19
Effects of a model on eating behavior: The induction of a restrained eating style' Janet Polivy, C. Peter Herman, Jonathan C. Younger, and Barbara Erskine, University of Toronto ABSTRACT An experiment was conducted to assess the effects of a same-sex model on females' eating behavior. The model ate either a large or small quantity along with the subject in an ad lib satiation context, and either did or did not identify herself as a dieter. Subjects were 86 female undergraduates, split into normally dieting or nondieting sub- groups. Number of sandwich quarters consumed ad lib following a small fixed preload was found to vary as a function of (a) model's consumption (b) model's dieter status and (c) subject's dieter status; there were no significant interactions. A subsequent taste-rating as- sessment of nut consumption, in which the model was present but could neither see nor be seen by the subject, indicated that the three factors which had previously affected sandwich consumption inde- pendently combined to affect nut consumption interdependently. The results were interpreted in terms of the effect of the model on tbe quantity and pattern of consumption, and conclusions were drawn about the dynamics of restrained and unrestrained eating and implica- tions for therapy. A series of studies by Herman, Polivy, and their colleagues (Herman & Mack, 1975; Herman & Polivy, 1975; Herman, Polivy, & Silver, 1979; Hibscher & Herman, 1977; Polivy, 1976; Polivy & Herman, 1976a, b, c) has explored the control of eating behavior in chronic dieters (restrained eaters). This research has shown that such factors as forced preloading (Herman & Mack, 1975; Hibscher & Herman, 1977; Polivy, 1976), emotional upset (Her- man & Polivy, 1975; Polivy & Herman, 1976b) and alcohol con- 1. This research was supported by a grant from the Clarke Institute of Psy- chiatry Research Fund to the first two authors. The authors would like to thank Linda Scott for her assistance as the confederate and Patricia Pliner for her com- ments and criticisms on the manuscript. Reprint requests should be sent to Janet Polivy, Department of Psychology, University of Toronto, Erindale College, Mis- sissauga, Ontario, L5L 1C6 Canada.

Effects of a model on eating behavior: The induction of a restrained eating style1

Embed Size (px)

Citation preview

Effects of a model on eating behavior:

The induction of a restrained eating style'Janet Polivy, C. Peter Herman, Jonathan C. Younger, andBarbara Erskine, University of Toronto

ABSTRACT

An experiment was conducted to assess the effects of a same-sexmodel on females' eating behavior. The model ate either a large orsmall quantity along with the subject in an ad lib satiation context,and either did or did not identify herself as a dieter. Subjects were 86female undergraduates, split into normally dieting or nondieting sub-groups. Number of sandwich quarters consumed ad lib following asmall fixed preload was found to vary as a function of (a) model'sconsumption (b) model's dieter status and (c) subject's dieter status;there were no significant interactions. A subsequent taste-rating as-sessment of nut consumption, in which the model was present butcould neither see nor be seen by the subject, indicated that the threefactors which had previously affected sandwich consumption inde-pendently combined to affect nut consumption interdependently. Theresults were interpreted in terms of the effect of the model on tbequantity and pattern of consumption, and conclusions were drawnabout the dynamics of restrained and unrestrained eating and implica-tions for therapy.

A series of studies by Herman, Polivy, and their colleagues(Herman & Mack, 1975; Herman & Polivy, 1975; Herman, Polivy,& Silver, 1979; Hibscher & Herman, 1977; Polivy, 1976; Polivy &Herman, 1976a, b, c) has explored the control of eating behaviorin chronic dieters (restrained eaters). This research has shownthat such factors as forced preloading (Herman & Mack, 1975;Hibscher & Herman, 1977; Polivy, 1976), emotional upset (Her-man & Polivy, 1975; Polivy & Herman, 1976b) and alcohol con-

1. This research was supported by a grant from the Clarke Institute of Psy-chiatry Research Fund to the first two authors. The authors would like to thankLinda Scott for her assistance as the confederate and Patricia Pliner for her com-ments and criticisms on the manuscript. Reprint requests should be sent to JanetPolivy, Department of Psychology, University of Toronto, Erindale College, Mis-sissauga, Ontario, L5L 1C6 Canada.

Effects of a model on eating behavior 101

sumption (Polivy & Herman, 1976a, c) interfere wi\h a dieter'smaintenance of restraint and result in overeating.

