Emotional Eating and How It Relates to Dietary

0 downloads 0 Views 1MB Size Report
Herman and Polivy's lO-item Restraint Scale has been the most widely ... ence of these factors—Concern for Dieting and Weight Fluctuation—has led .... An additional 10% of the sample was recruited from Weight Watchers' centers in dose ...

Emotional Eating and How It Relates to Dietary Restraint, Disinhibition, and Perceived Hunger

Richard M . Ganley, P h . D .

During the past decade much research on eating disorders has been devoted to instrument development. Herman and Polivy's lO-item Restraint Scale has been the most widely used instrument, but it has both conceptual and psychometric limitations (Ruderman, 1986). To improve the measurement of restraint, and to combine it with the concept of latent obesity, Stunkard and Messick (1982, 1985) used itemselection and factor-analytic techniques to develop their Eating inventory (El). The current paper reports a factor analysis of this instrument on a large sample (N = 442) of relatively unselected, adult women (age 25-40). The results confirm the ETs Dietary Restraint and Perceived Hunger factors. However, the Disinhibition factor split to form two new factors: Weight Lability and Emotional Eating. The Emotional Eating factor may be particularly important because of the growing body of evidence that emotions-—perhaps even more than cognitions—affect the eating of obese individuals. The Weight Lability factor was able to differentiate between two obese subgroups, suggesting that it may be useful in research on obesity.

Attempts to tmderstand obesity have led to the development of many conceptual models w i t h v a i y i n g explanations for individual differences i n body weight and eating behavior. T w o of the dominant models i n recent decades— Nisbett's (1974) set p o i n t theory and Schachter and Rodin's (1974) externality theory—led H e r m a n and Mack (1975) to develop the concept of restrained eati n g . This concept, elaborated most fully b y H e r m a n and Polivy (e.g., 1980), has generated a great deal of research (see Ruderman, 1986). The restraint concept, w h i c h also focuses o n normal-weight individuals, offers t w o basic h y potheses: (1) " t h a t restrained eaters develop anomolous eating patterns characterized b y dieting and periodic overindulgence" and (2) " t h a t differences i n level of restraint underlie obese-normal differences i n behavior" (Ruderman,

Richard M. Ganley, Ph.D., is Senior Psychologist, The Renfrew Center. Address reprint requests to Richard M. Canley, Ph.D., The Renfrew Center, 475 Spring Lane, Philadelphia, Pennsylvania 19128. international journal of Eating Disorders, © 1988 by John Wiley & Sons, Inc.

V o l . 7, No. 5, 6 3 5 - 6 4 7 (1988) C C C 0276-3478/88/050635-13$04.00



1986, p p . 248 and 249, respectively). I n her review of the restraint titeratnre, Ruderman (1986) concluded that normal-weight restrained eaters exhibit anomolous eating patterns consistent w i t h the restraint concept. However, Ruderman also concluded that the restraint concept, as currently measured, does not seem useful i n understanding obesity or i n explaining obese-normal differences. Most of the studies reviewed by Ruderman used H e r m a n and Polivy's 10i t e m , rationally developed. Restraint Scale (Herman, Polivy, Pliner, Threlkeld, & M u n i c , 1978). A l t h o u g h restraint was originally conceived of as a unitary, complex concept, and although most of the research has used the total restramt score to differentiate between groups on restraint, the Restraint Scale has repeatedly been s h o w n to have at least t w o factors (Ruderman, 1986). The presence of these factors—Concern for Dieting and Weight Fluctuation—has led several researchers to argue that further instrument development is needed to measure restraint and other eating-related concepts (Drewnowski, Riskey, & Desor, 1982; Ruderman, 1986; Stunkard, & Messick, 1985). I n an attempt to improve the measnrement of restraint, and to combine i t w i t h a related concept, "latent obesity" (Pudel, Metzdorff, & Getting, 1975), Stunkard and Messick (1982, 1985; Stunkard, 1981) used item-selection and factor-analytic techniques to develop a n e w Eating Inventory (El). Stunkard and Messick (1982) reported that the E l contains three factors. The 20-item Dietary Restraint factor focuses on conscious restriction of eating behavior and includes most of H e r m a n and Polivy's Concern for Dieting items. The 20-item Disinhibition factor assesses w e i g h t lability and eating triggered by emotional and social cues. These items are believed to measure the degree to w h i c h conscious control over eating is interfered w i t h , that is, " d i s i n h i b i t e d " by social, emotional, and other influences. A l l of H e r m a n and Polivy's Weightfluctuation items are included on this factor of the E l . The 15-item Perceived Hunger factor contains items related to perceptions of hunger and its relationship to eating. The Dietary Restraint and Perceived Hunger factors have been replicated across several groups, whereas the Disinhibition factor has shown some variability and is i n need of further refinement (Stunkard & Messick, 1982, 1985). Stunkard and Messick (1985) recommended d r o p p i n g the El's weight-fluctuation items f r o m the Disinhibition factor. This recommendation was based o n D r e w n o w s k i et al.'s (1982) f i n d i n g that items tike these on the Restraint Scale had an inordinate influence o n the scores of obese subjects. These weight-lability items were the only items added to or deleted from the EI based on information that did not come from Stunkard and Messick's own data. I n line w i t h this recommendation, the 1985 51-item version of the E l contains a reduced 16item Disinhibition factor. Stunkard and Messick (1985) reviewed studies using the E l . I n several studies, Disinhibition has been a better predictor of behavior than Dietary Restraint. This has been f o u n d i n studies of " w e i g h t change dirring depression" (Weissenburg, Rush, Giles, & Stunkard, 1985), "binge severity" (Marcus & W i n g , 1983), and "overeating d u r i n g a laboratory study of food i n t a k e " (Shrager, Wadden, Miller, Stunkard, & Stellar, 1983). I n addition, Disinhibition, b u t not Dietary Restraint, has been significantly correlated w i t h weight changes following the cessation of smoking (HaH, Ginsburg, & Jones, reported i n BjorveU, Rossner, & Stunkard, 1986). Thus, despite its need for further refinement.


