Behavior and weight changes in three obese adolescents

Behavior and weight changes in three obese adolescents

BEHAVIOR THERAPY 12, 383-399 (1981) Behavior and Weight Changes in Three Obese Adolescents THOMAS J . COATES The Johns Hopkins University School of ...

854KB Sizes 0 Downloads 20 Views

BEHAVIOR THERAPY 12, 383-399 (1981)

Behavior and Weight Changes in Three Obese Adolescents THOMAS J . COATES

The Johns Hopkins University School of Medicine

CARL E. THORESEN Stanford University The present study explored the relationships between treatment strategies, changes in eating behaviors and home environments, and weight loss. Two obese females were trained in weight management techniques over 10 weeks. Treatment for $2 was lagged 1 week behind treatment for SI. A third subject received a placebo treatment to control for the reactive effects of therapist contact and observation. Changes in eating behaviors and home environments were measured using self-reports and observations in the subjects' homes by nonparticipant observers. Ss 1 and 2 lost 21 and 11.5 pounds respectively, while $3 gained 5 pounds. Ss 1 and 2 showed different behavior changes: S1 modified variables associated with the home environment, food portions, and exercise, while $2 changed eating behaviors and food service. These outcomes suggest that behavioral prescriptions are associated with weight loss and that treatments tailored to individuals on the basis of behavior analyses may be useful.

Behavioral techniques have shown some promise for treating obesity among young persons (Brownell & Stunkard, 1978a). Their application with children and adolescents, however, has generally shown the same problems as applications with adults: Weight losses have rarely reached clinical significance, variability among subjects has been high, and maintenance of weight losses has been disappointing (Coates & Thoresen, 1978a, 1980). The issue is more than academic, as obesity among children and adolescents is prevalent, persistent, and positively implicated in in-

This research was supported in part by the Spencer Foundation, the Proctor and Gamble Fund, and by the Boys Town Center at Stanford. We thank the following persons whose help and advice were critical: Rosedith Sitgreaves, W. Stewart Agras, David Gillem, and Ann Gladstone. Judi Komaki, Kelly Brownell, and G. Terence Wilson offered invaluable feedback on earlier drafts of this paper. Requests for reprints should be sent to Thomas J. Coates, Division of Pediatric Cardiology, Johns Hopkins Hospital, 600 North Wolfe Street, Baltimore, MD 21205. 383 0005-7894/81/0383-039951.00/0 Copyright 1981by Associationfor Advancementof BehaviorTherapy All rightsof reproductionin any form reserved.

384

COATES A N D THORESEN

creases in risk factors for cardiovascular disease (Coates & Thoresen, 1980). Behavioral treatments for obesity are based on the assumption that external environmental variables (e.g., sight, accessibility and social cues to eat, quality of food and time of day) and behaviors associated with the act of eating (e.g., rate of eating, location, and extraneous activities associated with eating) influence eating and activity patterns. Prescriptions to modify the act of eating and the external environment form the basis of most behavioral approaches to weight loss (cf. Stunkard, 1975; Stunkard & Mahoney, 1976). When investigated empirically, however, the reported relationships between behavior changes and weight loss among adults have been supported weakly (Brownell & Stunkard, 1978b; Jeffery & Coates, 1978). Three studies have shown that eating behaviors were modified in behavioral treatment programs (Hagen, 1974; Mahoney, 1974; Wollersheim, 1970). Brownell and Stunkard (1978b) point out that " . . . these three studies constitute the total empirical evidence that weight change results from prescribed behavior change. Each relies on a single self-report by subjects whose answers may have been biased by their knowledge of their relative success and failure in treatment" (p. 301). Other studies using more comprehensive measures have not demonstrated relationships between behavior and weight changes (cf. Bellack, Rozensky, & Schwartz, 1974). Jeffery, Wing, and Stunkard (1978) analyzed data from daily eating records. Subjects showed significant changes on 7 of 9 variables from week 1 to week 5 of treatment. Correlations between changes in behavior and weight loss from week 1 to week 5 were nonsignificant. Brownell, Heckerman, Westlake, Hayes, and Monti (1978) required subjects and spouses to record their own and their spouses' eating behaviors every day. Correlations between weight loss and (1) spouses' ratings of subject behavior, (2) spouses' self-reports of behavior, and (3) subjects' report of spouse behavior were not significant. Correlations between subjects' ratings of their own behaviors and spouses' ratings of the same behaviors were not significant. Finally, Stalonas, Johnson, and Christ (1979) collected daily records from 44 subjects and correlated 10 categories of behaviors with weight changes. Nine of the ten behavior scores and an overall adherence score failed to correlate with weight loss. The present study was designed to examine relationships between treatment strategies, changes in eating behaviors and home environments, and weight loss. We were interested in documenting answers to two questions: (1) Do obese subjects really perform the behaviors they are taught to perform, and (2) Do those subjects who perform the behaviors lose weight?