Herman, Polivy, and Silver (1979) attempted to strengthen(rather than disrupt) restraint, or self control. Instead of leav-ing all subjects alone during the ad lib eating phase of the ex-periment, as in most previous eating studies, the experimenterremained in the room as an explicit observer for half of the sub-jects while they ate. The investigators predicted that the pres-ence of an observer during the ad lib eating period would not af-fect unrestrained eaters (nondieters) but would make restrainedeaters self-conscious and inhibit their consumption. The ad libeating occurred in the context of alleviating hunger before the"actual experiment" on taste perception began and followed eithera small (5 oz.) or large (15 oz.) mandatory Nutrament preload.Subjects were instructed by the experimenter to eat as mueh moreNutrament as they needed to "fill yourself up." Unrestrained sub-jects, as predicted, were unaffected by the presence or absenceof the experimenter; they responded to the preload by drinkingsignificantly more Nutrament after a small preload than after alarge one. Restrained subjects, on the other hand, were clearlyinfluenced by the experimenter's presence. When the experi-menter was not present, they behaved as in previous preloadingstudies (Herman & Mack, 1975; Hibscher & Herman, 1977; Polivy,1976), eating more after a large preload and less after a small pre-load (coimterregulating for preload size). When the experi-menter remained in the room however, restrained eaters behavedlike unrestrained eaters, responding to preload size by eating moreafter a small preload and less after a large one. Thus, the presenceof the experimenter induced normal (i.e. compensatory) eatingbehavior in restrained eaters, causing them to eat enough after asmall preload to plausibly achieve satiety but preventing themfrom overeating following a large preload; rather than uniform-ly inhibiting consumption regardless of preload size, the observers'presence induced "sensible" eating. Unfortimately, this sensibleeating in restrained subjects endured only as long as the experi-menter remained in the room. When left alone later to "taste andrate" nuts, restrained eaters reverted to their usual counterreg-ulatory pattern, eating more nuts the more Nutrament they hadpreviously consumed (r = .44, p < .02). Unrestrained subjects,having filled up on Nutrament, ate a uniformly minimal amount

102 Polivy et al.

of nuts, producing a negligible correlation between Nntramentand nut consumption (r = -.07, n.s.). Thus, while observationhad no effect on the behavior of unrestrained subjects, its effeclon normally restrained subjects suggests that restraint is sus-ceptible to social factors. The manifestation of sensible (regu-lated) eating, though less likely in chronic dieters, is clearly fea-sible in certain circumstances.

It should be noted that the effect of the observer in this par-ticular situation may have been mediated by the observer's voleas the experimenter. The experimenter's demand that the sub-jects "fill themselves up'" may well have suggested the sensiblebehavior of the restrained subjects in the observed condition, andthe experimenter's presence may have reinforced this demand,forcing tlie restrained subjects to stop eating after the large pre-load (instead of overeating as they customarily do once their dietsare broken), and to eat what they perceived as a normal amountfollowing the small preload (rather than undereat as their normalrestraint would dictate). Moreover, the observer did not eat alongwith the subjects, further accentuating her status in the situation.Thus, the situation in Herman, Polivy, and Silver's (1979) studywas fai- from a natmal eating situation for the subjects. Althoughit is difficult to create a natural eating situation in the laboratory,it is possible to construct a closer approximation. The presentstudy attempted to do so by using a confederate (purporting tobe another subject in the experiment) as the observer. The con-federate (observer), then, ate along with the subject, which pro-cedure, of course, changed the confederate from a mere observerinto a potential model for the subject to imitate.

Nisbett and Storms (1972) found that both obese and normalweight subjects were responsive to the eating behavior of a model,eating more when the model ate a lot and less when the modelate a minimal amoimt. Subjects in the present study, then, couldbe expected to be influenced by the behavior of the model. It waspredicted that restrained subjects would be especially responsiveto a model, given their greater reliance on social cues in Herman,Polivy, and Silver's (1979) study.

The literature on modeling effects indicates that character-istics of the model can be expected to influence subjects' respon-siveness to the model's behavior. Similarity between model andobserver tends to enhance modeling (Maccoby & Wilson, 1957;

Effects of a model on eating behavior 103

Rosekrans, 1967; Stotland, Zander & Natsoulas, 1961; Tannen-baum & Gaer, 1965). Thus, a dieting (restrained) model shouldbe imitated more by restrained subjects than a nondieting modelregardless of her behavior. For unrestrained subjects for whomdieting status is a less salient dimension of personality, the iden-tification of the model as a dieter or not should not be as im-portant a characteristic as sex, age, and the model's apparentstatus as another introductory psychology student. Therefore, itwas expected the unrestrained female subjects would be fairlyuniformly responsive to a similar (i.e., college age female) modelregardless of her identification as a dieter or not.

The present study was designed to investigate these predictedeffects of a model on restrained and unrestrained eaters. A nor-mal weight, female, college age confederate ate either 2 or 8 sand-wich quarters (following a preload of Nutrament) and eitheridentified herself as a dieter or didn't mention dieting. If thesensible behavior of restrained subjects in Herman, Polivy, andSilver's (1979) study was induced by the confounding of the ob-server's presence with the experimenter's presence (i.e., a matterof demand characteristics, an experimenter effect, as opposed tothe effect of an observer per se, irrespective of "demand"), thissensible pattern should not appear in the present study. Instead,restrained subjects' intake should be inhibited, especially in thepresence of a dissimilar (nondieting) model. A similar (dieting)model should produce imitation (i.e., eating little when she eatslittle and excessively when she overeats). Unrestrained subjectsshould be less responsive to the model overall, although theywould be expected to show some response to the amount eatenby the model (regardless of her dieting status). As in Herman,Polivy, and Silver (1979), any effects of the model or observershould dissipate if she leaves the room or ceases to be an observer(i.e., cannot see the subject); the present study, accordingly, in-cluded a subsequent phase in which subjects ate while the modelwas physically present, but unable to observe the subject's con-sumption during that time.

