between the restraint measures and total intake on the high-calorie instruction day. The TFEQ scoring (Stunkard & Messick, 1985) was used to derive three subscales (restraint, disinhibition, and hunger) and an additional two subscales (rigid control, flexible control) described by Westenhoefer (1991). Secondary data analysis Similar to Polivy (1976) and Spencer and Fremouw (1979), we divided participants into high- and low-restraint groups using a median split on each of the measures of dietary restraint to allow for comparisons with previous research on the restraint hypothesis. One-way ANOVAs were calculated between the dependent variables of intake on the low- or high-calorie instruction day, or the change in intake between the low- and high-calorie instruction days, and the independent variable of the median split (low-scoring vs. highscoring) for each of the measures of dietary restraint (DIS, TFEQ disinhibition, TFEQ restraint subscale, TFEQ hunger subscale, TFEQ flexible control subscale, TFEQ rigid control subscale, DEBQ restrained eating subscale, or EDE-Q restraint subscale). Finally, Pearson correlation coefficients were calculated between each of the self-report measures of dietary restraint and total intake on the low-calorie instruction day to determine whether there was a relationship between degree of dietary restraint and intake when participants believed the shake was low calorie. A power analysis conducted before recruitment began indicated that using a within-subjects design and a total sample size of 33, an a level of 0.05, and assuming a population standard distribution of 1.0, there was an 80% chance of finding a medium effect (Cohen’s d ¼ 0.5), suggesting adequate power to detect a clinically meaningful difference (Murphy & Myors, 2004). All statistical calculations were performed using SPSS for Windows, version 11.5. All means are reported7standard deviations. Results
Statistical analysis Demographic characteristics Primary statistical analyses Means and standard deviations were calculated for the participants’ demographic characteristics. To test the first main hypothesis, average amounts consumed on the highand low-calorie instruction days were compared using paired t-tests. Similarly, pre- and post-meal VAS ratings and the estimates of intake during the test meal sessions
Of the 44 participants who completed both test meal sessions, 38 (86.3%) met the criteria for the manipulation check and are included in the analyses described below. The meals were, on average, 13.4 days apart (range of 2–38 days). The mean age of the participants was 19 years and the mean body mass index (BMI; kg/m2) was 22.4973.44
ARTICLE IN PRESS 34
R. Sysko et al. / Appetite 49 (2007) 30–37
(range 18.43–37.42). Nineteen participants (50.0%) were Caucasian, 13 (34.2%) were Asian, 3 (7.9%) were Hispanic, 2 (5.3%) were African-American, and 1 (2.6%) was Indo-Caribbean and Afro-Caribbean. The mean scores7standard deviations and ranges, in parentheses, for the self-report measures, including restraint, shape and weight concerns, and other psychological symptoms are as follows: DIS, 19.3076.06 (9–31); TFEQ restraint subscale, 9.8472.15 (6–14); TFEQ disinhibition subscale, 8.9172.08 (3–13); TFEQ hunger subscale, 7.4172.05 (2–11); TFEQ flexible control subscale, 0.5270.63 (0–2); TFEQ rigid control subscale, 2.1671.63 (0–6); DEBQ restrained eating subscale, 23.5078.02 (9–38); EDE-Q restraint subscale, 1.4071.21 (0–4.2); EDE-Q eating concern subscale, 0.677 0.83 (0–3.4); EDE-Q shape concern subscale, 2.0371.39 (0–4.5); and EDE-Q weight concern subscale, 1.4771.13 (0–4.2). Only two of 38 (5.3%) participants reported being on a diet to lose weight during either of the test meal sessions. On the EDE-Q, two individuals reported self-inducing vomiting in the month (28 days) prior to completing the self-report assessments (1 episode each), and one of the individuals had also used laxatives once in the prior month to control her shape or weight. Sixteen participants reported objective bulimic episodes (OBEs), or