Journal of Personality and Social Psychology 1978, Vol. 36, No. 5, 536-548

Distractibility in Dieters and Nondieters: An Alternative View of "Externality" C. Peter Herman, Janet Polivy, Patricia Pliner, and Joyce Threlkeld University of Toronto, Toronto, Canada

Donna Munic Northwestern University Two experiments were performed in an investigation of the effects of distraction and emotional arousal on the proofreading performance of dieting female subjects. In Experiment 1, it was found that distraction initially impaired the performance of dieters and facilitated the performance of nondieters, a pattern previously shown by Rodin to apply to obese and normal weight subjects, respectively, and interpreted as evidence of greater externality in the obese. Subsequent retesting of the same subjects in succeeding months, however, revealed a complete reversal of the original results. In Experiment 2, the reaction to distraction found in the first phase of Experiment 1 was obtained when subjects were in a situation of minimal threat. In a situation of high threat, the relative distractibility of dieters was reversed, as in the latter phases of Experiment 1. An explanation is offered for these data in terms of the greater emotionality of dieters, the susceptibility of cognitive performance to arousal (distraction, anxiety) manipulations, and the potentially competing effects of distraction and anxiety. Implications for the prevailing "trait" view of externality (stimulus binding) are discussed.

Recent studies by Herman and Polivy (1975) and Hibscher and Herman (1977) have supported the contention that behavioral and physiological differences between obese and normal weight individuals might reflect the fact that the majority of obese individuals are chronic dieters, whereas such is not the case for normal weight persons. When obese and normal weight populations are divided into dieting and nondieting subgroups (Hibscher & Herman, 1977), characteristics initially regarded as particularly descriptive of the obese—such as elevated plasma levels of free fatty acids (an index of hunger or stress) and failure to show caloric compensation to a preload—appear to be related more closely to dieting than to obesity per se. Experiment 1 was done as a senior thesis study at Northwestern University; Experiment 2 was a senior thesis study at the University of Toronto. Requests for reprints should be sent to C. Peter Herman, Department of Psychology, University of Toronto, Toronto, Ontario, Canada MSS 1A1.

Similarly, overreaction to emotional provocation, well established as a characteristic of obese individuals (see Schachter & Rodin, 1974, for a review), was found to characterize normal weight dieters but not normal weight nondieters (Herman & Polivy, 197S). These studies, while clearly suggesting that dieting rather than obesity is the critical determinant of at least some "obese/normal" differences, did not determine precisely what aspect of dieting produces these effects. It has been suggested (Nisbett, 1972) that most obese persons, because of their extensive dieting, are in fact hungry, in the sense of maintaining a body weight substantially below "set point" (the weight that is normally defended by the body's long-term weight regulation mechanisms). In this view, dieting produces hunger, from which the other properties may be derived. Alternatively, Hibscher and Herman (1977) suggested that dieting may act as a stressor—and possibly produce the hypersensitivity to salient stimuli that

Copyright 1978 by the American Psychological Association, Inc.

536

0022-3514/78/3605-0536$00.7S

DISTRACTIBILITY IN DIETERS AND NONDIETERS

Schachter and Rodin (1974) describe as the prime characteristic of the obese—even if dieting is unsuccessful in producing substantial weight loss. In either case, the suggestion that the characteristics of the obese are actually the characteristics of dieters, and that they will be manifested by dieters irrespective of weight classification, warrants further exploration. One purported "obese" characteristic is heightened distractibility. Rodin (1973; also reported in Schachter & Rodin, 1974) has investigated the reaction of obese and normal weight individuals to the imposition of various distractors while performing proofreading and reaction time tasks. It was assumed that tasks of this sort require the subject's full concentration for optimal performance and that distracting stimuli would interfere with performance, especially for the obese, who were portrayed as being particularly sensitive to salient cues in the immediate perceptual environment. The results of the study generally conformed to prediction; as the degree of distraction increased (from no distraction to a voice reciting random numbers, to a voice reciting a bland descriptive passage, to a voice reciting a gruesome, emotionally upsetting passage) performance deteriorated sharply for the obese. Normal weight subjects, however, did not show a decline in performance; if anything, they actually improved as a direct function of the degree of distraction. Rodin concluded that the obese are more distractible than are normal weight individuals, at least as long as the distractors are perceptually salient. This same reasoning was applied to the finding that in terms of self-report, the obese were more strongly affected by the emotionally involving distractor; their general tendency to become more emotional in situations involving a salient "external" emotional stimulus (Schachter & Rodin, 1974) was interpreted as a specific case of their general reactivity to prominent environmental stimuli. One difficulty with the foregoing, stimulusbinding interpretation of the greater distractibility of the obese is that it fails to account for the "somewhat facilitative effect on nor-