METHOD

Overview

Eighty-SLx female subjects participated in an experiment purported-ly concerned with "taste perception." A female confederate, posing

104 Polivy et al.

as a naive subject, participated in the experiment ostensibly complyingwith procedural demands made by the experimenter. Subjects andconfederate were instnicted to consume completely a fixed amount ofliquid food (6 ozs.) in the presence of the e.xperimenter. Subjects andconfederate were then presented with a platter of sandwich quarterswith which to fill themselves while the experimenter was absent. Dur-ing this period, the confederate identified herself as a dieter or non-dieter and consumed either two or eight sandwich quarters. Follow-ing sandwich consumption, subjects and the confederate "tasted andrated" a variety of nuts, consuming them ad lib in the absence of theexperimenter and out of sight of each other. Subjects were designatedas restrained or unrestrained on the basis of a questionnaire adminis-tered after the experiment.

Suljjects

Subjects were 89 female undergraduates at the University of To-ronto, who received credit towards an introductory Psychology experi-ment participation option. During recruitment, all subjects were re-quested to refrain from eating (since the experiment involved "tasteperception") for five hours before arriving at the laboratory. Of the 89subjects, 3 were eliminated from the analysis of the data, one for failureto comply with instructions, one for having been previously informedof the observational nature of the experiment and one due to proceduralerror by the experimenter.

Confederate. The confederate was a female secondary school stu-dent in Toronto aged 18 years, 67 inches in height and weighing 120pounds.

Procedure

Subjects and confederate were scheduled to arrive at the laboratory,one subject and the confederate in any given session, during the after-noon (12 noon until 5 P.M.), the subjects having fasted for five hoursbeforehand to insure that they would all be comparably hungry. Sub-jects were told that the experiment involved an investigation of "tasteperception" and that they had been requested to refrain from eatingso that aU subjects would be equivalent with respect to prior taste ex-periences. It was explained that all subjects would be asked to tastefour similar foods and to make perceptual judgments about their taste.However, it was explained that since differing levels of hunger wereknown to influence ta.ste perception, it was necessary to ensure that.subjects were not differentially hungry. Subjects and the confederatewere required to consume a liquid food (Dutch Chocolate Nutrament)and as many sandwich quarters as necessary to make them "comfort-

Effects of a model on eating behavior 105

ably full." Thus, all subjects arrived at the experiment at least some-what food-deprived, consumed a fixed preload and were instructed tofill themselves up in preparation for a perceptual experiment.

Subject and confederate were seated side by side at two standardsize school desks. The liquid preload was served in opaque styrofoamcontainers by the experimenter, one serving each to subject and con-federate. The subject's serving was 6 ozs. and that of the confederatewas 2 ozs., allowing the confederate to convincingly simulate consump-tion of the larger amount. The purpose of the preload, which was ap-proximately the amount drunk by unobserved restrained subjects inHerman, Polivy, and Silver's (1979) study, was to make any sand-wiches eaten exceed the amount a restrained subject with restraintintact would normally consume. After consuming the preload in thepresence of the experimenter, the subject and confederate were offereda platter of 48 sandwich quarters made with white bread, cruststrimmed, consisting of 16 quarters each of roast beef, turkey breast,and cheese filling and an 8 oz. glass of water each. The experimentergave instructions to subject and confederate to fill themselves com-pletely with sandwich quarters taking as many as necessary to achievecomfortable satiation. The experimenter also requested that the sub-ject and confederate not talk so as not to disturb their eating.

Subjects were told that they would have ten minutes to consume thesandwiches while the experimenter was in another room preparing the"taste perception" materials. Subjects were randomly assigned to oneof the four modelling conditions. (The experimenter was blind to boththe subject's restraint level and the condition to which she had beenassigned.) The confederate posed as either a dieter or nondieter, andate either two or eight sandwich quarters. In the dieter condition, theconfederate stated either, "That's enough sandwiches for me. I'm stick-ing to my diet." or "These sandwiches are terrific! To heck with mydiet." Statements were made by the confederate after consumption oftwo quarters; however, after the first statement no more quarters weretaken and after the second, six additional quarters were consumed. Inthe nondieting condition, the confederate stated either, "That's enoughfor me. I'm full." or "These sandwiches are terrifici I could eat aton of them." Again, both statements were made after the consump-tion of two quarters. After the first statement, no more quarters wereconsumed and after the second statement, six additional quarters wereeaten by the confederate. Following the ten minute "satiation" phase,the experimenter removed the sandwich platter; the remaining quar-ters were later counted and the number and type consumed by the sub-ject recorded.