METHOD

Subjects Subjects, recruited by advertisements placed in local newspapers, were required first to obtain a medical examination to rule out physical con-

BEHAVIOR AND WEIGHT CHANGES

385

ditions possibly complicating treatment. A pool of 10 respondents showed no medical problems and agreed to participate. The three heaviest subjects were then chosen to participate and randomly assigned to treatment conditions. Subject 1 (S 1) (16 years of age, 286 lbs., 65½") was 128.4% above average weight at the beginning of treatment. All siblings and her father were normal weight, while her mother was obese. Subject 2 ($2), the oldest of four children (16 years of age, 194 lbs., 62"), was 71.4% above average weight. All siblings were normal weight but her mother and stepfather were overweight. Subject 3 ($3) (15 years of age, 215 lbs., 66"), 72.9% above average weight, was the youngest of four children. Her siblings were all normal weight, while her parents were slightly overweight.

Design A time-series design was used. Control was established using three design features: (1) Time-lagged treatments: Ss 1 and 2 were administered treatment over a 10-week period, yielding two independent replications of the treatment program. Treatment for $2 was lagged 1 week behind treatment for S 1 to control for possible delayed reactive effects of contact and observation. (2) Uninterrupted time series: $3 met with the therapist at the same intervals and for the same amounts of time as did the Ss 1 and 2 and was subjected to the same data collection procedures. $3 was a control for the reactive effects of therapist contact and the intensive observation system. $3 was also given written summaries of the experimental treatment program at the same intervals as they were administered to S 1 to control for demand and information effects. (3) Multiple baseline observations: Treatments were lagged across weeks within each subject. Treatment The 10-week treatment program was designed (1) to teach subjects skills for controlling eating and exercise behaviors; and (2) to teach family members specific procedures for supporting and encouraging their children's use of appropriate eating behaviors. Program components and the time line for their implementation are presented in Table 1. S s 1 and 2 met twice weekly with the experimenter for approximately 1 hour each time. Sessions were devoted to reviewing progress as determined by body weight reductions and eating habit changes, discussing problems, and learning and implementing new skills for controlling eating and exercise behaviors. Lectures, modeling, and guided practice were used to teach these skills. Family sessions (marked with an asterisk in Table 1) were conducted in Ss' homes with all family members present. These sessions encouraged family members to learn and apply changes in family food storage and buying patterns, in the foods served at meals, and in family interaction patterns related to eating. $3 was seen by the experimenter twice weekly for the duration of the study, and her family was visited in their home five times. All observation

386

COATES

"~

AND

6-~."

THORESEN

2.~-~.~

~ ~.~

~

~.~

~ 0

•r . . ~

o ~ ~.=.=

~

~ i ,. . . . . . .

~.~ oo °

.~

.~

~

==

~

~

m

~.~ ~ . o

=o~

0

Z ua b, < oa

"g m e~ N

Z

m

"

=

"g

g

~

o

e~

387

BEHAVIOR A N D W E I G H T C H A N G E S

•~ "~

~ ~

.~ ~.~

~ ~

~

r, ~

~ ~

"~ c~

"

o0'~

E

6 r',

E •~

?',

0

.E

"?., 0

6

C c~

~h C~

l

C~

Z

388

COATES AND THORESEN

procedures were conducted on the same schedule as Ss 1 and 2. Treatment sessions were devoted to nonspecific discussions of weight and nutrition and the social and personal implications of being overweight. $3 and her family received written summaries of each major component in the experimental program offered to Ss 1 and 2. These summaries were not reviewed verbally in the treatment sessions. All subjects were informed at the onset that subjects would be receiving either a true treatment or a placebo treatment. Subject 3 was informed at week 10 that she had been receiving placebo and that a procedure shown effective with the other two subjects could now be tried with her. She declined, saying that she preferred " . . . not to worry about weight loss right now."