537

mals" (Schachter & Rodin, 1974, p. 103) of increasing degrees of distraction. While there was no separate analysis reported for this effect (and while the magnitude of the effect, by visual inspection, was clearly smaller than the corresponding impairment effect on the obese), Rodin reported a linear trend of improvement for normals that clearly contributed to the overall interaction between weight and distraction as joint determinants of proofreading performance efficiency. Furthermore, the tendency for normals' performance to be facilitated by distractors that impair the performance of obese subjects has been found in at least two other contexts. Rodin (1973) found that obese subjects made more errors in a reaction time task when distracted but that the imposition of distraction decreased the error rate for normals. And Rodin and Slochower (1974) found that whereas distraction interfered with memory in an incidental learning task for obese subjects (for at least certain sorts of words), the same distraction improved memory for normal weight subjects for all types of words. The facilitative effect of distraction on normals, then, appears to be as reliable a phenomenon as the impairment effect on the obese and requires its own explanation ; such an explanation does not follow in any obvious manner from the stimulus-binding model in its present form. An alternative interpretation of the results of Rodin's (1973) experiment can be derived from the assumption that performance in the situation under investigation might be mediated by arousal rather than by distraction per se. Sanders and Baron (1975) have suggested that distraction serves to elevate drive level, perhaps because of its frustrative effect; distractors, then, might exert their effects on performance not simply by redirecting attentional mechanisms but perhaps also by increasing drive-induced arousal. The effect of a distracting stimulus, in this view, would be to increase the overall level of arousal of the subject; but the effect of such increased arousal on performance might be either helpful or harmful, depending on the subject's initial arousal level. It is assumed that the relation between performance and

538

HERMAN, POLIVY, PLINER, THRELKELD, MUNIC

arousal is roughly an inverted-U function, with optimal performance achieved under intermediate levels of arousal. Thus, a subject who was in a state of arousal optimal for perforamnce before the imposition of distractors (arousers) might be hindered by such additional arousal. By the same token, additional arousal might improve the performance of subjects who were initially at an arousal level less than optimal for performance. Such a formulation, long familiar to activation theorists (e.g., Hebb, 1955), requires only one additional assumption in order to account for Rodin's data; namely, that obese individuals have relatively elevated arousal levels (as compared to normals) and that they initially be located closer to the optimal performance point on the arousal/performance curve. The evidence concerning the general arousal status of the obese is scanty; and it is not entirely clear what the most appropriate measure of arousal might be. Nevertheless, there is some indication that obese individuals are in fact more aroused (on the whole) than are normals. Obese subjects have been reported as more strongly conditionable (Yaremko, Fisher, & Price, 1975) and have also been repeatedly shown to have elevated blood levels of free fatty acids (e.g., Bjorntorp, Bergman, & Varnauskas, 1969). While there is some confusion as to the precise significance of elevated free fatty acid (FFA) levels (see Hibscher & Herman, 1977, for a discussion), the consensus is that FFA elevations indicate a stress response. Similarly, greater strength of conditioning has long been regarded as an indication of greater drive or arousal. In any case, to the extent that an arousal interpretation of the effects of distraction is capable of accounting for improvements as well as disruptions in performance, it seems worth pursuing. Our first concern, then, was to repeat (with some modifications) Rodin's (1973) experiment, to ascertain whether the improvement in performance shown by some distracted subjects was a reliable phenomenon and also, of course, to demonstrate the impairment of performance in other subjects as a result of exposure to the same distractors.

Experiment 1 Rather than attempt an exact replication of the original Rodin (1973) study, the present investigation, following the rationale of Herman and Polivy (1975), was designed to assess the response to distraction of dieting and nondieting normal weight individuals. Extrapolating from previous work on restrained and unrestrained eaters, we should expect to find that normal weight dieters behave in a manner similar to Rodin's (1973) obese and that normal weight nondieters behave in a fashion similar to Rodin's normals. Furthermore, it was predicted that if Nisbett (and Herman and Polivy) were correct, that is, if sensitivity to "external" stimuli were in fact a result of dieting or underweight rather than simply a cause or correlate of overweight, it should follow that individuals who lose weight over time ought to show increased sensitivity to distractors, while individuals who are initially sensitive ought to become less sensitive to the extent that they gain weight over time. These considerations, strictly, ought to apply only to individuals who have not exceeded their "set point" for weight. It is a reasonable assumption, however, that the college-age sample used in the present study (as in Rodin's and Schachter's studies) were not in a state of hypertrophic obesity (i.e., obesity beyond the "demands" of a hypothalamically defended set point). Most instances of supraset-point obesity are likely to occur after adolescence, when a sedentary lifestyle becomes more prevalent. The first study, then, was designed to (a) replicate the pattern of Rodin's (1973) results, using dieting and nondieting normals rather than obese and normals, and (b) show that an individual's response to distraction changes systematically as a function of changing body weight. Method Overview College-age female l students, identified as dieters or nondieters, performed two proofreading tasks, one 1 Although subjects in the present study were exclusively female, there is no theoretical or empirical

DISTRACTIBILITY IN DIETERS AND NONDIETERS with distraction and one without, in a counterbalanced order. This procedure was repeated twice, at 1-month intervals.

Subjects Subjects were 60 female freshman students at Northwestern University, who received credit toward an experimental participation requirement. Of the original 60 subjects, 2 participated only in the first (October) session, and 5 other subjects changed their dieting status during the course of the experiment.