consuming what other people would regard as an unusually large amount of food (given the circumstances) with a sense of having lost control over eating (at the time the eating occurred), in the month prior to completing the self-report assessments. Five participants reported 1 OBE, three reported 2 OBEs, one reported 3 OBEs, two reported 4 OBEs, two reported 5 OBEs, one reported 6 OBEs, one reported 14 OBEs, and one reported 15 OBEs on the EDE-Q. Two participants endorsed both binge eating and purging (one participant with 1 OBE and 1 episode of vomiting, one participant with 6 OBEs and 1 episode each of vomiting and laxative abuse). Comparison of the low- and high-calorie instruction test sessions Participants consumed an average of 402.677166.87 g (range 105.0–819.0) on the low-calorie instruction day, and an average of 382.127155.74 g (range 107.0–634.0) on the high-calorie instruction day, t(37) ¼ 0.97, p ¼ 0.3, d ¼ 0.13. Participants’ estimates of caloric consumption on the high-calorie instruction day were significantly higher than on the low-calorie instruction day (low-calorie instruction day: 270.447160.04 calories, high-calorie instruction day: 387.707206.54 calories, t(37) ¼ 3.47, po0.001, d ¼ 0.61); however, estimates of the number of ounces of shake consumed were not different between the high- and low-calorie instruction sessions (low-calorie instruction day: 10.1276.84 ounces, high-calorie instruction day: 10.6277.08 ounces, t(37) ¼ 0.49, p ¼ 0.6, d ¼ 0.07). The means, standard deviations, and results of paired t-tests of the pre-meal and post-meal VAS ratings on the low- and high-calorie instruction days for the participants
are presented in Table 1. On the high-calorie instruction day, participants rated the shake as significantly more sweet, t(37) ¼ 2.49, p ¼ 0.02, d ¼ 0.21, and rich, t(37) ¼ 5.42, po0.001, d ¼ 0.65, than on the lowcalorie instruction day. There were no significant differences in pre- or post-meal VAS ratings of hunger, fullness, sickness, anxiety, depression, or desire to eat, or in postmeal VAS ratings of bitterness or novelty. In addition, there were no significant differences observed for total intake on the low-and high-calorie instruction days between ethnic groups or between shake flavors. Correlations between dietary restraint and intake on the high-calorie instruction day There was a significant correlation between the total intake on the low- and high-calorie instruction days (r ¼ 0.64, po0.001). Correlation coefficients were calculated between total intake on the high-calorie instruction day and measures of dietary restraint (DIS, TFEQ restraint subscale, TFEQ disinhibition subscale, TFEQ hunger subscale, TFEQ flexible control and rigid control subscales (Westenhoefer, 1991), DEBQ restrained eating subscale, and EDE-Q restraint subscale). Other correlations for the restraint scores and total intake on the high-calorie instruction day ranged from 0.005 to 0.26, but none of the correlations was statistically significant. Secondary analyses Median split based on dietary restraint measures Similar to previous studies of the restraint hypothesis, we analyzed the data using a median split. The a values for this Table 1 Means, standard deviations (SD), and results of paired t-tests for the preand post-meal Visual Analog Scale (VAS) line ratings of participants (n ¼ 38) on the low- and high-calorie instruction days
Hunger Fullness Sickness Anxiety Depression Desire to eat Like what eating Pleasant Sweetness Richnessa Bitternessa Novelty
Low-calorie instruction day
High-calorie instruction day
Pre-meal
Post-meal
Pre-meal
Post-meal
M
SD
M
SD
M
SD
M
SD
8.6 3.7 2.3 3.4 2.0 9.3
3.1 3.1 2.9 3.2 3.8 3.3
2.3 11.2 3.5 2.3 1.2 2.5 10.5 10.6 10.2 8.6 1.3 5.8
2.7 2.7 3.7 2.7 2.1 2.7 3.1 2.8 2.9 3.6 2.1 3.6
8.8 3.4 2.3 2.9 1.9 8.7
2.7 2.6 3.3 2.9 2.6 3.7
2.2 11.0 3.5 1.9 1.7 2.1 9.7 9.8 11.4 11.6 1.2 5.9
2.5 2.7 3.3 2.6 2.7 2.5 3.7 3.9 3.2 2.5 2.4 3.8
Note: maximum VAS rating was 15.0 cm. a Post-meal paired t-tests between low- and high-calorie instruction days, po0.05.