Dependent Measures Body weight was measured once weekly, at approximately the same time each week, by the experimenter using a 300-lb. capacity fulcrum scale. Measurements were rounded to the nearest ½ pound. Height was measured at the same time using a height calibrator (Holtrain, Ltd.) and rounded to the nearest tenth of a centimeter. These data were used to compute percentage overweight using the U.S. Department of Health, Education, and Welfare (1973) standards. The Daily Energy Diary, a self-report measure, was designed to serve both clinical and research purposes. Prior to eating a meal, ingesting a snack, or engaging in exercise, the subjects recorded the time, place, food, amount, and associated activities. Feelings of hunger were rated on a scale of " 0 " (not hungry) to " 3 " (extremely hungry). Subjects also recorded the type and amount of any special exercise engaged in during the study. Thus, the instrument yielded data on six dependent measures: average meal duration (total self-reported duration of meals in minutes divided by number of meals eaten); number of places at home food eaten; number of times food eaten; number of activities while eating; minutes of exercise; and average hunger rating (sum of hunger ratings divided by number of hunger ratings). The Eating Analysis and Treatment Schedule, (EATS) 1 (Coates & Thoresen, 1978b) was designed to be used by nonparticipant observers in subjects' homes at mealtimes to collect data on the physical environment and on subject eating behaviors. Observations were collected four times weekly (3 randomly chosen weekdays and 1 weekend day) during the 14 weeks of the study. Observers arrived prior to the meal to record foods out in various parts of the house and to list foods in the first row of the cupboards, refrigerator, and freezer. The observers recorded the kinds of foods prepared for the meal and the manner in which it was served. They then sat away from the table during the evening meal to record client eating behaviors. The frequency of bites, sips, utensils down, and talks were counted for the duration of the entire meal.

1 It is important to note that the families were cooperative and willing to participate despite the pessimism o f our colleagues that the study was feasible.

BEHAVIOR AND WEIGHT

CHANGES

389

The following variables were derived from the EATS: foods displayed in house (average caloric ratings of food sitting out); cupboard caloric rating (average caloric rating of foods in the first row of food cupboards); refrigerator caloric rating (average caloric rating of foods in first row of refrigerator shelves or on refrigerator door); freezer caloric rating (average caloric rating of foods in first row of freezer); food service ratio (food items not served on table divided by total food items served for meal observed; a score of " 1 " indicated that all foods were served from the kitchen counter and not from the table); helpings per average family helping (subject helpings divided by average number of separate servings taken by all family members; a score of " 1 " indicated that the subject took the same number of helpings as other family members, " 2 " indicated twice as many helpings, etc.); time eating dinner; position finishing dinner (relative position in finishing meal in comparison to other family members); bites plus sips per minute; talks per minute; delayed eating index (number of bites plus sips preceded by another bite or sip divided by number of bites plus sips preceded by a placing utensil down). These variables were selected to reflect content of the treatment program. Observer training and procedures. Observers first studied observation manuals and then practiced recording role-played meals, videotapes of actual meals, and live observations of nontreatment families. As a final criterion check, each O rated a standardized videotaped meal. Each O's agreement with criterion ratings on each of the dependent measures reached or exceeded .90 (agreements divided by agreements + disagreements). A table of random numbers was used to determine the nights on which families would be observed and the observer assigned to that family for a given night. Three constraints to complete random assignment were permitted: (1) Observation nights could be rescheduled if the family was not going to eat dinner at home on a given night, (2) Each O was assigned to only one family each night, and (3) If the O was unable to observe on a given night, the next O, as determined by a random numbers table, was assigned. Os were four community college students. They were not familar with obesity research in general and were not informed about the purposes and design of this study. They were instructed that they would be going to homes of overweight adolescents to make selected observations. They were warned of the problems of observer bias and were asked not to question the E or the Ss about the purposes or nature of the study. While unobtrusive reliability checks are usually desirable (O'Leary & Kent, 1973), this was not possible given the demands of observing in the natural environment. Each O was checked weekly in the home by one experimenter and by another O. The random numbers table was used to determine each O's check night and companion O. A 90-min training session was conducted each week to provide feedback to Os regarding inter-observer agreements. Videotaped meals were coded by all Os to assess and reduce drift. Finally, to minimize reactivity and bias, the Os were instructed to minimize communication with the family and the families were asked to

