‘MSU

LIBRARIES

 

 

RETURNING MATERIALS:
Place in book drop to
remove this checkout from
your record. FINES will
be charged if book is
returned after the date
stamped below.

 

 

 

 

THE ROLE OF PATIENT PREFERENCE FOR TREATMENT TYPE

IN THE MODIFICATION OF WEIGHT LOSS BEHAVIOR

BY

Thomas C. Fuller

A DISSERTATION

Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of

DOCTOR OF PHILOSOPHY

School of Health Education, Counseling Psychology
and Human Performance

1988

550‘837X

ABSTRACT

THE ROLE OF PATIENT PREFERENCE FOR TREATMENT TYPE
IN THE MODIFICATION OF WEIGHT LOSS BEHAVIOR
BY
Thomas C. Fuller

The purpose of this study was to determine the role of
treatment choice in modifying outcomes in weight reduction
programs. Three treatment types (nutritional education,
behavior management, and exercise training) were utilized in
a ten week weight loss program offered to a sample of
overweight adults. The subjects were randomly assigned to
two treatment choice conditions: half received an informed
choice of treatment type, half were randomly assigned to
treatments.

The dependent variables of interest were body weight
change, attrition, attendance, adherence, treatment efficacy
perceptions, program satisfaction and reactance, and self-
efficacy. Research assessments were gathered by
questionnaires at pre-program, mid-program, and post-program

sessions. The major hypotheses involved comparisons between

assigned subjects and subjects receiving a choice of
treatments on the eight outcome variables.

Results indicated that subjects receiving a choice of
treatments had a significantly lower attrition rate and
higher adherence ratings, whereas subjects assigned to
treatments had greater weight loss. There were no
significant differences between subjects choosing treatments
and assigned subjects on attendance, perceptions of
treatment efficacy, program satisfaction, reactance, or
self-efficacy.

Theories of self-efficacy, reactance, and decision
making provided a basis for explaining the results.
Treatment choice was seen as promoting a perception of
restricted freedom, which generated psychological reactance.
Outcomes were consequently undermined for subjects receiving
a treatment choice relative to assigned subjects.

Additional exploratory hypotheses revealed that:

1) desirability of control did not interact significantly
with choice to affect outcomes, 2) weight locus of control
was unrelated to attendance and weight loss, 3) subjects
choosing exercise training had a non-significant tendency
toward internal locus of control relative to subjects
choosing the other treatments, and 4) history of prior
weight loss efforts had a small, significant and negative
relationship with pre-test and mid-program self-efficacy,

mid-program reactance, and adherence.

ACKNOWLEDGEMENTS

By design, a doctoral dissertation is a solo effort.
Conceived, developed, implemented, analyzed, described ---
the dissertation is an accomplishment of the student alone.
In theory, at least. In reality, many people contribute in
many ways throughout the process of conducting and
completing a research study of any size and scope. I owe
much to those "many people" who contributed in so many ways
to this research study, and to my doctoral education. They
are deserving of acknowledgement.

Dr. Bertram Stoffelmayr has played an integral part in
all aspects of this research study. He participated in the
initial conceptalization of this study and arranged for
important logistical support from the Department of
Psychiatry during the weight reduction program. In
addition, he has provided me with ongoing conceptual
stimulation and personal encouragement throughout the
dissertation process.

Dr. Robert Lent also participated in the initial
conceptualization of this study and it was his encouragement
that led me to make it my dissertation. More importantly,

Bob has been ever-available and accessible with his

ii

resourcefulness and advisement on issues ranging from the
empirical to the personal. In Bob I have found a friend and
a model to follow.

It was in Dr. Richard Johnson's research class that I
first explored weight loss research. Dr. Johnson rekindled
my spirit for treatment outcome research --- something few
students want to do anymore. Dr. Johnson made me aware of
the multiple facets of treatment outcome research, and this
made my actual work much easier.

Dr. William Hinds has been a mentor and guide --- a
guardian angel --- throughout my doctoral education. Dr.
Hinds provided me with most of the opportunities that shaped
my professional identity as a health psychologist. He has
trusted me, and allowed me a free-hand in making my
education what I wanted it to be. In the process, he came
through with vital support at critical junctures. It simply
would not have been the same without him.

These four individuals, my doctoral committee, have
respected my abilities and contributed to my integrity by
trusting me to do so much on my own. The quality of my work
has been driven by a desire to uphold their trust. Thus,
they have truly been a "guidance" committee.