Procedure Subjects were run individually, at approximately 1-month intervals; each session lasted for less than an hour. Upon arrival, subjects were informed that the experiment was aimed at examining factors affecting performance, and following the general introduction, the specific demands of the proofreading task were explained. Subjects were instructed to work as quickly and as accurately as possible, indicating with a mark the presence of errors in the proofreading passages. The passages were taken from Jane Jacobs's The Death and Life of Great American Cities; each passage was altered so as to contain 75 discriminable errors of spelling, punctuation, verb agreement, and so on. The subject was not required to correct the errors but merely to indicate their presence. At the end of the 10-minute period allotted to each proofreading passage, the subject was requested to indicate the point to which she had progressed in the passage. The two proofreading passages were equivalent in terms of difficulty (as assessed by the efficiency quotient derived from Rodin, 1973); order of presentation was counterbalanced between subjects. Each subject proofread one passage under distracted conditions and the other passage without distraction. The distractor, for all subjects, was a tape-recorded voice reciting a series of random numbers at a moderate volume. This distractor was intended to provide distraction without imposing any emotional arousal on the subjects. The order of presentation of the distractor was counterbalanced between subjects, half being distracted for the first passage they proofread and half being distracted on the second passage. restriction of the correlates of obesity (or dieting) to one sex or another. Previous studies have used male subjects (Hibscher & Herman, 1977; Rodin, 1973) and female subjects (Herman & Polivy, 1Q75; Rodin & Slochower, 1974); results have been similar for both sexes. Because males and females do differ in the degree to which they diet, studies based on a virtual median split of dieting scores, the present study included, avoid certain confoundings by not using male and female subjects in the same sample.

539

After completing the proofreading tasks within a session, each subject filled out the restraint questionnaire (see Herman & Polivy, 1975 for details), which assesses the extent to which respondents are currently concerned about and engaged in active dieting. In addition, the experimenter measured the subject's height and weight after every session. Following the initial (October) session, the experimenter reminded the subject that she had signed up for a three-part experiment and scheduled the subject for another session in 1 month's time. During the second (November) session, the procedure was identical to that in October; likewise, December sessions were identical to preceding sessions, with the addition of a complete debriefing at the end of the final session.

Results Classification oj Subjects Subjects were designated as restrained or unrestrained eaters on the basis of a median split of scores on the restraint scale (Herman & Polivy, 1975) administered at the end of the October session. As in the Herman and Polivy study, females scoring 17 or more were classified as restrained, and those scoring below 17 were considered to be unrestrained. Proofreading

Performance

Various criteria were available for judging the quality of proofreading performance. One measure, the absolute number of errors detected, yielded no statistically significant differences between restrained and unrestrained eaters in response to the imposition of a distractor. The measure used by Rodin (1973) was an "efficiency index," which was calculated as the ratio of errors correctly detected to the number of errors contained in that part of the proofreading passage which the subject covered during the 10-minute period; this ratio was expressed as a percentage. The data on the efficiency measure, for the initial (October) session only, are presented in Table 1. It is clear from this pattern of data that the results with restrained and unrestrained subjects paralleled fairly closely the results found by Rodin with obese and normal subjects, respectively. Restrained subjects' performance deteriorated when the distractor was imposed,

540

HERMAN, POLIVY, PLINER, THRELKELD, MUNIC

Table 1 Proofreading Efficiency Session Only]

Percentage (October

Condition Quiet

Group

Distracted

Unrestrained eaters

M SD n

64.37 13.61

66.83 12.07 30

Restrained eaters

M SD n

65.60 14.51

62.00 13.23 30

whereas the performance of the unrestrained subjects actually improved with the addition of the distractor. Statistically, the only significant effect in the analysis of variance was the interaction between the restraint variable and the distraction variable, jF(l, 56) = 4.55, p < .05. In order to compare performance across time (October to November to December), a distraction index was constructed for each subject at each session. This index was calculated as the ratio of distracted-proofreading efficiency score over undistracted-proofreading efficiency score. Thus, the greater the extent to which distraction interfered with performance, for a given subject, the lower would be her distraction index. The distraction in-

x

1.10

UJ Q

UNRESTRAINED EATERS

1.0;

o cr

IOO

Q

.95

\

RESTRAINED EATERS

OCTOBER

NOVEMBER

DECEMBER

Figure 1. Distractibility over time. (Lower scores represent more impairment of performance under distracting conditions.)

dex of restrained subjects in October, then, was lower than the distraction index for unrestrained subjects. For the 53 subjects who participated in all three sessions and who remained in their original restraint classification (either restrained or unrestrained) for the duration of the experiment, the course of distractibility over time is depicted in Figure 1. It may be seen that whereas restrained eaters were initially more distractible than were unrestrained eaters (as is evident from Table 1), this pattern was reversed during the second (November) session, and this reversal was maintained during the third (December) session. The heightened distractibility of the restrained eater, then, while clearly evident initially, was reversed in subsequent sessions; in November and December, unrestrained eaters were debilitated by the imposition of the distractor, whereas restrained eaters' performance actually improved somewhat when the distractor was in effect. Statistically, this reversal was reflected in a significant interaction between restraint and time (sessions), F ( 2 , 98) = 3.16, p < .OS.2 Effects of Weight Change It was initially predicted that the extent of distractibility might be related to fluctuations in weight over time. Within-subject correlations were performed to assess the relationship between weight and distraction index for restrained subjects (r — — .26, ns), for unrestrained subjects (r = .17, ns), and for all subjects combined (r = —.07, ns). - The use of the distraction index precludes analysis of absolute changes in performance over time. A separate analysis of the efficiency data, with performance under distracted and quiet conditions as a factor in the analysis, revealed that there was a significant effect of time (sessions) on the absolute performance level, F ( 2 , 98) =3.91, p < .OS. The means for performance efficiency were: October, 64.09; November, 62.30; and December, 65.50. From October to November, there was an insignificant decline in performance, F(l, 98) = 1.20, ns, a by-product of the complex changes in distractibility indicated in Figure 1. From November to December, there were no major changes in distractibility, but there was a marginally significant overall improvement in proofreading performance, F(l, 98) =3.84, p < .10, reflecting an apparently uncontaminated practice effect.