390

COATES AND THORESEN s

p

0

0.-201

~ - C "''r-~" ~ . ~)'-'0



i~ -'s

. 0 C> <}--©--©--.:>--0

20



Sl

1

2

3

4

5

6

7

8

9

10 11 12 13

weeks

FIG. 1. Pounds lost, percent above average weight lost, and reduction index for Subjects I, 2, and 3 across the 14 weeks of the study.

refrain from talking to the Os. The Os had no responsibility for summarizing data or computing reliabilities. The Os were questioned at weekly meetings regarding the experimental procedures. The Os did refrain from communicating with families but did comment on noticeable differences in subjects' behaviors and environments and began to learn about the study from family conversations at dinner. However, the Os revealed no knowledge regarding the experimental design, or subject assignment to treatment conditions. If knowledge regarding the experiment did bias observations, it seems reasonable that these should have biased the data equally across all subjects.

RESULTS Changes in Weight Fig. 1 presents pounds lost, percent overweight lost, and the Reduction Index (weight loss/pounds overweight x initial weight/target weight × 100; Feinstein, 1959). S1 lost 20 pounds in 14 weeks, while $2 lost 11.5 pounds and $3 gained 5 pounds. Weight was stable for Ss 1 and 2 until each was introduced into the program (weeks 3 and 4 respectively). Per-

BEHAVIOR AND W E I G H T CHANGES

391

cent overweight lost and the Reduction Index tend to compensate for the initial differences in weight among the subjects. S 1 lost 16.9~ overweight (RI = 24.17), while $2 lost 10.6% overweight (RI = 22.45). $3 failed to lose weight, despite therapist time and written summaries of the treatment components. She reported, however, that she had been gaining weight prior to joining the program; maintenance may have been a positive therapeutic outcome for her. Until informed of her assignment, $3 did not appear aware of her status as a placebo subject. Neither she nor her family expressed disbelief in the treatment approach, but some frustration at the rate at which weight loss occurred was expressed. The therapist responded by telling her that weight loss would come in time if she stayed with the program.

Generalizability Analysis 2 Agreement between Os (agreement divided by agreement + disagreements) averaged .88 (range = .85 to .92) for foods available and .93 (range = .87 to .97) for eating behaviors. Coates and Thoresen (1978b; see also Cronbach, Gleser, Nanda, & Rajaratnam, 1972; Strossen, Coates, & Thoresen, 1979) conducted a generalizability study of baseline observations from this study to assess percent of variance in observed outcomes that could be attributed to observers, occasions of observation, and differences among subjects. On all variables, differences among observers contributed relatively little variance. On five variables (food displayed in house, refrigerator calorie rating, food service ratio, bites and sips per minute, talks per minute), there was considerable variation due to occasions of observation. Generalizability theory also permits the researcher to determine the relative precision and reliability of obtained scores. Relatively precise estimates were obtained on all EATS variables except one, number of helpings (see Coates & Thoresen, 1978b, Table 2). Changes in Behaviors and Environments Two procedures were used to assess the statistical significance of changes in behaviors and environments across treatment phases. The Autoregressive Integrated Moving Averages (ARIMA) analysis (Glass, Willson, & Gottman, 1975; Jones, Vaught, & Weinrott, 1977) was used to assess changes in level and slope of each time series between baseline and treatment. The ARIMA analysis assesses the degree and kind of serial dependency in the data and transforms the data to remove the dependency. Standard t tests are then performed on the transformed data. Markov Chain analyses (Gottman & Notarius, 1978) were used for those variables in which lack of variability in the data did not make it possible to remove serial dependency. It is possible to describe points in 2 Generalizabilitydata (i.e., reliabilityand validity)are availablein Coates and Thoresen (1978b). Copies of the EATS may be obtained from the first author.

392

COATES

AND

THORESEN

~D t'4

b3~

.1

,<

,< [Z Z 0

r~

Z e~ Z <

r~

Z

c~ ,ff o-; ~

Z

,-1 M

Z r.~. Z 0 <

©

Z .<

© .~

u

.-= ¢.,

Z <

._=

I. N

=~z

z t~

BEHAVIOR AND W E I G H T CHANGES

393

[..