ARTICLE IN PRESS R. Sysko et al. / Appetite 49 (2007) 30–37
analysis were divided by 14 in accord with the Bonferroni correction for multiple comparisons (p ¼ 0.05/14, or pp0.004). With this a level, none of the comparisons of high- and low-scoring individuals on the restraint measures (DIS, TFEQ disinhibition, TFEQ restraint subscale, TFEQ hunger subscale, TFEQ rigid control subscale, TFEQ flexible control subscale, DEBQ restrained eating subscale, or EDE-Q restraint subscale) were significant for total meal intake. Correlations between dietary restraint and intake on the low-calorie instruction day The correlations between intake on the low-calorie instruction day and the TFEQ rigid control subscale (r ¼ 0.36, p ¼ 0.03) and the DEBQ restrained eating subscale (r ¼ 0.33, p ¼ 0.04) were statistically significant. However, when one outlier with a TFEQ rigid control score of 6 was removed from the analysis, the correlation was no longer significant (r ¼ 0.20, p ¼ 0.2). While there was a trend toward significance of the correlation between intake on the low-calorie instruction day and the DIS (r ¼ 0.30, p ¼ 0.07), none of the other correlations between measures of dietary restraint and intake on the low-calorie instruction day was significant (range of r ¼ 0.04 to 0.28). Discussion The purpose of the current study was twofold: first, to examine the effect of a cognitive manipulation on total shake consumption during two experimental meal sessions, and second, to evaluate the relationship between selfreported measures of dietary restraint and intake during the meals. The data indicated that the cognitive manipulation was successful, with 86% (n ¼ 38 of 44) participants meeting the criteria for the manipulation check. In addition, on the high-calorie instruction day, participants rated the shake as significantly more sweet by VAS and reported consuming significantly more calories during the session. However, despite the success of the cognitive manipulation and the differences in ratings of sweetness and calories consumed on the high-calorie instruction day, the instructions had little effect on total food intake, as no significant differences in shake consumption were observed between the low- and high-calorie instruction days. We, therefore, did not find support for the hypothesis that participants would consume less when they believed they were consuming a shake made with high-calorie ingredients. In addition, since none of the correlations between the measures of dietary restraint and intake were significant on the high-calorie instruction day, we found no support for our second main hypothesis. When participants with higher scores on measures of dietary restraint were presented with a food believed to be higher in calories, their self-reported attempts at restraint, dieting, or weight loss did not translate into reduced food intake.
35
As many studies of dietary restraint and eating behavior divide participants into low- and high-scoring groups using a median split, we conducted a secondary analysis to determine whether similar results would also be obtained in the current study. No significant relationships were observed between intake and dietary restraint for highscoring participants. Thus, when analyzing the self-report measures of dietary restraint as continuous variables (described above) or using a median split, the results were the same. Another secondary analysis, comparing measures of dietary restraint and intake on the low-calorie instruction day, found that scores on the TFEQ rigid control subscale and the DEBQ restraint subscale were significantly correlated with intake, with higher scores associated with greater consumption. However, this correlation was no longer significant when an outlier was removed from the analysis. As both subscales include items measuring efforts at behavioral restraint to avoid weight gain, this finding suggests that high-restrained participants on the TFEQ and DEBQ exhibited disinhibited eating when they believed that the shake included low-calorie ingredients. The disinhibited eating may have occurred because participants thought they could drink more of the low-calorie shake without consuming a large number of calories or gaining weight. Thus, with the recent proliferation of foods marketed as ‘‘low-calorie,’’ restrained consumers could be triggered to overeat when eating food believed to be low calorie, which could result in weight gain, but the finding needs to be replicated. Taken together, the two secondary analyses found only small effects. Given the size of the current study, the analyses should have been adequately powered to detect a clinically meaningful difference, but the interpretation of these analyses do not support a strong relationship between measures of dietary restraint and eating behavior on either the low- or high-calorie instruction days. The restraint hypothesis posits that restrained eaters are triggered to eat after breaking a dietary rule, or consuming a forbidden food. This theory was not specifically tested in this study. The current study provided participants with a lunch meal instead of using a pre-load paradigm, or the consumption of a food, like a shake, prior to an experimental meal or taste test. A direct comparison of the results of the current study and studies testing the restraint hypothesis is therefore not possible. However, the findings of the current study and two previous studies (Polivy, 1976; Spencer & Fremouw, 1979) can be compared. Although these studies were designed to test the restraint hypothesis, their focus on cognition provides potential explanations for the discrepancies between studies testing the theory of Polivy, Herman, and colleagues and other eating behavior studies. Both Polivy (1976) and Spencer and Fremouw (1979) demonstrated that restrained subjects increased their tastetest intake when they believed they had consumed a highcalorie preload. Therefore, among restrained individuals, the effect of dietary restraint on eating behavior may be
ARTICLE IN PRESS 36
R. Sysko et al. / Appetite 49 (2007) 30–37