Subject and confederate were then instructed to turn the desks so

106 Polivy et al.

that the confederate and subject were seated back to back with a desk infront of each. Subject and confederate were eacli presented with fourbowls containing cashews, almonds, peanuts, and sunflower seed ker-nels respectively. These bowls were placed in front of the participantsand beside each bowl was placed a rating sheet containing a series often-point rating scales on which tlie subject was to judge the four typesof nuts with respect to how salty, sweet, sour, l^itter, and good tastingthey were and how much the subject liked them. This phase of the ex-periment was presented a.s a taste perception task, requiring an atten-tion to the taste properties and as much accuracy of description as thesubject could give. It was stressed that the bowls had been placed in aspecific order, and that the taste-rating should proceed in that orderonly. Subjects were told to taste as many nuts as they wished from eachbowl while rating that type of nut. Once they had proceeded to thene.xt type of nut, however, they were not to go back to previous bowlsfor the purpose of additional tasting or changing ratings. Once all theratings had been completed subjects were allowed to help themselvesto any of the remaining nuts; they were not to alter their taste ratingsat that point, however. During the "taste-rating" phase of the experi-ment, the experimenter was absent. The confederate commented onthe flavor of the almonds and the peanuts, the two nuts previouslyfound to be moderately liked by most subjects (Herman, Polivy. andSilver, 1979). For half the subjects she praised the almonds saying"Umm, these almonds are delicious," and complained about the pea-nuts saying "Ugh, these peanuts are stale"; for the rest of the subjectsshe did tlie reverse. Thus, the subjects' imitation of the model whenshe was neither observed nor observer could be assessed. The "taste-rating" phase lasted for ten minutes, which allowed ample time for adlih consumption of nuts after completing the scales. During this time,the experimenter waited in an adjoining room with the connecting doorshut.

After the ten-minute "taste-rating" phase, the experimenter returnedand administered two questionnaires, one conceming eating habits andthe other requiring the subject to recall the number and type of sand-wiches eaten by herself and the confederate. Ostensibly to remove thetemptation of the participants to consult on the second questionnaire,the experimenter asked that one person follow her to a different roomto complete the forms, and, after feigning deliberation, the experi-menter asked the confederate to leave with her. The subject was thenleft in isolation to fill out the questionnaires. When the adjoining roomwas reached, the confederate told the experimenter which conditionhad been assigned to the subject. Following completion of the ques-tionnaires by the subject, the experimenter asked for the subject's ad-dress to which a statement of debriefing would b( sent and requested

Effects of a model on eating behavior 107

secrecy about the experiment. Then the subject was weighed and herheight taken. When the subject had left, the experimenter reweighedher four bowls of nuts, and calculated the amount of nuts con.sumed.

RESULTS

Classification of Subjects

The Restraint Scale developed by Herman and Polivy (1975)was modified slightly in response to psychometric analyses (Po-livy, Howard, & Herman, Note 1); a median split on the revisedscale was used to classify subjects as restrained or unrestrainedeaters. The revised scale follows.

Restraint Scale (Revised'^

1. How often are you dieting? (Circle one)

Never Rarely Sometimes Usually Always

2. What is the maximum amount of weight (in pounds)you have ever lost within one month? (Circle one)

0-4 5-9 10-14 15-19 20+

3. What is your maximum weight gain within a week(Circle one)

0-1 1.1-2 2.1-,3 3.1-5 5.1+

4. In a typical week, how much does your weight fluc-tuate? (Circle one)

0-1 1.1-2 2.1-3 3.1-5 5.1+

5. Would a weight fluctuation of 5 lbs. affect the way youlive your life (Circle one)

Not at all Slightly Moderately Very much

6. Do you eat sensibly in front of others and splurge alone?(Circle one)

Never Rarely Often Always

7. Do you give too much time and thought to food?(Circle one)

Never Rarely Often Always

8. Do you have feelings of guilt after overeating? (Circleone)

Never Rarely Often Always

108 Polivy et al.

9. How conscious are you of what you re eating? ( Circleone)

Not at all Slightly Moderately Extremely

10. How many pounds over your desired weight were youat your maximum weight? (Circle one)

0-1 1-5 6-10 lt-2() 21+

Subjects scoring 16 or more were designated as restrained, whilesubjects scoring below 16 were designated unrestrained. Therewere 41 restrained and 45 unrestrained subjects in the presentsample, the unequal distribution being due to tied scores at themedian.