Emotional Eating

Disinhibition has proven to be a better predictor of behavior than Dietary Restraint under several conditions. O n the other hand, a retrospective study that focused on behavioral techniques aimed at increasing cognitive control over eating f o u n d that obese subjects w i t h higher scores on the Dietary Restraint factor were more successful i n maintaining a weight loss (BjorveU et al., 1986). The Disinhibition factor was not predictive i n this study. Dietary Restraint has also been related to w e i g h t maintenance and other weight-related variables i n t w o unpublished studies (A. J. Stunkard, personal commurucation). Thus, i t appears that the Disinhibition and Dietary Restraint factors both provide useful information, albeit under different conditions. The Perceived Hunger factor has not had either differential or predictive significance i n any of the above studies. Stunkard and Messick (1985) argued that the Dietary Restramt and Disinhibition factors of the E l distinguish and clarify t w o concepts that were not identified as being distinct o n H e r m a n and Polivy's Restraint Scale. I n addition, the El was developed using sophisticated psychometric techniques, and w i t h adult subjects w h o varied i n their orientation to food and weight control, including b o t h dieting and obese individuals. This contrasts w i t h the predominant use of college students i n studies using the Restraint Scale and may make the E l more directly relevant to treatment studies because adult w o m e n are the most frequent subjects w h o seek treatment. The present paper presents the first postdevelopment factor analysis of the El on a large sample of adult w o m e n .

METHODS Subjects A major goal of subject selection was to obtain a sample of Caucasian w o m e n , 25-40 years o l d , f r o m the broad midranges of socioeconomic status (SES). These criteria were used so that subgroups selected f r o m the f u l l sample w o u l d be similar on race, sex, and SES: variables that have been associated w i t h weight status i n previous research (Rodin, 1982). The subgroups were needed for a study investigating a family systems model of obesity (Ganley, 1986, 1988a). National O p i n i o n Research Center (NORC, 1972) Occupational Prestige Scores were used to assess SES. A l t h o u g h the above criteria place some restrictions on the subject pool, the sample still covers a w i d e age range and a broad spectrum of SES groups and is uruestricted on variables such as weight and dietary status. I n order to achieve the above goals, various selection procedures were considered, including recruitment f r o m businesses, community organizations, shopping mails, supermarkets, and door-to-door canvasing. After some initial w o r k i n these areas, the supermarket recruitment procedure stufaced as the best w a y to gain access to a large sample f r o m a relatively rmselected population ( w t h i n the confines of the selection parameters). Subjects were selected f r o m 20 supermarkets i n w o r k i n g - and middle-class neighborhoods i n and around a large eastern d t y . Very poor and very affluent areas were avoided. Recruitment times were staggered (morning, afternoon, r u s h hour, and evenings o n b o t h weekdays and weekends) i n order to minimize selection biases