,4,,4,4

,-1 ["-' r~

~ . v v ~

,,I

VV

~ V V

,-; rq ,.,4 4 N M e-,

6 o

=

0

~ r~

,"

~

0

;>.

II © H r..q

~

r.,,q

,,.4 ,,,d

394

COATES

AND

THORESEN

%

E ......

0 0 0

0 0 0

ggg ,..a

"r-

z

z

~

ca ce~

r.)

. . .

> o

~

~

~ - ~

..= 0 z

M

O ta

r,

0

0 0 0

0 0 0

E

[-- < Z

<

~

< >,

Z

e

~a

m

.~._=

e~ <

VVV ~

Z <

Z < m

N

&-

N

I

~a

.~

I

•= ~

~'-

7= .g'g

:~ " g 6

e "g6

~z

g~

g~

>

2e~

e'~ v

g :-: == No

BEHAVIOR AND WEIGHT CHANGES

395

a series as recurrent or transitory, and to compare the probability of going from one point to another during baseline with respective probabilities during treatment. To complete this analysis, the transition matrix during baseline and the transition matrix during treatment were calculated. In this procedure, the baseline transition probabilities are used to generate the expected frequencies in the transition matrix following treatment. Expected and obtained frequencies in the posttreatment series were then compared using the Chi-square (df = 1). The series was considered to have changed following treatment if the observed recurrent state changed and if the Chi-square was statistically significant. The treatment program was correlated with behavioral and environmental changes in each of the treated subjects. However, each subject showed different patterns of change. Tables 2 and 3 present means, standard deviations, and the results of the statistical analyses for each of the dependent measures.

Subjects 1 and 2 S 1 showed significant changes in cue elimination, food storage, food portions, and exercise. Significant and stable reductions were observed in number of places, number of activities, and number of times food was eaten (cue elimination), the caloric ratings of the cupboard, refrigerator, and freezer (food storage), food displayed in the house, average hunger rating (food portions), and minutes of exercise per day. S 1 and her family had difficulty implementing the food service strategy (keeping all foods off the table); some changes were noted but these tended to be variable and inconsistent. S 1 showed no changes in eating behavior or in meal duration. $2 showed significant and stable changes in cue elimination (number of places, activities, and times associated with eating), food service, and eating behaviors. Especially striking, in contrast to S 1, were the reductions in bites + sips per minute, Delayed Eating Index, and meal duration. These are displayed graphically in Fig. 2. Food storage patterns showed statistically significant changes in level from baseline to treatment for $2. As shown in Fig. 3, however, the data show both variability and shifts in trend during treatment (see Fig. 3). Cupboard Caloric Rating showed a significant change in level with the introduction of treatment and a significant change in trend from baseline to treatment. The Refrigerator Caloric Rating and the Freezer Caloric Rating, by contrast, drifted downward in baseline but the trend leveled out during treatment. Subject 3 The procedures of the experiment were partially reactive as $3 also showed some positive behavior changes during the program. The caloric density of foods in the cupboards and refrigerator declined during baseline (cf. Table 2 and Fig. 3) and $3 decreased the number of times per day that she ate (Table 3). The statistical analyses and inspection of the

396

COATES

AND

THORESEN

'i 4

~1

+

"

sl ""

" . ... . . . . . . . .. .

~"

t,/ :&,, VtC', i", I j~....

I'--4



~!

"~

~

.,

~,

• rate per downs irate

minute

per

t

$3

I

i 1o

l 2o



1~ I I I ~o 4o so 6o TREATMENT DAYS

I

70

I

80

FI~. 2. Subject eating behavior as assessed by observers on the Eating Analysis and Treatment Schedule. Triangles indicate the rate of bites + sips per times utensils were placed on the plate;"circles indicate the rate of bites + sips per minute.

data showed that caloric densities were returning to baseline, however (cf. Fig. 2).

Patterns of Change There is some evidence to suggest that behavior changes for S 1 and $2 occurred in areas in which each was most deviant. As shown in Fig. 2, for example, S 1 ate at a moderate rate prior to treatment and showed no apparent need to slow the rate of her eating. However, copious amounts of high calorie foods were available in her household (Fig. 3). Food at meals was plentiful and the portions served by her mother were large. Her father and two brothers consumed these large, high-calorie portions easily and remained at normal weights because they expended considerable energy as self-employed construction workers. The subject was sedentary and ate similar portions. The result was predictable. Limiting the number, times and places of eating, increasing the availability of lower calorie foods, and implementing an exercise program seemed to be necessary for successful short term weight losses. S2's household was quite different; high calorie foods and large portions were not characteristic family habit patterns. But she did eat rapidly and in several locations and was also quite sedentary. Reductions in the rate of eating, number of places where food was eaten, and food portions seemed sufficient to produce weight loss.