DISTRACTIBILITY IN DIETERS AND NONDIETERS

Discussion The results of Experiment 1 indicated that the parallel between obese and normal weight subjects, on the one hand, and restrained and unrestrained normal weight subjects, on the other hand, extends to proofreading performance and the influence of distracting stimuli. Dieting normals, like Rodin's (1973) obese, were initially hampered by the imposition of a distractor, whereas nondieters (like Rodin's normals) showed an improvement when distracted. These results suggest that the greater distractibility of the obese, like some of their other distinctive characteristics, may well be a function of dieting rather than of obesity per se. The main unexpected finding of the study was that the heightened distractibility of the restrained eaters did not persist beyond the first experimental session. In subsequent sessions, the same subjects, despite consistency in their dieting and weight levels, showed a significant reversal, with the result that the nondieters were considerably more distractible; the dieters actually showed an improvement in performance under distracted conditions in the second and third experimental sessions. It was initially predicted that the heightened distractibility of the restrained eaters would be somewhat variable over time and that that variability would be coordinated with changes in body weight. There was a remarkable shift in distractibility, but it was clearly uncoordinated with body weight. Dieters and nondieters reversed their positions with respect to the influence of distractors on performance over time; and while it is at present unclear what the controlling variable (s) for this shift might be, it is apparent that they are uncorrelated with shifts in body weight.3 Only when such controlling variables are identified and held constant will the conditions for a direct test of the body-weightfluctuation hypothesis have been met. The reversal of relative distractibility over time for restrained and unrestrained eaters has a number of implications of theoretical importance. First, it suggests that the elevated distractibility of the obese may also be a more ephemeral trait than has heretofore been

541

assumed. This suggestion depends, of course, on the assumption that the parallels between restrained eaters and the obese extend beyond the pattern of data from the October session to the November and December sessions as well. The original Rodin (1973) study, of course, was a single-session study, as has been the case for most studies of obese/normal differences. The present results suggest, at the least, extreme caution in extrapolating from one particular measurement occasion. This same caution, of course, requires that the present results themselves be replicated. A further implication of the present set of results is that there may be some merit in the arousal interpretation of distractibility differences between obese and normal subjects (or dieters and nondieters). If we consider distractibility to be a function of arousal level, then it follows that changes in arousal level might be responsible for changes in distractibility, such as were observed in Experiment 1. While it was initially argued that obese subjects are in a chronic state of elevated arousal (relative to normals), it is probably reasonable to assume that overall arousal level in a given individual is best regarded as a product of both chronic and acute components. Changes in the acute (situational) component might plausibly be expected to occur over the short term and produce corresponding changes in overall arousal level and thus in performance and the effects of distractors. The foregoing speculations suggest that an appropriate extension of Experiment 1 might be one in which we actively manipulate arousal level (in the acute sense) independently of distraction and assess the effects of arousal on distractibility. It is not obvious why or how subjects in the first study might 3 Only 18 subjects underwent what might be called significant weight changes, 11 losing and 7 gaining 5 or more pounds. The predicted relation between weight change and distractibility did not obtain even for this restricted sample, who exhibited random fluctuations with respect to distractibility scores. It seems unlikely, then, that a consistent relation between body weight change and distractibility would have been found even if all subjects had undergone significant shifts in weight.

542

HERMAN, POLIVY, PLINER, THRELKELD, MUNIC

have shifted in their arousal levels over time; such shifts were simply assumed as a working hypothesis. Experiment 2 was designed specifically to manipulate acute arousal and bring it under experimental control. Ideally, we would expect to find that directly manipulating arousal (as with a fear manipulation) would produce a pattern of results similar to that found over time in Experiment 1. Experiment 2 This experiment was designed to extend and clarify the findings of the initial experiment, in which distractibility was found to decrease over time in restrained normal weight eaters and to increase in their unrestrained (nondieting) counterparts. It was hypothesized that distractibility shifts might be coordinated with shifts in arousal; an equivalent increase or decrease in arousal level might be expected to have differential (or even opposite) effects on subject groups who differ initially or chronically in arousal. In addition to the imposition of distraction on proofreading performance, then, Experiment 2 involved the imposition of an emotional stressor, which was expected to have an independent effect on arousal level during performance. In the original Rodin (1973) study, both distraction and emotional arousal were imposed on the subjects during proofreading performance; but in that situation, those two potential sources of arousal were experimentally confounded, inasmuch as the various distractors differed in their degree of emotional impact. In the present study, the orthogonal manipulation of distraction and emotional arousal permitted an assessment of their influence alone and in combination. Furthermore, the imposition of an emotional arouser provided for replication of the finding of Herman and Polivy (1975) that restrained eaters overreact to emotionally provocative stimuli or at least react more strongly than do unrestrained eaters in objectively equivalent emotional situations.

in a counterbalanced order. In addition, half the subjects performed the tasks while anticipating painful electric shocks, whereas the remainder did not anticipate painful stimulation. Measures of proofreading efficiency and emotional reactivity were recorded, and subjects were retrospectively designated as restrained or unrestrained.

Subjects Subjects were 82 female undergraduates at the University of Toronto, who were recruited on a volunteer basis. Sixteen of the subjects were students in an introductory psychology course and received a course credit bonus for experimental participation.