that m i g h t result f r o m different groups using the markets at different times. A n additional 10% of the sample was recruited f r o m Weight Watchers' centers i n dose geographic proxunity to the supermarkets. These subjects were i n d u d e d because the research design for investigating the family system model (Ganley, 1988a) called for one obese group currently i n weight-related treatment. Data for the frdl sample on age, SES, education, and weight range are presented i n the Results section. Instrument •The 55-item Eating Inventory (El; Stunkard & Messick, 1982) was completed by all subjects. The major factor w o r k o n the E l has been done on a sample of 220 subjects (97 males and 123 females) using a 67-item preliminary version of the scale. Following the factor analysis of the 67-item scale, item analysis procedures were used to produce the 55-item version. The 55-item version was then factor analyzed o n a second sample consisting of 98 subjects (12 males, 59 females, and 27 sex n o t reported; Stunkard & Messick, 1982, 1985). Subjects for b o t h factor analyses were selected so that about one-third represented an extremely restrained group (dieters w h o h a d lost large amounts of weight), one-third were extremely tmrestrained ("free-eaters"), and one-third were i n termediate (from the same geographic area b u t unselected for any restraint characteristic). The 51-item version of the E l (Stunkard & Messick, 1985) was not used i n the current study (i.e., the 4 weight-lability items discussed earlier were not dropped) because the 55-item version was used i n the original factor analysis, there continues to be interest i n w e i g h t lability i n eating disorders (Ruderman, 1986), data were available for all 55 items, and no reliability data has been presented for the reduced, 16-item Disinhibition scale. The 55-item E l has been reported to have substantial reliability; Dietary Restraint, alpha .93, test-retest .93 (Stunkard & Messick, 1985); Disinhibition, alpha .91, test-retest .80; and Perceived H u n g e r , alpha .85, test-retest .83. I n Stunkard and Messick's sample of 98 subjects Dietary Restraint was significantly correlated w i t h Disinhibition (r = .43, p < .01), b u t n o t Perceived Hunger (r = — .04, ns); Disinhibition and Perceived Hunger were significantly correlated (r = .40, p < .01).. The scoring procedures used b y Stunkard and Messick (1982, 1985) were used here. The present sample of 442 subjects constitutes a substantially larger sample t h a n previously used to factor analyze the E l . Thus, the current research has an i m p r o v e d subject-to-item ratio (8.0:1 for this analysis versus 3 . 2 : 1 for the 67-item scale and 1.8:1 for the 55-item version), and, consequently, the correlation matrix is likely to be more stable. The current subjects also differ f r o m Stunkard and Messick's (1982, 1985) subjects i n that the ones used here were homogeneous for sex (all women) and represent a relatively rmselected sample f r o m the general population (supermarket goers). Procedures Subjects were recruited as they entered or left supermarkets (Weight Watchers' centers for 10% of the sample). Five minutes of their time was requested to discuss a research project on family relationships, assertiveness, modes of expressing anger, and eating habits. Subjects were given a packet of question-


Emotional Eating

naires (including the El) to complete at home and return i n a preaddressed, stamped envelope. Subjects were i n f o r m e d that, as a gesture of appreciation, two $100 stipends w o u l d be given away i n a r a n d o m drawing of completed questionnaires (both prizes were awarded). Nonrespondents were contacted by phone at 1 - and 2-week intervals to encoturage participation. These procedures resulted i n questiormaires being returned b y 79% of the subjects. Following data collection, the E l was factor analyzed using a principal factors procedrue w i t h oblique rotation (Nie, HuU, Jenkins, Steinbrermer, & Bent, 1975). This m e t h o d of factor analysis uses an iterative process, that begins w i t h squared multiple correlation coefficients, to produce estimates of commimahty i n the diagonals of the correlation matrix. The purpose of this analysis was to explore the structure of the E l i n order to find the most representative factor solution. Criteria p r o v i d e d b y Comrey (1978) and A i k e n (project consultant, see first footnote) were used to select the "best" solution. These criteria i n cluded focusing on solutions w i t h factors containing more t h a n a few items (i.e., avoiding w h a t Comrey calls first-level factors that are spuriously produced b y t w o items w i t h similar content), using a minimum criterion of .30 to consider an item as contributing to a given factor, examining various solutions for clarity of factors (i.e., comparing item-factor loadings to determine w h i c h solutions have fewer items loading above the .30 criterion on t w o or more factors simultaneously), and choosing between solutions w i t h different n u m bers of rotated factors based o n b o t h statistical (e.g., size of factor loadings) and rational (e.g., interpretabhity of factors) criteria. This factor analytic approach is very similar to the one used b y Stunkard and Messick (1982).

RESULTS Subject Variables W i t h i n the f u l l sample of 442 w o m e n , 345 were currently married while the remaining 97 were single or divorced. A l l subjects were between the ages of 25 and 40 ( M = 33.2, SD = 4.38) and ranged i n years of education f r o m 10 to 18 ( M = 13.4, SD = 2.09). The m a i n income of subjects came f r o m employment, and all subjects were i n the midranges of SES as indicated b y N O R C Occupational Prestige Scores between 16 and 82 ( M = 46.5, SD = 12.0). This range of scores excludes only the lowest (unemployed and unskilled laborers) and highest (corporate executives and high-paid professionals) social classes. The income of the m a i n providers i n the subjects' households ranged f r o m $10,000 to $75,000 ( M = $27,254, SD = $10,818). Subjects' current, self-report percentage above or below desheable w e i g h t , using the 1983 Metropolitan Life Height and Weight Tables, ranged f r o m - 2 0 . 6 % to 4-100.0%. F r o m the above data i t is dear that, as intended w i t h the selection criteria, subjects covered a broad range of age, education, sodal dass, and w e i g h t status. Eating Inventory The initial factor solution revealed 14 eigenvalues greater that 1.0 (9.20, 6.32, 2.07, 1.70, 1.69, 1.47, 1.41, 1.34, 1.27, 1.20, 1.16, 1.10, 1.08, and 1.02). Inspect i o n of the 14-factor solution according to the criteria discussed earlier made i t