DISCUSSION This study might be regarded as an innovative time-series quasi-experimental investigation that demonstrates the feasibility of home obser-

BEHAVIOR AND W E I G H T CHANGES



.--."

/

=

._.

397

. ~. I

\

"

-.

\ 7".

,refrigerator

-• '~

~./'"

-j~'~ ;-. s3

7

~-~...~,

I

r

10

20

I ! ° i 30

40

TREATMENT

icupboard

!

I

I

F

50

60

70

80

DAYS

FIG. 3. Caloric ratings of the refrigerator a n d c u p b o a r d s as a s s e s s e d by o b s e r v e r s on the E a t i n g A n a l y s i s and T r e a t m e n t S c h e d u l e (1 = low calorie, 2 = m e d i u m calorie, 3 = high calorie).

vation procedures and time-series analyses. The conclusions that can be drawn are limited, of course, by the small number of subjects studied (we limited ourselves to 3 because of resources) and the pioneering nature of the instruments. It is not possible to know what subject characteristics or pretreatment variables interacted with components of the treatment program to produce the results observed for each subject. The treatment program was effective in producing short-term weight losses and differential behavior changes in the two treated subjects. While the procedures of the experiment were partially reactive, changes in Ss 1 and 2 cannot be accounted for by assessment alone. $3 and her family were treated similar to Ss 1 and 2 except for participation in the active treatment program. She and her family received the same amount of therapist attention and time, were given printed descriptions of the experimental treatment program, and were subjected to the same intensive observations as Ss 1 and 2. Some behavior changes did occur, but these were relatively short-lived. By contrast, Ss 1 and 2 each showed stable behavior changes in a variety of areas. Finally, $3 lost no weight during the program and actually gained weight at the end. Because this was designed to study the process of behavior change and weight loss, no formal follow-up was planned. S 1, however, continued to visit the therapist bi-weekly and reached 200 at 12 months following treatment. She chose to continue on her own after that, and had reached 185 pounds 4 months later. $2 chose not to continue in treatment, and returned to baseline weight. S3's failure to continue participation was disappointing. During the 10-week placebo treatment period, she appeared increasingly frustrated at her failure to lose. Her parents ques-

398

COATES AND THORESEN

tioned the utility of our treatments. We responded by indicating the weight loss takes time and alternatives would be tried if original strategies were unsuccessful. By the time we approached week 14, however, she expressed the opinion that she did not feel like continuing to try at the present time. S3's failure to continue in treatment may represent what happens to persons in placebo treatments and less successful treatment programs. The decision to lose weight is not omnipresent, and there is some evidence indicating that weight loss early in a program promotes continued weight loss (Jeffery et al., 1978). Given that we have vast evidence on the results of no-treatment and the relative efficacy of different types of treatment (cf. Wing & Jeffery, 1979), it may make little sense to continue subjecting participants to failure experiences. It is possible that previous studies have failed to find relationships between behavior and weight change, not because there is no relationship, but because the behavior being measured is not specific to an individual's eating patterns. Poor compliance to behavioral strategies may result because persons are asked to engage in myriad activities having no relation to the subject's problem. As Brownell (Note 1) pointed out, " . . . subjects might be displaying more wisdom in not following some of our advice than we display in giving it." It might be possible to structure individual and group programs so that basic components could be supplemented with generalized training in behavior analysis and problem solving, and intense individualized behavior analysis could be used to pinpoint problem areas so that cost effective prescriptions could be developed and used. There are variables not assessed here that might contribute, in future studies, to a more complete understanding of the relationships between treatment components and weight loss: calories ingested and expended, and mood states, the emotional associations of food. The instruments and techniques used here might be applied to such assessments before proceeding with treatment for given subjects. Certainly the foundation is laid for basic and clinical studies to examine this proposition further.

REFERENCE NOTE 1. Brownell, K . D .

Personal communication, May 24, 1979.