Procedure Subjects were recruited from four undergraduate courses in which students had completed a number of personality assessment measures. Subjects for Experiment 2 were selected from this pool at random, with the constraint that they be females and that they scored in the top or bottom of the pool distribution on the Spielberger Trait Anxiety Scale (Spielberger, Gorsuch, & Luchene, 1970). Since the experiment was to involve a manipulation of anxiety, it was thought best to ensure that subjects assigned to the two levels of the acute anxiety manipulation be balanced as well as possible with respect to trait anxiety.4 Subjects were recruited by telephone and told that the experiment involved the effects of sensory stimulation on performance and would take about 45 minutes to complete. Upon arrival, subjects were seated and the experimenter explained that the study would involve their performing a proofreading task under four different sensory conditions, in an attempt to discover which factors helped and which factors hindered performance. Subjects were told that the first two proofreading tasks would be performed under different levels of auditory stimulation (quiet and noisy) and the second two tasks would be performed under different levels of tactile stimulation (no stimulation and stimulation). It was explained to all subjects that the auditory portion would involve listening to either random numbers or nothing (a control condition), in a counterbalanced order. Instructions regarding the subsequent portion of the experiment (tactile stimulation conditions) were varied according to whether the subject had been randomly assigned to the high-anxiety or low-anxiety condition. Assignment to condition was constrained only to the extent of ensuring that equal numbers of high-trait and low-trait anxious subjects were as-

Method Overview College-age female subjects performed two proofreading tasks, one with distraction and one without,

4 The variable of trait anxiety was not confounded with the variable of restraint in the present study; the correlation between the two trait measures was far from significant, r(T&) = .18, ns.

DISTRACTIBILITY IN DIETERS AND NONDIETERS signed to the high-anxiety and low-anxiety conditions. Assignment to the high- or low-anxiety condition constituted a between-subjects factor, whereas the distraction manipulation was a within-subjects factor, as in the first experiment. Subjects assigned to the high-anxiety condition were told that in order to affect performance it was necessary to administer "quite high levels of shock" and that while the shock would not be administered until later in the procedure, the electrodes would be attached immediately. The subject was requested to remove any rings, bracelets, and so forth, from her nondominant hand, and two finger electrodes were then attached to the subject's index and middle fingers, following the application of a small amount of electrode paste. The shock generator was plainly visible to the subject and was set at an intermediate intensity. Subjects in the low-anxiety condition were told simply that the amount of stimulation required was minimal and would be felt as a tickle or tingle, if at all. Subjects could see the shock generator, which was set at its lowest intensity, but were not confronted with electrodes and electrode paste. Before performing the proofreading tasks, subjects were asked to complete a "mood scale" in order that "we might take individual reactions into account." The mood scale was a graded version of the Multiple Affect Adjective Check List (Zuckerman & Lubin, 196S), which includes a measure of state anxiety. This scale was later readministered to each subject after she had performed the "first" two proofreading tasks and was ostensibly waiting to perform the second set of proofreading tasks. The proofreading tasks that the subjects worked on were passages taken from Alvin Toffler's Future Shock; two passages had been prepared, with 100 errors in each passage of approximately the same type and distribution throughout the text. Order of administration of the two passages was counterbalanced, as was order of exposure to the distractor (i.e., distracted first or distracted second). As in Experiment 1, the distraction manipulation was effected by having subjects listen over headphones to either a tape-recorded voice reading random numbers or nothing.

543

Following the completion of the two proofreading tasks and the readministration of the mood scale, subjects were informed that there would be no further proofreading and that they would not be receiving electrical stimulation or shock. Subjects completed the restraint scale, as in Experiment 1, and then were fully debriefed, with a complete explanation of the necessity of deceptively threatening shock. During debriefing, information was collected that resulted in 2 subjects being eliminated from the data analysis, 1 because she did not believe that she would receive electric shock and 1 because she was receiving a form of shock treatment for cancer. Results Classification of Subjects The version of the restraint scale used in Experiment 2 was altered somewhat from that used in Experiment 1, on the basis of psychometric refinement (Polivy, Howard, & Herman, Note 1). The revised version has established predictive validity (Polivy, Herman, Younger, & Erskine, Note 2), as well as substantial reliability; it differs from the scale originally used in Experiment 1 and the previous restraint research in that it eliminates one item from the scale, uses a forced-choice response format, and generally produces a lower distribution of scores (see Table 2). In the present study, subjects were designated as restrained or unrestrained on the basis of a median split of restraint scores; subjects scoring 14 or less (n = 42) were considered to be unrestrained, and subjects scoring 15 or more (n — 38) were designated restrained. The slight imbalance in numbers of restrained and unrestrained subjects was attributable to ties at the median.

Table 2 Revised Restraint Scale 1. How often are you dieting? Never; rarely; sometimes; often; always. (Scored 0-4) 2. What is the maximum amount of weight (in pounds) that you have ever lost within 1 month? 0-4; 5-9; 10-14; 15-19; 20+. (Scored 0-4) 3. What is your maximum weight gain within a week? 0-1; 1.1-2; 2.1-3; 3.1-5; 5.1 +. (Scored 0-4) 4. In a typical week, how much does your weight fluctuate? 0-1; 1.1-2 ; 2.1-3 ; 3.1-5; 5.1+. (Scored 0-4) 5. Would a weight fluctuation of 5 Ibs. affect the way you live your life? Not at all; slightly; moderately; very much. (Scored 0-3) 6. Do you eat sensibly in front of others and splurge alone? Never; rarely; often; always. (Scored 0-3) 7. Do you give too much time and thought to food? Never; rarely; often; always. (Scored 0-3) 8. Do you have feelings of guilt after overeating? Never; rarely; often; always. (Scored 0-3) 9. How conscious are you of what you're eating? Not at all; slightly; moderately; extremely. (Scored 0-3) 10. How many pounds over your desired weight were you at your maximum weight? 0-1; 1-5; 6-10; 11-20; 21+. (Scored 0-4)

HERMAN, POLIVY, PLINER, THRELKELD, MUNIC

544

were more reactive to the imposition of anxiety, in terms of both intensity and endurance of response.