Sandwich Consumption

The principal hypotheses of the experiment concerned the ex-pected effects of the model on restrained and unrestrained eaters.Restrained subjects were expected to be especially responsive tothe models consumption (2 or 8 sandwich quarters) when themodel was similar to them (i.e., also a dieter); a dissimilar modelwas expected to suppress consumption for restrained eaters uni-formly. Unrestrained eaters were expected to be moderately re-sponsive to the amount consumed by the model, but to be unaf-fected by the model's dieting status.

The results, presented in Table 1, indicate that there were nointeractions involving restraint, model's consumption, and/ormodel's dieting status, contrary to prediction. There were, how-ever, three main effects. The subject's restraint classification sig-nificantly affected consumption, F(l,78) =6.41, p < .02, withrestrained subjects consuming less than unrestrained subjects inall comparable conditions. This finding was unexpected and isunprecedented in previous research comparing restrained and un-restrained eaters.

A second significant effect was found for model's consumption(2 versus 8 sandwich quarters), with subjects paired with the 8-quarter model eating much more than comparable subjects pairedwith the 2-quarter model, F( 1,78) = 36.23, p < .001. This effectreplicates the findings of Nisbett and Storms (1972), and fails toconform to expectation for the present study only in that thepowerful effect of the model's behavior on the subjects' consump-

Effects of a model on eating behavior 109

Table 1. Number of sandwich quarters consumed ad lib.

Dieting model Nondieting model

Quarter consumed by model:

Unrestrained eatersMSDN

Restrained eatersMSDN

2

3.000.33

11

2.400.84

10

8

5.291.86

14

4.202.04

10

2

4.002.05

10

2.781.30

9

8

5.901.85

10

5.251.54

12

tion was not differentially strong for the restrained eaters, whohad been expected to be more responsive to such "social" cues.

A third main effect was found for the model's dieting status,with subjects paired with a "dieter" model consuming significant-ly less, F( 1,78) =4.68, p < .05, than subjects paired with a"nondieter" model. This effect was entirely unanticipated, therebeing no prior theoretical basis for predicting that being in thepresence of a dieter would produce a relative suppression of eat-ing, regardless of the normal dieting behavior of the subject oreven of the actual behavior of the model.

Nut Consumption

The influence of restraint, model's sandwich consumption, andmodel's dieting status was assessed in a situation in which the sub-ject and the model were in each other's presence but were un-able to observe each other's behavior. This aspect of the experi-ment, designed to determine the persistence of modeling effects,indicated that the strongest effect during the sandwich-eatingphase of the experiment did persist even in the absence of informa-tion about the model's current behavior. Subjects who had beenexposed to the 8-sandwich-quarter model ate significantly morenuts than did subjects who had previously been exposed to themodel who ate only 2 quarters, F(l,77) =4.72, p < .05. As isevident in Table 2, however, this main effect is subsumed under amore complex interaction between restraint, model's sandwich

2. One subject was eliminated from this phase of the analysis (and all anal-yses of nut consumption) because she was allergic to nuts and could eat only thesandwiches and sunflower seeds.

n o Polivy et al.

consumption, and model's dieter status, F( 1,77) = 3.48, p < .07.The pattem of tliis interaction indicates tliat the subjects who mostclosely tracked the model's prior consumption level (i.e. tite a lotof nuts following an 8-sandwich-quarter model and few nuts fol-lowing a 2-sandwich-quarter model) were restrained subjectspaired with a nondieting model and unrestrained subjects pairedvsdth a dieting model. If neither or both subject and model weredieters, tracking of the model's consumption was negligible.

Relation Between Sandwich and Nut Consumption

Prior research in a similar experimental situation (Herman,Polivy, & Silver, 1979) led to the expectation that restrained eat-ers would show a counterregulatory pattern when consumption ofsandwich quarters and nuts was jointly considered. Unrestrainedcaters, by the same precedent, were expected to show little rela-tion between sandwich and nut consumption. As anticipated, thecorrelation betvk een sandwich consumption and nut consumptionfor restrained eaters was strongly positive (r(38) = .48, p < .01);dieters who ate many sandwich quarters went on to consume arelatively large amount of nuts. Surprisingly, a significant positivecorrelation between sandwich and nut consumption was alsofound for imrestiained eaters (r(43) = .36, p < .02). A more de-tailed analysis of these same data revealed that the correlation wasrobust for restrained subjects whether they had been exposed tothe dieting model (r(18) = .52, p < .02) or to the nondietingmodel (r(18) = .43, p < .06); for unrestrained eaters, the cor-relation was strong only for subjects who had been exposed to thedieting model ir{23) = .53, p < .01), and was much attentuatedfor subjects who had been exposed to the nondieting model (r(18) =.24, n.s.).

DISCUSSION

The present study yielded a number of significant findings,many of them unexpected. The effect of the model on restrainedand xmrestrained eaters will be discussed separately for sandwichconsumption and for nut consumption. Finally, an attempt willbe made to provide an integration and draw conclusions.