primarily related to cognitive factors (Polivy, 1976; Spencer & Fremouw, 1979). The current study also supports the importance of a cognitive control over eating, as a relationship was observed between dietary restraint and total intake when participants received instructions intended to produce expectations about the caloric density of a shake, although only on the low-calorie instruction day. The findings of these studies support Stice et al.’s (2006) suggestion that the use of a cognitive manipulation may lead to a more pronounced relationship between dietary restraint and eating behavior. Other cognitive factors may be important in testing the restraint hypothesis, where participants are asked to consume a pre-load and complete a subsequent ‘‘taste test.’’ This type of food consumption is not typical of naturalistic eating, as it requires the breaking of a dietary rule, and focuses participants’ attention on the pre-load and taste test, with foods (e.g., shake, ice cream, cookies) and a meal structure that are not usual for meals outside of the laboratory. Although studies testing the restraint hypothesis usually observe disinhibited eating among high-restrained participants after consuming a pre-load, these experiments involve procedures that are relatively artificial. Therefore, the relationship between dietary restraint and eating behavior may only be apparent under special circumstances that are elicited in a laboratory meal paradigm under atypical eating conditions. As such, it is important to examine the role the experimental design in influencing the relationship between dietary restraint and eating behavior. The ability to predict eating outside of the laboratory from scores on measures of dietary restraint is an important issue. In three studies, using more naturalistic laboratory paradigms (Martin et al., 2005; Stice et al., 2004; Sysko et al., 2005), no relationship was observed between dietary restraint and eating behavior. In addition, a study assessing reported energy intake among restrained and unrestrained participants (Bathalon et al., 2000) found no effect of dietary restraint and total energy expenditure over an 18-day period. These studies indicate that scores on self-report measures of dietary restraint do not measure acute (Martin et al., 2005; Stice et al., 2004; Sysko et al., 2005) or chronic caloric restriction (Bathalon et al., 2000), and may be assessing the perception of efforts to lose weight as opposed to true weight loss dieting, or the behavioral restriction of food intake (Stice et al., 2006). Self-report measures of dietary restraint may not provide information about behavior during laboratory meals that lack a pre-load, or in studies without instructions designed to draw attention to a specific aspect of the experimental manipulation. Thus, there is only limited evidence that selfreport measures of dietary restraint are valid assessments of food restriction. The clinical significance of the findings from the current study is not clear. Most participants included in the study did not report clinically significant eating disturbances. In addition, only six participants had a BMI in the overweight
or obese range (BMIX25 kg/m2), and only two participants reported that they were dieting to lose weight at either test meal session. In addition, prior to participating in the study, all interested undergraduates were told that the experiment involved consuming a shake made with either high- and low-calorie ingredients. As a result, the sample likely included individuals who had fewer concerns with eating and weight than participants who would not volunteer for a study of eating behavior, as evidenced by the limited range of scores on self-report measures of dietary restraint among participants in the study. Therefore, the results from this study cannot be generalized to other populations, as the sample was primarily normalweight undergraduate women without high levels of dietary restraint or significant eating disorder symptoms. There are also several limitations to the design of the current study. Similar to other studies of eating behavior in a laboratory, it is not known whether the findings of the current study generalize to a naturalistic eating environment. The use of the single-item test meal (ice-cream shake) for lunch, the presentation of the test meal in a large opaque container with the amount and type of food unknown to the participant, is also inconsistent with a typical lunch. Future research should include additional experimental studies to help further elucidate the relationship between dietary restraint and eating behavior, such as a study with a within-subjects design investigating the effect of meal instructions on test meal intake. Participants could be asked in one session to complete a taste test like those used to test the restraint hypothesis (e.g., pre-load shake followed by tasting cookies) and in another to consume a more normal meal (e.g., breakfast as in the Stice et al. (2004) study). In this way, the explanation proposed by the current study for the discrepancy between the studies testing restraint theory and more recent studies of dietary restraint, and the importance of the instructions during the experiment, could be tested. In summary, this study focused on the relationship between a cognitive manipulation, self-reported dietary restraint, and total intake during two experimental meal sessions. Although the cognitive manipulation was successful in producing an expectation about the caloric content of the food used in the study, there was no significant difference in intake between the low- and high-calorie instruction days. There were also few significant relationships between self-report measures of dietary restraint and intake during either test meal. The effects of dietary restraint on food intake in this study and others (e.g., Bathalon et al., 2000; Martin et al., 2005; Stice et al., 2004; Sysko et al., 2005) are not robust, and are not supportive of the restraint hypothesis (e.g., Ruderman, 1986). Measures of dietary restraint may therefore be evaluating another construct (e.g., cognitive restraint over eating) instead of behavioral restraint. To resolve the discrepancies between the studies of dietary restraint and eating behavior, and to better understand the construct assessed by self-report dietary restraint measures, additional studies are needed.