dear that most of these factors had ordy one or t w o items loading above the r n i n i m u m criterion of .30, or that they were spurious, first-level factors (i.e., were produced by t w o items w i t h similar content). (Shmkard and Messick's, 1982, original solution was quite similar. I t had 9 eigenvalues greater than 1.0 and also contained several first-level, " d o u b l e t " factors.) A n inspection of the initial solution produced here suggested that a m a x i m u m of six factors could be meaningfully derived, and factor solutions were obtained for t w o , three, four, five, and six factors. The 6-factor solution contained no items loading above .30 on the final factor, whereas the 5-factor solution contained only t w o items above .30 o n the final factor (one at .35, the other at .33). Thus, i t is clear that there are less than five factors on the E l . The four-factor solution presented i n Table 1 contains six items loading above the .30 criterion on the f o u r t h factor. The average loading for items on Factor FV is .62, and none of these items loads above .30 o n any other factor. The four-factor solution also contained the fewest number of items of the t w o - , three-, or four-factor solutions loading above .30 o n more than one factor at a time. Thus, i t is clear that there are four well-defined factors o n the E l and that irfiormation w o u l d be lost if fewer than four factors are retained. Table 2 presents the five items loading highest on each of the four factors. Factor I V dearly deals w i t h eating d u r i n g periods of dysphoric affect and, therefore, has been labeled Emotional Eating. None of the items on this factor were f r o m H e r m a n and Polivy's original Restraint Scale and, thus, this factor may represent a n e w dimension that has received little attention outside of PudeTs w o r k o n latent obesity. The two items n o t dealing directly w i t h emotional eating on Factor I V , hinging w h e n not h i m g r y (see Table 2) and the remaining item, "Sometimes things taste so good that 1 keep on eating even w h e n 1 am no longer h u n g r y [T F ] , " suggests that some types of uncontrolled eating may be influenced more b y dysphoric emotion t h a n b y either hunger or cognitive restraint. This suggestion is further supported b y the l o w loadings for these items on the Dietary Restraint factor (.04 and .00, respectively) and o n the Perceived Hunger factor (.22 and .15, respectively). A l l of the items o n Factor I V were originally on Stankard and Messick's Disinhibition factor. Most of the remaining items f r o m the original disinhibition factor loaded on the current Factor 11. The five highest-loading items o n Factor 11 deal w i t h weight lability (see Table 2). A l l four of the items identified i n Ruderman's (1986) review as belonging to the Weight Fluctuation factor of the Restraint Scale appear among the five highest-loading items on Factor 11. Thus, Factor I t may be unambiguously labeled as Weight Lability. The splitting of the disinhibition items into t w o separate factors confirms Stunkard and Messick's (1985) difficulty replicating this factor across different groups. The current factor labels also conform w i t h t h e h suggestion to consider "interpretations that do n o t invoke prior inhibition as a prereqmsite" (p. 79). O n l y three items f r o m the original disinhibition factor d i d not load highest o n Factor 11 or I V , and no items f r o m the other original factors loaded highest on Factor 11 or I V . Thus, i t is clear that Stunkard and Messick's 20-item disinhibition factor split to f o r m the Weight Lability and Emotional Eating factors. The t w o remaining factors of the E l were fairly w e l l replicated. Factor 1 h a d loadings of .30 or above for 19 of the 20 original dietary restraint items. This

Table 1 . Factor loadings greater than .30 for Eating Inventory items on the four-factor solution. N e w factor

Items from the three original factors Factor II (Weight Lability)

Factor I (Dietary Restramt) Item" 5 9 [10] 13 18 21 23 28 30 32 33 35 37 39 42 44 46 48 50 52 54

Factor loadings'' .63 -(.27) [.32] .39 .42 .44 .50 .63 .58 .54 .65 .51 .51 .32 .58 .43 .41 .58 .36 .56 .66

Factor III (Perceived Hunger)


Factor loadings''


2 6 8 10 12 14 15 17 20 25 27 31 36 38 41 45 47 49 53 55

-(.06) -(.13);^ -(13)^ .43"' .36 -(.24)" -(.24) -(.11)" -(.06)" .47 -(.11)" -(.16) .32 .51 .43 .49 .48 .36 -(.18)" .35

1 3 4 7 11 [14] 16 19 22 24 26 29 34 40 43 51

Factor loadings'' .31 .47 .63 .37 .51 [.36] -(• .35 .40 .66 .60 -(• .44 .52

Factor IV (Emotional Eating) Item"

Factor loadings

17 20 27 53

.33 .86 .72 .78 .65 .40

-(• -(•

"Item numbers for Factors I, I I , and III correspond to the items appearing on the original scale for Dietary Restraint, Disinhibition, and Perceived Hunger, respectively. The only exceptions are the two items in brackets "[ ] " (number 10 on Factor I and 14 on Factor III) which loaded above .30 on the derived factor but were not on these factors in the original scale. ''Parenthetical values " ( ) " indicate factor loadings for items from the original factors that loaded less than .30 on the derived factor most similar to the original factor. "All items on Factor I V were on the Disinhibition factor in the original scale. "items from the original scale for this factor that loaded above .30 on a second factor.