The most powerful effect of the model on sandwich consump-tion was the direct behavioral modeling effect, whereby subjects

Effects of a model on eating behavior 111

ate considerably more in the presence of an "overeating" modelthan in the presence of a model who confined herself to 2 sand-wich quarters following the Nutrament preload. This result wasnot unanticipated, although there was initially some expectationthat this effect would be especially pronounced for restrained sub-jects, who were more strongly influenced by social cues (i.e., thepresence of an experimenter-observer) in a prior study (Herman,Polivy, & Silver, 1979). The nondifferential impact of the mod-el's behavior, however, is consistent with the finding of Nisbettand Storms (1972) that obese and normal weight subjects wereequally influenced by the behavior of a model, despite the alleged"external" eating style of the obese.

A second nondifferential effect of the model on restrained andunrestrained eaters was the relative suppression of eating in sub-jects paired with a "dieter" model, irrespective of the model's ac-tual consummatory behavior. It was initially expected that re-strained eaters would more closely imitate the behavior of amodel who was similar to them (i.e., a dieter), perhaps eatingmost when paired with a "dieter" model who ate a great deal.However, Baron (1970) has suggested that a similar model wiUbe imitated only to the extent that the model is not perceived asincompetent; and the "dieter" model in the present study whoabandoned her diet in the subject's presence might be regardedas a clear example of such incompetence, unhkely to be imitated.The most parsimonious way to interpret the main effect of themodel's dieter status on the subjects' consumption is to concludethat the presence of a dieter (or the mention of dieting) alertssubjects to the implicit norm of minimal consumption. This norm,it appears, apphes to restrained and unrestrained subjects ahke;even the unrestrained subjects were influenced by the presence ofa dieter, and it seems likely that restraining one's consumption isa behavioral option open to virtually all (female) subjects in ourculture. Normal differences between restrained and unrestrainedeaters, in the absence of direct or indirect social pressure, arelargely a matter of degree—or better, probability—with re-strained eaters normally encountering more occasions or cognitivepresstnres to exert inhibitions over eating.

Finally, while restrained and imrestrained eaters did not dif-fer in their pattern of consumption (i.e., susceptibility to model-ing effects), they did differ significantly in the absolute amount

112 Polivy et ol.

eaten, with restrained eaters consuming less than did unrestrainedeaters in all comparable conditions. This absolute difference in-dicates that the presence of an observer suppresses intake for re-strained eaters. Although there was no unobserved condition inthe present study for purposes of comparison, this is the first studyemploying restrained and imrestrained subjects in which therehas been a simple main effect of restraint on consumption. In theanalogous Herman, Polivy, and Silver (1979) study, the experi-menter-as-observer influenced behavior, but in a more complexway, presumably because the observer in that study was con-founded with the demand to engage in "sensible" (i.e., regulated)eating. In the present study, in which the observer was not per-ceived as "demanding" a particular pattern of eating, the effectwas one of simple inhibition for restrained eaters, compared totheir unrestrained counterparts.

In sum, the three main effects on sandwich consumption maybe regarded as representing three alternative techniques for in-ducing dietary restraint: (1) being (or regarding oneself as) achronic dieter leads to less eating, at least in the presence of oth-ers; (2) being with someone else who eats less makes one eat less;and (3) being with a dieter makes one eat less, regardless of thedieter's actual behavior. The remainder of the experiment can beviewed as bearing on the question: How do these factors in-fluence subsequent behavior, in a situation wherein the subjectretains her status as restrained or unrestrained eater, as does themodel, but the model neither observes the subject not indicateshow much she is eating?

In a certain sense, the subject was aware of at least one aspectof the model's behavior during the nut consumption phase of theexperiment; the model commented on the taste of two types ofnuts, praising one sort and derogating the other. To the extentthat modeling effects persisted in this phase, it was expected thatsubjects would eat more of the "good" than of the "bad" sort ofnuts. There was, however, no effect whatsoever of the model's

3. Eating less, or sensibly, in the presence of others, of course, is one of the de-fining characteristics of a dieter in the present context (Restraint Scale item # 6 ) ,In the present experiment, however, it does not appear to be the case that theclassification of subjects as restrained or unrestrained was seriously dependent ontheir response to that particular item. The correlation of restraint scores includingand excluding that item from the scale was 0.99, indicating that one's behavior inthe present experimental context had little if any contaminating effect on one'sdieting classification.

Effects of a model on eating behavior 113

Table 2. Grams of nuts consumed.