ARTICLE IN PRESS R. Sysko et al. / Appetite 49 (2007) 30–37
References Bathalon, G. P., Tucker, K. L., Hays, N. P., Vinken, A. G., Greenberg, A. S., McCrory, M. A., et al. (2000). Psychological measures of eating behavior and the accuracy of 3 common dietary assessment methods in healthy postmenopausal women. American Journal of Clinical Nutrition, 71, 739–745. Cooper, P. J., Taylor, M. J., Cooper, Z., & Fairburn, C. G. (1987). The development and validation of the Body Shape Questionnaire. International Journal of Eating Disorders, 6, 485–494. Fairburn, C. G., & Beglin, S. J. (1994). Assessment of eating disorders: Interview or self-report questionnaire? International Journal of Eating Disorders, 16, 363–370. Fairburn, C. G., Marcus, M. D., & Wilson, G. T. (1993). Cognitivebehavioral therapy for binge eating and bulimia nervosa: A comprehensive treatment manual. In C. G. Fairburn, & G. T. Wilson (Eds.), Binge eating: Nature, assessment and treatment (pp. 361–404). New York: Guilford Press. Herman, C., & Polivy, J. (1975). Anxiety, restraint, and eating. Journal of Abnormal Psychology, 84, 666–672. Lowe, M. R. (1993). The effects of dieting on eating behavior: A three factor model. Psychological Bulletin, 114, 100–121. Martin, C. K., Williamson, D. A., Geiselman, P. J., Walden, H., Smeets, M., Morales, S., et al. (2005). Consistency of food intake over four eating sessions in the laboratory. Eating Behaviors, 6, 365–372. Murphy, K. R., & Myors, B. (2004). Statistical power analysis: A simple and general model for traditional and modern hypothesis tests. Mahwah, NJ: Lawrence Erlbaum Associates, Publishers. Polivy, J. (1976). Perception of calories and regulation of intake in restrained and unrestrained subjects. Addictive Behaviors, 1, 237–243. Polivy, J., & Herman, C. P. (1985). Dieting and bingeing: A causal analysis. American Psychologist, 40, 193–201. Ruderman, A. J. (1986). Dietary restraint: A theoretical and empirical review. Psychological Bulletin, 99, 247–262.
37
Spencer, J. A., & Fremouw, W. J. (1979). Binge eating as a function of restraint and weight classification. Journal of Abnormal Psychology, 88, 262–267. Stice, E. (1998). Prospective relation of dieting behaviors to weight change in a community sample of adolescents. Behavior Therapy, 29, 277–291. Stice, E., & Agras, W. S. (1998). Predicting onset and cessation of bulimic behaviors during adolescence: A longitudinal grouping analyses. Behavior Therapy, 29, 257–276. Stice, E., Fisher, M., & Lowe, M. R. (2004). Are dietary restraint scales valid measures of acute dietary restriction? Unobtrusive observational data suggest not. Psychological Assessment, 16, 51–59. Stice, E., Presnell, K., Lowe, M. R., & Burton, E. (2006). Validity of dietary restraint scales: Reply to van Strien et al. Psychological Assessment, 18, 95–99. 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. Stunkard, A. J., & Messick, S. (1988). Eating inventory manual. San Antonio: The Psychological Corporation. Sysko, R., Walsh, B. T., Schebendach, J., & Wilson, G. T. (2005). Eating behavior among women with anorexia nervosa. American Journal of Clinical Nutrition, 82, 296–301. van Strien, T., Frijters, J. E. R., Bergers, G. P. A., & Defares, P. B. (1986). The Dutch Eating Behavior Questionnaire (DEBQ) for assessment of restrained, emotional, and external eating behavior. International Journal of Eating Disorders, 5, 295–315. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54, 1063–1070. Westenhoefer, J. (1991). Dietary restraint and disinhibition: Is restraint a homogeneous construct? Appetite, 16, 45–55.