642 T a b l e 2.

F i v e h i g h e s t - l o a d i n g i t e m s o n the f o u r factors of the E a t i n g I n v e n t o r y .


Factor I: Dietary Restraint


33. 5. 28. 48.

O n a scale of 0 to 10, where 0 means no restraint in eating (eat whatever y o u want, whenever you want it) and 10 means total restraint (constantly limiting food intake and never "giving i n " ) , what number would you give yourself? [Six statements accompany this question to anchor the ratings.] I do not eat some foods because they make me fat. [T F] I deliberately take small helpings as a means of controlling my weight. [T F] I consciously hold back at meals i n order not to gain weight. [T F] H o w Hkely are you to shop for low calorie foods? [1—unlikely to 4 ^ v e r y likely]


Factor H: Weight Labflity


H o w many pounds over your desired weight were you at your maximum weight? [0— zero lbs to 9—-greater than 150 lbs] What is your maximum weight gain within a week? [0—-zero lbs to 9—greater than 9 lbs] I n a typical week, how much does your weight fluctuate (maximum—rtiinimum)? [0—zero lbs to 6—-greater than 5 lbs] M y weight has hardly changed at all i n the last ten years. [T F] What is yoru maximum weight loss within one month? [1—-less than 2 lbs to 10—-greater than 30 lbs]

45. 47. 25. 41.


Factor l U : Perceived Hxmger

24. 4. 26. 40. 11.

I get so hrtngry that m y stomach often seems like a bottomless pit. [T F] Dieting is hard for me because I just get too hungry. [T F] I am always hungry so it is hard for me to stop eating before I finish the food on m y plate. [T F] H o w often do y o u feel hungry? [1—only at meal times to 4—almost always] I often feel so hungry that I just have to eat something. [T F]


Factor I V : Emotional Eating

8. 20. 17. 27. 53.

W h e n I feel anxious, I find myself eating. [T F] When I feel blue, I often overeat. [T F] Sometimes I get so nervous that I just have to eat something. [T F] W h e n I feel lonely, I console myself by eating. [T F] Do y o u go on eating binges even though you are not himgry? [1—never to 4—at least once a week]

factor contains four of the five items w h i c h Stunkard and Messick adapted f r o m H e r m a n and Polivy's Restramt Scale and w h i c h Ruderman's (1986) review identified as belonging to a "Concern for D i e t i n g " factor. Table 2 dearly shows that items o n Factor I deal w i t h consdous attempts to control the intake of food i n order to control weight. Thus, the "Dietary Restrainf' label chosen b y Stunkard and Messick (1985) appears appropriate. Factor H I "Perceived H r m ger" provides a reasonably good confirmation of the hunger f a d o r on the EI. Eleven of the 15 E I hrmger items loaded above .30 o n Factor I I I and none of the hunger items ( i n d u d i n g the 4 that d i d n o t load o n Fador ID) loaded higher on any other factor. As Table 2 reveals. Factor ID items deal w i t h consdous perceptions of hrmger and w i t h the perceived relationship of hrmger to eating behavior. The correlations between the four derived factors are presented i n Table 3. I t should be recaUed that an oblique rotation was used to determine the final factor solution so that the natural relationship between factors could emerge. As can readily be seen. Dietary Restramt is n o t significantly related to Weight

Emotional Eating


Lability, Perceived H u n g e r , or Emotional Eating, w h i l e there are small, significant correlations between each of these other three factors. A l t h o u g h factors I I , i n , and I V are significantly correlated, the size of these correlations indicate that there remains a great deal of independence between all of the factors (highest fr = .19). For comparative purposes, the correlations between the three original EI factors (Stunkard & Messick, 1985) are listed i n the lower portion of Table 3. First, let us compare the correlations for their sample (n = 98, b o t t o m left of table) w i t h the current, f u l l sample {N = 442, b o t t o m middle of table). I t can be seen that the results for the current study ( N = 442) are i n good agreement w i t h Stunkard and Messick's data for the relationship between Dietary Restraint and Perceived Hunger (current study r = .11, theirs r = .04, b o t h nonsignificant). T h e h correlation between Disinhibition and Perceived Hunger (r = .40, p < .01) is also quite similar to the relationship between these factors i n the current study (r = .41, p < .01). However, the lack of a significant relationship i n the current study between Dietary Restraint and Disinhibition (r = .07, nonsignificant) is v e r y different f r o m the significant relationship f o u n d by Stunkard and Messick (1985, r = .43, p < .01). I t can also be seen f r o m the top of Table 3 that there is little relationship between Dietary Restraint and the t w o factors derived f r o m the original Disinhibition factor (for Weight Lability r = .05, nonsignificant; for Emotional Eating r = .10, nonsignificant). Thus, i n agreement w i t h Stunkard and Messick's data. Dietary Restraint does not appear to be

Table 3. Interfactor correlations for the derived and original factors of the Eating Inventory. Derived factors (N — 442, current study)