Dr. Mary Noell and Tom Custer, both of the Michigan
Dairy Council, provided the informational and logistical
support necessary to deliver the nutritional education

treatments. Susan Dougherty's enthusiasm and interest in

iii

the study made the exercise training treatments a learning
experience for me, as well as for the subjects.

Dr. Elizabeth Becker and Roy Feber were valuable
resources during the statistical analysis of the data. Mr.
David Harley, always the resource person, also provided
statistical guidance. In addition, Dave's tolerance of my
dissertation anxieties helped free-up my occasional
paralyses and set me back on the road to clear thinking.

Dr. Brian Mavis contributed more to this study than can
be mentioned here. His dissertation paved the way for my
research and provided a constant guide for preparing the
final draft of this document. Brian provided support for
this research on all fronts --- finances, contracts,
policies and procedures, program resources, program
materials, and knowledge of delivering weight reduction
programs, were only a few of his contributions. His
personal support has been most meaningful to me. When and
where it counted, Brian was there. The friendship and
comaraderie that has developed between us is beyond measure.

Finally, but by no means least, my wife, Mary, deserves
acknowledgement. Her understanding, patience, and tolerance
of the many ups and downs, and highs and lows of the
dissertation process have pulled me through it all. Mary
was always there, always supportive, and always loving. She
has kept me balanced (emotionally and financially!). In the

lonely process of dissertation research, Mary has taught me

iv

that the best things aren't accomplished alone, but rather
together.

To all of these "many people" who have helped me
accomplish so much --- not alone, but together —-- I owe my

gratitude.

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . .

LIST OF FIGURES . . . . . . . . . . . . . . . . . .

LIST OF NOMENCLATURE AND ABBREVIATIONS . . . . . . .

I.

II.

III.

INTRODUCTION . . . . . . . . . . . . . . . . .

Need for the Study . . . . . . . .
Purpose of the Study . . . .
General Hypotheses of the Study. .
Theory . . . . . . . . . . . . . .
Overview of Remaining Chapters . .

O O O O O
O O O O O
O O O O O
O O O O 0

REVIEW OF THE LITERATURE . . . . . . . . . . .

Definitions and Measures of Body Weight. .
Personal Variables Associated with Weight
Reduction . . . . . . . . . . . . . . .
Weight Reduction Treatments. . . . .
Attrition, Attendance, and Adherence
Patient Choice of Treatment Type . .
Locus of Control in Weight Reduction
The Desirability of Control. . . . .
Self-efficacy. . . . . . . . . . . .
Summary. . . . . . . . . . . . . . .

“mommGY O O O O O O O O O O O O O O O O 0

Population . . . . . . .
Sample . . . . . . . . .
Procedures . . . . . . .
Subject Selection .
Subject Assignment.
Monetary Incentives and Costs
Informed Consent. . . . . .
Weight Loss Treatment Conditions
Treatment Providers . . . . .
Measurement Schedule. . . . .
Debriefing and Follow-up. . .
Instrumentation. . . . . . . . . .

vi

viii
xi

xii

10
11
23

24
24

28
35
53
61
71
80
83
94

100

100
101
102
102
104
106
107
108
111
113
114
114

IV.

V.

Body Weight .
Attendance. .
Attrition . .
Adherence . .
Self-efficacy

Perceptions of Treatment E

Program Satisfaction.
Program Reactance . .

Desirability of Control .
Weight Locus of Control .
Demographic Descriptive an

Variables. . . . . .
Experimental Design. . . .
Research Hypotheses. . . .
Analysis of the Data . . .

RESULTS 0 O O I O O O O O O O

Comparability of the Groups

Testing the Hypotheses . .
Hypothesis One . . . .
Multivariate Analyses.
Hypothesis Two . . . .
Hypothesis Three
Hypothesis Four.
Hypothesis Five.
Hypothesis Six .
Hypothesis Seven
Hypothesis Eight
Hypothesis Nine.
Hypothesis Ten . .
Hypothesis Eleven.
Hypothesis Twelve.
Hypothesis Thirteen.

DIstSION. O O O O O O O O O

APPENDICES

A.

B.

C.

D.

E.

f cacy

[10000000000

0

ooooaoooomooooo
ooooldoooopoooooo
H
m o
On
ooooP-o o
O
oooof’foooooooooo
0

Physical Activity Readiness Questionaire. . .

Description of Treatments and
Subjects . . . . . . . . . .

Consent Form. . . . . . . . .
Outlines of the Treatments. .

Measurement Instruments . . .