Table 3 Measures of Anxiety Low-anxiety measures

High-anxiety measures

Group Unrestrained eaters M SD n

38.83 9.8S

Restrained eaters M SD n

34.06 8.45

37.56 9.12

42.89 40.26 10.21 10.22 19

33.41 9.45

46.14 46.71 9.81 10.89 21

23

17

Anxiety Measures Two measures of state anxiety were taken, one immediately after the anxiety manipulation and one following the completion of the proofreading tasks. These data (see Table 3) were analyzed in a three-factor analysis of variance (anxiety, restraint, and measures), with the first two factors between-subjects and the last factor within-subjects. The analysis revealed a strong effect of the anxiety manipulation, with high-anxiety subjects showing a great deal more anxiety than did lowanxiety subjects, F(l, 76) = 14.54, p < .01. As is also evident from Table 3, the impact of the anxiety manipulation, while strong overall, was particularly dramatic for restrained eaters; analysis of variance revealed that there was a significant interaction between the factors of anxiety and restraint, F(l, 76) — 4.89, p < .05, reflecting the fact that restrained eaters were somewhat less anxious than were unrestrained eaters in the low-anxiety condition but were somewhat more anxious in the high-anxiety condition. A separate analysis of high-anxiety subjects across time revealed that there was a marginally significant interaction between the factors of restraint and measures, F(l, 36) =3.54, p< .10, indicating that whereas unrestrained eaters became somewhat less anxious over time, the initially elevated anxiety of the high-anxiety restrained subjects was maintained (and even increased slightly) over time. The restrained subjects in the present study, then,

Proofreading Measures Proofreading efficiency was measured in the same fashion as in Rodin's study (1973) and in our Experiment 1. The two proofreading passages, though designed to be equivalent in terms of number, type, and distribution of errors, and thus in terms of difficulty, were not at all equivalent in this regard; across all experimental conditions, only 2 of 80 subjects performed better on Passage A than on Passage B, *(79) = 15.43, p < .001. This unforeseen discrepancy between the difficulty levels of the two passages necessitated conversion of raw efficiency scores on a given passage to standard scores, so that the two passages might be considered as equivalent for purposes of statistical analysis of the effects of the various independent variables. Accordingly, the data were analyzed and will be presented in terms of standard scores, with a separate standardization for each passage. The effects of restraint, anxiety, and distraction on proofreading performance may be seen in Table 4; statistical analysis was based on a four-factor analysis of variance (restraint, anxiety, order of presentation of the distraction/quiet variable, and distraction/ quiet itself). 5 Analysis revealed that there were two significant effects: first, a significant interaction involving restraint and anxiety, F ( l , 72) = 5.48, p < .05, and second, a significant higher order interaction involving restraint, anxiety, and distraction, F(l, 72) = 4.70, p < .05. The former interaction is interpretable in terms of the latter effect, and discussion of the results will therefore be in terms of the three-way interaction. One way of viewing the complex interaction is to regard it as a demonstration that the manipulation of anxiety does in fact re5 A comparable analysis was performed using the control factor of trait anxiety rather than restraint as the individual difference factor. There were no statistically significant effects in this analysis involving the trait anxiety factor (or state anxiety factor) as either a main effect or a component in an interaction.

DISTRACTIBILITY IN DIETERS AND NONDIETERS

545

shown by both restrained eaters in high anxiety (M = —20.57) and unrestrained eaters in low anxiety (M = —10.01). The empirical demonstration of the reversal of relative distractibility by altering anxiety lends some support, of course, to the arousal interpretation of distractibility that prompted the present study. However, the fact that shifts in distractibility are coordinated in some fashion with shifts in anxiety or arousal does not explain the precise manner in which anxiety/arousal shifts produce their effects.

verse the relative distractibility of restrained and unrestrained eaters, as was originally predicted. In the low-anxiety condition, the addition of a distractor impaired performance for the restrained subjects and improved performance for unrestrained subjects. In the high-anxiety condition, the foregoing pattern was almost perfectly reversed. In short, it appears that the reversal of the elevated distractibility of restrained eaters, as first revealed in Experiment 1, might well have been mediated by alterations in anxiety or arousal levels. The reversal of distractibility as a function of the anxiety manipulation is made more graphic if the present data are converted into a "distraction index" equivalent to the index used in Experiment 1. Because of the standardization of scores, the most appropriate index was a difference score: Quiet Performance/Distracted Performance X 100. Thus, the higher the score, the greater the impairment due to the imposition of the distractor; negative scores represent improvement under the influence of the distractor. Analysis of variance of these scores revealed, as expected, only one significant effect: There was a significant interaction between restraint and state anxiety, F(l, 72) = 5.05, p < .05. Restrained eaters were most distractible in the low-anxiety condition, (M — 24.49); distractibility was greatest for unrestrained eaters in the high-anxiety condition (M — 21.15). Improved performance under distraction was