I. Dietary Restramt n. Weight Labflity m. Perceived Hunger IV. Emotional Eating





.05 —

-.08 .32" —

.10 .44" .43" •—

Stunkard and Messidds original factors Stunkard & Messick (n = 98) 1 1. Dietary Restraint 2. Disinhibition 3. Perceived Hrmger

2 .43" —

Current study (N = 442)



-.04 .40" _


Current study (n = 120) 3

.07 —

-.11 .41" _


2 —

3 .01 —

-.14 .66" _

Note: Sample size: N = 442 is the full sample from the current study; n = 120 is a select sample from the current study representing individuals at varying extremes on restraint and thought to resemble Stunkard and Messick's (1985) subjects; n = 98 is Stunkard and Messick's (1985) sample, also thought to represent varying extremes on restraint (see text for further discussion). "p < .01.



related to Perceived Hunger. However, contrary to their findings. Dietary Restraint also does not appear to be related to Disinhibition or to the t w o factors that were derived f r o m the disinhibition scale. I t is possible that the differences i n the interfactor correlations are the result of differences i n sampling. Stunkard and Messick's subjects were explicitly chosen to represent extremes o n restraint, w h i l e the current sample of 442 subjects was relatively unselected for any restraint-related characteristic (only 10% of the subjects, those selected f r o m Weight Watchers, were chosen based on a variable related to restraint). However, i n selecting subgroups to investigate the family systems model of obesity (for the Ganley, 1988a, study), a subsample of 120 subjects was obtained. This subsample varied w i d e l y o n weightand restraint-related variables and thus more closely resembles Stunkard and Messick's subjects. These 120 subjects included 30 obese w o m e n (>15% above desirable weight on the 1983 Metropolitan Life Tables: M = +37.5%) w h o were currently i n a weight-loss program, 30 obese w o m e n ( M = +34.0%) w h o were not dieting and had not dieted for an average of 6 months, 30 normal weight w o m e n (between ± 1 0 % of desirable weight, M = +2.1%) w i t h a history of weight lability and emotional eating, and 30 normal weight w o m e n ( M = —2.6%) selected for the absence of weight lability and emotional eating. Because these 120 subjects varied w i d e l y on treatment, weight, and dieting status, they more closely resemble Stunkard and Messick's group then the fuU sample of 442. The correlations for these 120 subjects on the original EI scales are also presented i n Table 3 (bottom right). The pattern of correlations is very similar to that for the f u l l sample, and the relationship between Dietary Restraint and Disinhibition remains nonsignificant (r = .01). Thus, i t appears that there is little relationship between Dietary Restramt and the other factors, at least among adult w o m e n . It should be noted that Stunkard and Messick's subjects included b o t h m e n and w o m e n , and this may be another sotuce of the discrepant correlations.

DISCUSSION The Eating Inventory appears to represent several significant advances i n the measiuement of variables related to eating and weight control. I n agreement v f i t h Stunkard and Messick (1985), one of the major advances is the ETs ability to differentiate between conscious dietary restraint and other factors related to eating. Their results, and their review of research using the EI, suggested that dietary restraint is only sHghtly related to w h a t they called disinhibition, and that of these t w o , disinhibition is more strongly related to current weight, depression-related weight gain, binge severity, and eating i n the laboratory. The current results support the differentiation of dietary restraint f r o m other factors involved i n eating b u t suggest that dietary restraint is even less related to the other dimensions measured b y the EI than t h e h data indicated. The present findings also suggest, as Sturikard and Messick had thought possible, that the disinhibition dimension is more complex than t h e h factor analytic studies had revealed. A l t h o u g h the Emotional Eating and Weight Lability factors w h i c h emerged