Quarter consumed by model:

Unrestrained eotersMSDN

Restrained eatersMSDN

Dieting

2

23.9111.35

11

31.5020.27

10

model

8

43.7115,61

14

32.2021.27

10

Nondieting

2

30.4014.47

10

29.7815.57

9

model

8

32.0019.60

10

40.3617.40

11

comments on nut quality. The only major influence on subjects'nut consumption was the number of sandvdch quarters that themodel had consumed earher in the experiment, with greater nutconsumption following greater sandwich consumption. This pat-tern conformed to the "counterregulatory" pattern which has beenthought to characterize restrained eaters. In the present experi-ment, however, the effect was not confined to normally restrainedeaters; even unrestrained eaters appeared capable of exhibitingcounterregulation. The precise extent and circumstances of thecounterregulatory behavior pattern, though, was conditional uponthe presence of a complex combination of factors, as is evidentfrom the marginally significant three-way Interaction (RestraintX Model's sandwich consumption X Model's dieter status) gov-erning nut consumption.

Inspection of the means in Table 2 indicates that restrainedsubjects paired with the dieting model, and imrestrained subjectspaired with the nondieting model, both ate nuts at the minimallevel established in previous research (Herman, Polivy, & Silver,1979). The other two combinations (restrained eaters with thenondieting model, and unrestrained eaters with the dieting mod-el) both ate more nuts following exposure to the model who ate8 sandwich quarters. This complex pattern of results is best con-sidered on a group-by-group basis. Restrained eaters paired withthe dieting model, of course, had eaten the least during the sand-wich consumption phase of the experiment; in short, their restrainthad been maintained intact, and they went on to consume a mini-mal amount of nuts. This generalization, of course, overlooks thefew subjects in this condition who had in fact "overeaten" duringthe sandwich phase; such subjects, in effect, were responsible for

114 Polivy et al,

the appearance ol the counterrcgulatory positive correlation be-tween sandwich and nut consumption that obtained for this group(;• = .52, p < .02). Re.strained eater.s expo.sed lo a nondietingmodel had eaten more dining the sandwicli phase of the experi-ment, especially when the model had eaten eight quarters. Moroof these subjects, then, had "broken restraint/' the result beingthat subjects previously exposed to the 8-sandwich quarters, non-dieting model went on to consume a relatively large quantity ofnuts. For these restrained subjects, as well, the expected counter-regulatory positive correlation ljetween sandwich and nut con-sumption was also robust ( r ---•- .43, p < .06),

A similar consideration of the patterns for unrestrained sub-jects revealed that unrestrained subjects exposed to a nondietingmodel ate the most sandwich quarters in the first phase, and thenwent on to eat only a minimal quantity of nuts, presumably be-cause they had in fact filled themselves up with sandwiches. Thisgroup did not exhibit a significant positive correlation betweensandwich and nut consumption (r = .24, n.s.). The remaininggroup of subjects, unrestrained eaters exposed to the dieting mod-el, had of course shown some inhibition of eating during sand-wich consumption; exposuie to a dieting model, it will be remem-bered, had decreased consumption for both restrained andunrestrained subjects. During the nut consumption phase of theexperiment, these unrestrained subjects behaved virtually iden-tically to the restrained subjects in the group where restraint wasbroken for a high proportion. That is, the unrestrained subjectsexposed to a dieting model showed a strong counterregulatory cor-relation (r = .53, p < .01); and subjects exposed to the 8-sand-wich-qiiarter model went on to consume the greatest amount ofnuts. In short, the behavior of imrestrained subjects exposed to adieting model shows some striking parallels to the behavior of nor-mally restrained subjects, including caloric counterregulation. Itappears that a "restrained orientation" towards eating was in-duced in normally unrestrained eaters in the present experiment.

Although previous research has considered restraint to be acharacteristic trait of a sizable portion of the population, and al-though this trait has been demonstrated to be strongly correlatedwith the behavioral and physiological characteristics that prevail inthe obese (Hibscher & Herman, 1977), it appears that restraint isperhaps best regarded as a state, rather than a trait. This state.

Effects of a model on eating behavior 115

admittedly, appears to endure in some individuals to the extentthat it becomes indistinguishable from a permanent aspect of theircharacter. But an emphasis on the transience (at least in prin-ciple ) of the restrained eating style permits both a clearer under-standing of the experimental data and a more functional approachto clinical application.

In the first case, it may be seen that the present experimentcomplements the study of Herman, Polivy, and Silver (1979); inthe present study, unrestrained subjects were induced to behavein a restrained manner (e.g., to exhibit counterregulation), where-as in the Herman, Polivy, and Silver study, restrained eaters,through a somewhat different manipulation of the individual'sinterpersonal circumstances, were induced to behave as if theywere normally unrestrained eaters (including the adoption ofcaloric compensation). Thus, the degree of restraint that an in-dividual will exhibit, while fairly stable over time (holding thesituation constant) is clearly susceptible to social or other environ-mental influences, and must therefore be regarded in context. Thefact that the Restraint Scale reflects a continuum ought to em-phasize the eapacity of an individual to "slide" on the scale.