REFERENCES Bellack, A. S., Rozensky, R., & Schwartz, J. A comparison of two forms of self-monitoring in a behavioral weightreductionprogram. Behavior Therapy, 1974,5, 523-530. Brownell, K. D., Heckerman,C. L., Westlake, R. J., Hayes, S. C., & Monti, P.M. The effect of couples training and partner cooperativeness in the behavioral treatment of obesity. Behaviour Research and Therapy, 1978, 16, 323-333. Brownell, K. D., & Stunkard, A. J. The behavioral treatment of obesity in children. American Journal of Diseases of Children, 1978, 132, 403-412. (a) Brownell, K. D., & Stunkard,A.J. Behaviortherapy and behaviorchange: Uncertainties in programsfor weight control. Behaviour Research and Therapy, 1978, 16, 301. (b)

BEHAVIOR AND WEIGHT CHANGES

399

Coates, T. J., & Thoresen, C.E. Obesity in children and adolescents: A review. American Journal of Public Health, 1978, 68, 143-151. (a) Coates, T. J., & Thoresen, C.E. Using generalizability theory in behavioral observations. Behavior Therapy, 1978, 9, 605-613. (b) Coates, T. J., & Thoresen, C . E . Obesity in children and adolescents: The problem belongs to everyone. In B. Lahey & A. Kazdin (Eds.), Advances in clinical child psychology. New York: Plenum, 1980. Cronbach, L. T., Gleser, G. C., Nanda, H., & Rajaratnam, N. The dependability of behavioral measures. New York: Wiley, 1972. Feinstein, A. R. The measurement of success in weight reduction. Journal of Chronic Diseases, 1959, 10, 439--456. Glass, G. V., Willson, J. L., & Gottman, J. M. Design and analysis of time series experiments. Boulder, CO: Colorado Associated Universities Press, 1975. Gottman, J., & Notarius, C. Sequential analysis of observational data using Markov chains. In T. Kratochwill (Ed.), Strategies to evaluate change in single subject research. New York: Academic Press, 1978. Hagen, R . L . Group therapy versus bibliotherapy in weight reduction. Behavior Therapy, 1974, 5, 222-234. Jeffery, R. W., & Coates, T . J . Why aren't they losing weight? Behavior Therapy, 1978, 9, 856-860. Jeffery, R. W., Wing, R. R., & Stunkard, A.J. Behavioral treatment of obesity: The state of the art, 1976. Behavior Therapy, 1978, 9, 189-199. Jones, R. R., Vaught, R. S., & Weinrott, M . R . Time-series analysis in operant research. Journal of Applied Behavior Analysis, 1977, 10, 151-166. Mahoney, M . J . Self-reward and self-monitoring techniques for weight control. Behavior Therapy, 1974, 5, 48-57. O'Leary, K. D., & Kent, R . N . Behavior modification for social action: Research tactics and problems. In L. A. Hamerlynck, L. C. Hardy, & E. J. Mash (Eds.), Behavior change: Methodology, concepts, and practice. Champaign, IL: Research Press, 1973. Stalonas, P. M., Johnson, W., & Christ, M. Behavior modification for obesity: The evaluation of exercise, contingency management and program adherence. Journal of Consuiting and Clinical Psychology, 1979, 46, 463-469. Strossen, R. J., Coates, T. J., & Thoresen, C. E. Extending generalizability theory to single subject research. Behavior Therapy, 1979, 10, 606-614. Stunkard, A. J. From explanation to action in psychosomatic medicine: The case for obesity. Psychosomatic Medicine, 1975, 37, 195-236. Stunkard, A. J., & Mahoney, M . J . Behavioral treatment of eating disorders. In H. Leitenberg (Ed.), Handbook of behavior modification. New York: Appleton-CenturyCrofts, 1976. United States Department of Health, Education, and Welfare. Height and weight of youths 12-17 years (DHEW Publication No. (HSM) 73-1606). Rockville, MD: U.S. Public Health Service, 1973. Wing, R. R., & Jeffery, R . W . Outpatient treatments of obesity: A comparison of methodology and clinical results. International Journal of Obesity, 1979, 3, 261-279. Wollersheim, J. P. Effectiveness of group therapy based on learning principles in the treatment of overweight women. Journal of Abnormal Psychology, 1970, 76, 462-474. RECEIVED: APRIL 10, 1980; REVISED: JUNE 24, 1980 FINAL ACCEPTANCE:SEPTEMBER4, 1980