Emotional Eating


f r o m the disinhibition items w i l l need replication, there are some reasons to believe that they w i l l remain stable. First, the sample used here was AVz times the size of the group used by Stunkard and Messick to analyze the 55-item version of the E l . Second, the current sample was homogeneous for sex and more representative of the general female population (supermarket goers) than their subjects. The current sample, even given the restrictions o n age, sex, and SES, is also more representative of adult w o m e n than the highly restricted college student population used i n m u c h of the research o n restramt. A l t h o u g h it is not k n o w n i f the current results wiU extend to males, they should apply to adult females, the group of greatest interest for research on eating disorders. T h i r d , the items o n the Emotional Eating factor were very homogeneous, and three of the four items specifically deahng w i t h eating related to dysphoric affect h a d the highest loadings of any item o n any factor (.86, .78, and .72; w i t h the f o u r t h one also loading h i g h , .65). Finally, the emergence of a weightlability factor is i n good agreement w i t h many studies that have f o u n d a similar factor on the Restraint Scale (see Ruderman, 1986, for a review). Some researchers have argued that weight fluctuation is a conformding variable (e.g., D r e w n o w s k i et al., 1982) and, indeed, i t probably is w h e n w e i g h t fluctuation items are indiscriminantly summed w i t h other items as i n H e r m a n and Polivy's Restraint Scale. Based o n D r e w n o w s k i et al's. (1982) data, Stunkard and Messick (1985) recommended that the weight-lability items be dropped f r o m their scale and that a reduced, 16-item disinhibition factor be used. However, w h e n weight lability is treated as a separate factor, as the current results indicate i t should be, i t may provide valuable data. For example, d u r i n g a perusal of the data i t was noticed that the t w o obese groups included i n the subsample of 120 d i d not differ on current percentage overweight (f = .96, 116 df, nonsignificant) or on the Emotional Eating factor (f = .12, nonsignificant) b u t were significantly different on the Weight Lability factor {t = 3.3, p < .001). This may indicate that there are different subgroups among the obese population w h i c h can be distinguished b y the Weight Lability factor. The consistent emergence of a weight-lability factor on H e r m a n and Polivy's Restraint Scale, the emergence of a similar factor o n the EI i n the current study, and the ability of this derived factor to discriminate between t w o obese subgroups argue against Stunkard and Messick's (1985) recommendation to simply drop the weight-fluctuation items f r o m the E I . D r o p p i n g these items w o u l d dearly result i n a loss of valuable data. A t the present time i t seems that the most reasonable approach is to use the 55-item version of the EI and to obtain scores for all four of the factors formd i n the present study. However, further scale development w i l l be needed to refine and possibly expand the Emotional Eating and Weight Lability factors. The Emotional Eating factor of the E l may be particularly interesting because of the growing body of research suggesting that emotions strongly influence eating i n obese subjects. Positive finding have been reported i n clinical studies (e.g., Bruch, 1973; Swanson & Dinello, 1970), questioimaire studies (e.g., Lowe & Fisher, 1983; Rand & Stunkard, 1978), laboratory research (e.g.. Pine, 1985; Slochower & Kaplan, 1980; Slochower, Kaplan, & M a n n , 1981), and i n a comprehensive review of the obesity hteralure (Ganley, 1988b). There is also broader renewed interest i n emotion as a primary, precognitive determinant of behavior (Rachman, 1981; Zajonc, 1984) and i n not avoiding emotion as a var-



iable i n research simply because i t is difficult to investigate (Peele, 1981). Thus, the Emotional Eating factor may provide one w a y to further research o n emot i o n and h u m a n behavior w h i c h w i l l be of interest not only to eating-disorder researchers b u t also to the broader professional commuruty. The Dietary Restraint factor may be valuable m treatment outcome studies as suggested b y BjorveU, et al. (1986). The present study confirmed the psychometric quaUty of this factor and, thus, supports its continued use. Little has been said about the Perceived Hunger factor of the EI. This is because Uttle is k n o w n about i t and i t is difficult to relate to other areas of research. I t d i d repUcate reasonably weU i n the current study and, thus, may be w o r t h fruther research. However, Stunkard and Messick (1985) reported that i t is not related to current weight or to overeating i n the laboratory, and, beybnd this, n o t m u c h was reported about i t . Perceived Hunger has also been significantly correlated w i t h Stunkard and Messick's Disinhibition factor, and the relationship between these dimensions (and the t w o factors derived here f r o m the Disinhibition factor) needs to be clarified. Concepts involving hunger, for example, chronic himger/semistarvation, cognitive restraint of eating behavior believed to be caused b y hunger, and various abnormal hunger/satiety mechanisms (e.g., BlundeU, 1980; Booth, 1981; H e r m a n & Polivy, 1980; N i s bett, 1974), continue to flourish i n theories of eating disorders. Perhaps the Perceived H u n g e r factor w i U provide a w a y to study these concepts empirically. I give spedal thanks to Leona Aiken, who at the time of this study was professor of quantitative psychology at Temple University and codirector of the Lrstitute for Srrrvey Research. Her assistance i n the research design and data analysis portions of the study was extremely valuable. I also thank the Department of Psychology at Temple University for providing many resources that aided this project. I thank Stephen Berk and Thomas Shipley for their expert consultation, and Thomas Wible, Jane BurreU, and Nivea Torres for their valuable assistance.

REFERENCES BjorveU, H . , Rossner, S., & Stunkard, A . (1986). Obesity, weight loss, and dietary restraint. Intermtioml Journal of Eating Disorders, 5, 727-734. BlundeU, J. E . (1980). Phannocological adjustment of the mechanisms underlying feeding and obesity. I n A . J. Stunkard (Ed.), Obesity. PhUadelphia: Saimders. Booth, D . A . (1981). Hunger and satiety as conditioned reflexes. I n H . Weiner, M . A . Hofer, and A . J. Stunkard (Eds.), Brain, behavior, and bodily disease. N e w York: Basic Books. Bruch, H . (1973). Eating disorders. N e w York: Basic Books. Comrey, A . L . (1978). Common methodological problems i n factor analytic studies. Journal of Consulting and Clinical Psychology, 46, 648-659. Drewnowski, A . , Riskey, D . , & Desor, J. A . (1982). Feeding fat yet unconcerned: Self-reported overweight and the restraint scale. Appetite: Journal for Intake Research, 3, 273-279. Ganley, R. M . (1986). Epistemology, famUy patterns, and psychosomatics: The case of obesity. Family Process, 25, 437=451. Ganley, R. M . (1988a). FamUy systems i n the assessment and treatment of obesity: With consideration of restrained eating. Submitted for publication. Ganley, R. M . (1988b). Emotional eating m obesity: A review of the literature with implications for research. Submitted for publication. Herman, C . P., & Mack, D . (1975). Restrained and unrestrained eating. Journal of Personality, 43, 647-660.