As for clinical application, the social induction of a restrainedor unrestrained eating style has clear, if somewhat complex, im-plications for therapeutic intervention with respect to eating dis-orders. The present study indicates that normally unrestrainedeaters may be induced to inhibit their eating by the presence of adieting model; if such an unrestrained individual were faced withthe necessity of losing weight, the desirability of instituting a"buddy system" for meals, with the partner being a dieter of rela-tively inflexible restraint, suggests itself rather directly. It shouldbe remembered, however, that the induction of a restrained eatingstyle appears to carry along with it a proclivity to counterregula-tion, so that failures to inhibit eating might become more likely tolead to subsequent bingeing. Future research must concentrate onthe development of an eating style that is sensible in terms of bothquantity consumed and regulatory pattern.

It appears likely that the achievement of sueh practical goalswill depend critically on the degree to which dieters can be in-duced to monitor their consumption accurately. The occurrenceof bingeing is closely associated with inaccurate monitoring. Inthe present experiment, dieters "binged" (i.e., overate nuts) foi-

116 Polivy et al

lowing the consumption of 5.25 sandwich quarters (in the 8-quar-ter/nondieting model condition), but retained their restraint fol-lowing consumption of 4.20 sandwich quarters in the 8-quarter/dieting model condition. It seems highly unlikely that the dif-ference of 1.05 sandwich quarters was sufficient in itself toaccount for the production of subsequent overeating. More plau-sible is the possibility that subjects' overindulgence in the non-dieter condition was triggered by some cognitive event, such asthe perception that they had exceeded the limits of appropriaterestraint. Postexperimental debriefing included a question aboutthe number of sandwich quarters that subjects could rememberhaving consumed; 18 out of 20 restrained subjects in the dieter-model condition were accurate, whereas only 10 out of 21 re-strained subjects in the nondieter-model condition were accurateX^ = 3.93, p < .01). (Unrestrained subjects' accuracy did notvary as a function of the type of model.) Cognitive monitoring,then, as well as overt behavior, appears to be related to social in-fluence; the further investigation of the effects of social circum-stances on thought and behavior, and the mutual influence ofthought and behavior on each other, must serve as a guide fortherapeutic intervention.

REFE«ENCE NOTE

1. Polivy, J., Howard, K. I., & Herman, C. P. Psychometric analysis of therestraint scale. Unpublished manuscript, Loyola University, 1976.

REFERENCE.^

Baron, R. A. Attraction toward the model and model's competence as determinantsof adult imitative beha-vior. Journal of Personality and Social Psychology, 1970,14, 345-^51.

Herman, C. P., & Mack, D. Restrained and unrestrained eating. Journal of Per-sonality, 1975, 43, 647-660

Herman, C. P., & Polivy, J. Anxiety, rfstraint, and eating behavior. Journal of Ab-normal Psychology, 1975, 84, 66^672.

Herman, C. P., Poli\'y, J., & Silver, R. Effects of an observer on eating behavior:the induction of sensible eating. Journal of Personality, 1979, 47, 85-99.

Hibscher, J. A., & Herman, C. P. Obesity, dieting, and the expression of obesecharacteristics. Journal of Comparative and Physiological P.sychology, 1977,91, 374-380.

Maccoby, E. E., & Wilson, W. C. Identification and observational learning fromfilms. Journal of Abnormal and Social P.iyc}iology, 1957, 55, 76-87.

Nisbett, R. E., & Storms, M. D. Cognitive and social determinants of food intakein Harvey S. London and Richard E. Nisbett (Ed,s,). Cognitive AUemation ofFeeling States, Chicago: Aldine, 1972.

Polivy, J. Caloric perception and regulation of intake in restrained and unre-strained subjects. Addictive Behaviors, 1976. 1. 2.37-243.

Effects of a model on eating behavior 117

Polivy, J., fit Herman, C. P. Clinical depression and weight change: A complexrelation. Journal of Abnormal Psychology, 1976, 85, 338-340. (a)

Polivy, J., & Herman, C. P. The effects of alcohol on eating behavior: Disinhibi-tion or sedation? Addictive Behaviors, 1976, 1, 121-125. (b)

Polivy, J., & Herman, C. P. Effects of alcohol on eating behavior: Influence ofmood and perceived intoxication. Journal of Abnormal Psychology, 1976, 85,601-606. (c)

Rosekrans, M. A. Imitation in children as a function of perceived similarity to asocial model and vicarious reinforcement. Journal of Personality and SocialPsychology, 1967, 7, 307-315.

Stotland, E., Zander, A., & Natsoulas, T. The generalization of interpersonalsimilarity. Journal of Abnormal and Social Psychology, 1961, 62, 250-256.

Tannenbaum, P. H., & Caer, E. P. Mood change as a function of stress of pro-tagonist and degree of identification in a film-viewing situation. Journal ofPersonality and Social Psychology, 1965, 2, 612-616.

Manuscript received Jurte 28,1977.