Discussion Experiment 2 succeeded in replicating previous research in a number of respects. First, it was found that restrained eaters, like the obese, react more strongly to the imposition of emotional stimuli. This finding is congruent with the findings of Rodin and her colleagues (Rodin, 1973; Rodin, Elman, & Schachter, 1974) and other investigators of obese/normal differences (Pliner, Meyer, & Blankstein, 1974). The present findings are an almost exact replication of the findings of Herman and Polivy (1975) with respect to restrained and unrestrained eaters. There seems to be little lingering doubt as to the elevated emotional responsiveness of obese and dieting subjects, although there is still some debate as to the exact nature and dynamics of such elevated emotionality (see Polivy, Herman, &

Table 4 Proofreading Efficiency

(Standardized Scores) Condition Low anxiety

Group Unrestrained eaters M SD n Restrained eaters M SD n

Quiet

Distracted

-.24 .91

-.13 .85

High anxiety Quiet

.42 .94

.10 1.05 17

.22 .92 19

23 .34 .89

Distracted

-.21 1.16

-.33 1.09 21

546

HERMAN, POLIVY, PLINER, THRELKELD, MUNIC

Warsh, in press, for a discussion). Experiment 2 extended previous work in that it demonstrated that the elevation of anxiety in response to threat was not only more intense for restrained subjects but was also more enduring. Experiment 2 also replicated the findings of Experiment 1, in which the elevated distractibility of restrained subjects was found to be present on one occasion and absent on subsequent occasions. The presence of elevated distractibility in restrained eaters was shown not only to be systematically variable but to be under the control of anxiety. Restrained eaters are more distractible than are unrestrained eaters only when they are calm; the induction of anxiety produces a reversal of the pattern. There is no direct evidence regarding the emotional state of the subjects in Experiment 1—there being no reason (except in retrospect) to investigate emotionality in that study—but the data are consistent with the hypothesis that subjects were relatively calm during the first (October) session and became relatively anxious during the subsequent (November and December) sessions. It might be reasonable to suppose that initially cavalier freshman students became anxious over the course of the experiment in light of the advent of mid-term (November) and final (December) exams; such conjectures, of course, are entirely ad hoc. The pattern of results in Experiment 2, while consistent with the data from Experiment 1, is still essentially unexplained. The initial impetus for the study was the "arousal interpretation" of the original Rodin (1973) results. While we do not have any direct evidence concerning the degree of arousal prevailing in subjects under the various experimental conditions, we may nevertheless attempt to develop a tentative explanatory scheme relating individual differences and experimental variables (and their presumed effects on arousal levels) to performance. First, unrestrained eaters are assumed to be less aroused, all things equal, than restrained eaters. This assumption, at least, is compatible with the little information available regarding the chronically elevated arousal of the restrained eater (Hibscher & Herman, 1977;

Yaremko et al. 1975). c The subject with the lowest level of arousal, then, is the unrestrained eater who is neither distracted nor anxious. Such subjects exhibit quite a low level of performance. The addition of either distraction or anxiety serves to facilitate performance in unrestrained subjects, presumably because each of these variables increases arousal level toward the optimal value for good performance. In the present data, it must be argued that the effect of anxiety is stronger than that of distraction, in terms of arousal induction; but that assumption seems fairly plausible. Finally, the imposition of both anxiety and distraction yields a performance level somewhat below that obtained under anxiety alone; perhaps the addition of two sources of arousal creates a degree of arousal that is now greater than optimal for the proofreading task at hand. Considering the restrained eaters, we find that their performance is best when they are neither distracted nor anxious; we may assume that in the absence of these experimental arousers, their arousal level is close to optimal for the proofreading task. The addition of distraction or anxiety produces a smaller or larger decrement in performance, respectively, quite comparable in magnitude to the increments in performance shown by unrestrained eaters exposed to distraction or anxiety manipulations. The only discrepancy in this speculative mapping of performance onto a theoretical arousal/performance curve is that for the restrained eaters, the effect of distraction and anxiety in combination produces a performance level that is slightly superior to that 0 This assumption seems at first glance to be incompatible with the evidence regarding restrained eaters' lower self-reports on anxiety measures in lowanxiety experimental conditions. It is perhaps possible to argue, however, that chronic elevations of arousal will not necessarily manifest themselves subjectively and, furthermore, that self-reports of emotional activation are based on judgments of current state relative to the normal baseline state characterizing a given individual. Thus, the restrained eater, despite a chronic elevation of arousal, may not report a heightened state of emotion except under conditions of salient external provocation. Admittedly, we have insufficient data to discriminate between these various hypothetical possibilities.

DISTRACTIBILITY IN DIETERS AND NONDIETERS

found under anxiety alone. If we are to retain a relatively simple arousal/performance explanation, we must argue that subjects in the distraction-plus-anxiety condition are less aroused than subjects in the anxiety-only condition. Such a contention is plausible, of course, if we postulate that distraction, in the present context, serves to distract subjects not only from the proofreading task but also from the anticipation of electric shock, with a possible net decline in arousal. Such a distraction-from-anxiety hypothesis, of course, requires independent substantiation but seems compatible with the logical notion that if subjects' attention is redirected away from the source of anxiety, then anxiety-induced arousal will be effectively lessened. A similar distraction-from-anxiety mechanism may also be applied to the unrestrained eaters; their performance is best under anxiety alone, and the addition of a distractor may lessen rather than increase arousal, with a resultant decline in the quality of performance. Consideration of the data in terms of the arousal/performance curve, along with assumptions about the effects on arousal of anxiety, distraction, and their combination, serves to render intelligible a welter of data that otherwise defies interpretation. At present, the foregoing argument can be defended only in terms of a set of mutually supportive assumptions; but the degree of parsimony achieved by tentative acceptance of the argument and its assumptions is attractive. The model outlined above is capable of accounting simultaneously for the following phenomena: (a) impairment of performance by distractors in the obese/restrained eaters; (b) improvement of performance by distractors in normal/unrestrained eaters; (c) the reversal of these effects over time; and (d) the coordination of these effects to anxiety levels. It is unlikely that the stimulus-binding model of distractibility offered by Rodin (1973) can account for as wide a range of phenomena without severe distortion. Conclusion The present study suggests—even demands —research in support of the assumptions re-