Emotional Eating


H e m a n , C . P., & PoUvy, J. (1980). Restrained eating. I n A . J. Stunkard (Ed.), Ofcesiiy. Philadelphia: Saunders. Herman, C . P., Polivy, J., Pliner, P;, Threlkeld, J., & Munic, D . (1978). Distractabihty i n dieters and nondieters: A n alternative view of "Externality." Journal of Personality and Social Psychology, 36, 536-548. L o w e , M . R., & Fisher, E . B. (1983). Emotional reactivity, emotional eating, and obesity: A naturalistic study. Journal of Behavioral medicine, 6, 135-149. Marcus, M . , & Wing, R. R. (1983). Binge eating and dietary restraint. Paper presented at the Annual Meeting of the Society of Behavioral Medicine. Baltimore, M D . Metropolitan Life Insurance Company. (1983). 1983 Metropolitan Height and Weight Tables. N e w York: Metropolitan Life, Health, Safety, and Education. National Opinion Research Center ( N O R C ) . (1972). Occupational classification: A s currently used by N O R C (revised). Chicago: University of Chicago Press. Nie, N . H . , HuU, C . H . , Jenkins, J. G . , Stembrenner, K . , & Bent, D . H . (1975). Statistical package for the social sciences. N e w York: McGraw-HUI. Nisbett, R. E . (1974). Starvation and the behavior of the obese. I n G . A . Bray and J. E . Bethrme (Eds.), Treatment and management of obesity. Hagerstown, M D : Harper & Row. Peele, S. (1981). Reductionism in the psychology of the eighties: C a n biochemistry eliminate addiction, mental iUness, and pain? American Psychologist, 36, 807-818. Pine, C . J. (1985). Anxiety and eating behavior in obese and nonobese American Indians and White Americans. Journal of Persorudity and Social Psychology, 49, 774-780. Pudel, U . , Metzdorff, M . , & Oetting, M . (1975). The personaUty of the obese as shown i n psychological tests with special consideration of latently obese subjects (in German). Zeitschrift fur Psychosomatische Medizin und Psychoarmlyse, 21, 345-361. Rachman, S. (1981). The primacy of affect: Some theoretical impUcations. Behavior Research and Therapy, 19, 279-290. Rand, C . S., & Stunkard, A . J. (1978). Obesity and psychoanalysis. American Journal of Psychiatry, 135, 547-551. Rodin, J. (1982). W h y the losing battle? I n B. B. Wolman (Ed.), Psychological aspects of obesity: A handbook. N e w York: V a n Nostrand Reinhold C o . Ruderman, A . J. (1986). Dietary restraint: A theoretical and empirical review. Psychological Bulletin, 99, 247-262. Schachter, S., & Rodin, J. (1974). Obese humans and rats. Potomac, M D : Lawrence Erlbaum. Shrager, E . E . , Wadden, T . A . MUler, D . , Stunkard, A . J., & SteUar, E . (1983). Compensatory intrameal response of obese women to reduction i n the size of food units. Abstracts, Society for Neurosdence, 201, No. 62.8. Slochower, J . , & Kaplan, S. P. (1980). Anxiety, perceived control, and eating i n obese and normal weight persons. Appetite, 1, 75-83. Slochower, J . , Kaplan, S. P., & Marm, L . (1981). The effects of life stress and weight on mood and eating. Appetite, 2, 115-125. Stunkard, A . J. (1981). Restrained eating: What it is and a new scale to measrue it. I n L . A . Qoffi, W . B. T . James, and T . B. V a n ItaUie (Eds.), The body weight regulatory system: Normal and disturbed mechanisms. N e w York: Raven Press. Stunkard, A . J . , & Messick, S. (1982). Psychological factors i n the control of eating behavior. U n published manuscript. Stunkard, A . J . , & Messick, S. (1985). The three-factor eating questiormaire to measure dietary restraint, disinhibition, and hunger. Journal of Psychosomatic Research, 29, 71-83. Swanson, D . W . , & DineUo, F . A . (1970). Severe obesity as a habituation syndrome: Evidence during a starvation study. Archives of General Psychiatry, 22, 120-127. Weissenburger, J . , Rush, A . J . , GUes, D . E . , & Stunkard, A . J. (1985). Weight change i n depression. Psychiatry Research, 17, 275-283. Zajonc, R. B. (1984). O n the primacy of affect. American Psychologist, 39, 117-123.

Suggest Documents