547

quired to explain the data of these and Rodin's (1973) experiments. First, more information is needed with respect to arousal levels in obese/normal and restrained/unrestrained groups. Whether the obese/restrained are more aroused in general, as we have assumed, or are simply more arousable (as suggested by the emotionality data) 7 and what the effect of eating and deprivation might be on arousal levels remain questions of empirical and theoretical importance. Above and beyond differences between subgroups, we are left with a suggestion that distraction (an arouser) from anxiety (an arouser) produces less arousal than anxiety alone; such a hypothesis ought to be relatively easily testable in a simpler context. Finally, the present experments suggest that other examples of stimulus-bound behavior in the obese/restrained might also be under the control of situational influences; in short, the "external" cognitive orientation of the obese/restrained may be reversible in general as well as in the specific instance of distractibility. There appears to be no question that there are enduring differences between the subgroups of individuals considered in the obesity/restraint literature; these differences, insofar as they bear on noneating behaviors, however, may simply be derivatives of the heightened arousal/arousability of the obese/restrained. The "reduction" of cognitive externality to elevated arousal levels remains as a challenge for future research. Reference Notes 1. Polivy, J., Howard, K. I., & Herman, C. P. Psychometric analysis of the restraint scale. Unpublished manuscript, University of Toronto, 1976. 7 It has been assumed that subjects in the present study exposed to neither the anxiety nor the distraction manipulation were therefore not in a situation-specific state of elevated arousal; it must be remembered, however, that participation in a psychological experiment probably functions as a minor stressor. The "resting" differences between restrained and unrestrained eaters, then, might be an acute differential response to the arousing effects of the experiment per se. This speculation, of course, requires that such experiment-induced arousal not be given a specific emotional label (e.g., "anxiety") by the subject.

548

HERMAN, POLIVY, PLINER, THRELKELD, MUNIC

2. Polivy, J., Herman, C. P., Younger, J. C., & Erskine, B. Effects of a model on eating behavior: The induction of a restrained eating style. Unpublished manuscript, University of Toronto, 1977.

References Bjorntorp, P., Bergman, H., & Varnauskas, E. Plasma free fatty acid turnover rate in obesity. Acta Medico Scandinavica, 1969, 185, 351-356. Hebb, D. 0. Drives and the C.N.S. (conceptual nervous system). Psychological Review, 1955, 62, 243254. Herman, C. P., & Polivy, J. Anxiety, restraint, and eating behavior. Journal of Abnormal Psychology, 1975, 84, 666-672. Hibscher, J. A., & Herman, C. P. Obesity, dieting, and the expression of "obese" characteristics. Journal of Comparative and Physiological Psychology, 1977, 91, 374-380. Nisbett, R. E. Hunger, obesity, and the ventromedial hypothalamus. Psychological Review, 1972, 79, 433-453. Pliner, P., Meyer, P., & Blankstein, K. Responsiveness to affective stimuli by obese and normal individuals. Journal of Abnormal Psychology, 1974, 83, 74-80. Polivy, J., Herman, C. P., & Warsh, S. Internal and external components of emotionality in restrained

and unrestrained eaters. Journal of Abnormal Psychology, in press. Rodin, J. Effects of distraction on performance of obese and normal subjects. Journal of Comparative and Physiological Psychology, 1973, 83, 68-75. Rodin, J., Elman, D., & Schachter, S. Emotionality and obesity. In S. Schachter & J. Rodin (Eds.), Obese humans and rats. Potomac, Md.: Erlbaum, 1974. Rodin, J., & Slochower, J. Fat chance for a favor: Obese-normal differences in compliance and incidental learning. Journal of Personality and Social Psychology, 1974, 29, 557-565. Sanders, P. S., & Baron, R. S. The motivating effects of distraction on task performance. Journal of Personality and Social Psychology, 1975, 32, 956-963. Schachter, S., & Rodin, J. (Eds.). Obese humans and rats. Potomac, Md.: Erlbaum, 1974. Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. Manual for the State-Trait Anxiety Inventory. Palo Alto, Calif.: Consulting Psychologists Press, 1970. Yaremko, R. M., Fisher, M. L., & Price, J. M. Pavlovian galvanic skin response conditioning in overweight and normal weight women. Journal of Abnormal Psychology, 1975, 84, 429-432. Zuckerman, M., & Lubin, B. Manual for the multiple affect adjective check list. San Diego, Calif.: Educational and Industrial Testing Service, 1965.

Received August 18, 1977 •

Distractibility in dieters and nondieters: an alternative view of "externality".

Journal of Personality and Social Psychology 1978, Vol. 36, No. 5, 536-548 Distractibility in Dieters and Nondieters: An Alternative View of "Externa...
1MB Sizes 0 Downloads 0 Views