BIBLIOGRAPHY . . . . . . . . . . . .

vii

Instructions to

114
115
115
115
116
116
117
118
118
119

120
121
122
128

131

131
137
138
149
153
168
171
175
183
190
195
202
207
209
210
212

215

254

255

266

267

271

281

9.

10.

11.

12.

13.

14.

15.

LIST OF TABLES

Pretest Means and F Ratios by Experimental
Condition. . . . . . . . . . . . . . . . . . .

Chi-Square Tests on Pre-Program Measures by
Experimental Condition . . . . . . . . . . . . .

Chi-Square Analysis of Attrition at the First
Program Meeting, During the Course of the Program,
and Total Program-Related Attrition. . . . . . .

Multiple Comparisons of Attrition Between Weight
Loss Methods During the Course of the Program and
Total Attrition O O O O O O O O O O O O O O O O O

Reasons for Attrition at the First Program Meeting

Reasons for Attrition During the Program. . . .

Means, Standard Deviations, and T-Tests for
Participants Vs. Drop-Outs During the Program. .

Chi-Square Analyses of Participants Vs. Drop-Outs

Multivariate Analysis of Variance on Post-Program
measures O O O O O O O O O O O O O O O O O O O O O

Multivariate Analysis of Variance on Mid-Program
measures O O O O O O O O O O O O O O O O O O O O O

Multivariate Analysis of Variance on Pro-Program
Mea sures O O . O O O O O O O O O O O O O O O O O O

Cell Means of Body Weight Measures at Pre-Test,
Post-Test, and Change at the End of the Program .

Analysis of Variance for Body Weight Change in
Pounds o o o 0 O o o o o o o o o o o o o o o o o 0

Analysis of Variance for Body Mass Index Change. .

Analysis of Variance on Percent Over Ideal Weight
Change 0 o o 0 O o o o o O 0 O 0 O o o o o o o o 0

viii

132

134

140

144

145

145

147

148

151

151

152

155

156

156

157

16.

17.

18.

19.

20.

21.

22.

23.

24.

25.

26.

27.

28.

29.

30.

Summary of Multiple Regression Analysis with Change
in Body Weight (Pounds) as Criterion Variable and
Pre-Weight, Choice Factor, and Treatment Factor as
Predictor Variables . . . . . . . . . . . . . . .

Summary of Multiple Regression Analysis with Change
in Body Mass Index as Criterion Variable and
Pre-BMI, Choice Factor, and Treatment Factor as
Predictor Variables . . . . . . . . . . . . . . .

Summary of Multiple Regression Analysis with Change
in Percent Overweight as Criterion Variable and
Pre-Percent Overweight, Treatment Factor and
Choice Factor as Predictor Variables. . . . . . .

Multiple Comparisons for Treatment Types on Body
Weight Change Measures. . . . . . . . . . . . . .

Analysis of Variance on Attendance . . . . . . . .

Cells Means and Standard Deviations on Attendance
and Adherence . . . . . . . . . . . . . . . . . .

Analysis of Variance on Adherence. . . . . . . . .

Repeated Measures Analysis of Variance on Percep-
tions of Treatment Efficacy Over Time . . . . . .

Cell Means and Standard Deviations of Pre, Mid,
and Post Measures of Perceptions of Treatment
EfficaCYO O O O O O O O O O O O O O O O O O O O O

Multiple Comparisons for Treatment Types on
Perceptions of Treatment Efficacy at Pre-, Mid-,
and Post-Program Measures . . . . . . . . . . . .

Repeated Measures Analysis of Variance on Program
Satisfaction Over Time. . . . . . . . . . . . . .

Cell Means and Standard Deviations for Mid and
Post Measures of Satisfaction and Reactance . . .

Multiple Comparisons Between Treatment Types on
Mid- and Post—Measures of Program Satisfaction. .

Repeated Measures Analysis of Variance on Program
Reactance Over Time . . . . . . . . . . . . . . .

Repeated Measures Analysis of Variance on Self-
Efficacy Over Time. . . . . . . . . . . . . . . .

ix

162

163

164

166

169

169

172

176

177

180

185

186

189

192

197

31.

32.

33.

34.

35.

36.

37.

Cell Means and Standard Deviations of Pre-, Mid,
and Post-Measures of Self-Efficacy. . . . . . . .

Multivariate Analysis of Variance for the Post-
Program Measures With Choice, Treatment Type, and
Desirability of Control as Factors. . . . . . . .

Multivariate Analysis of Variance for the Mid-
Program Measures With Choice, Treatment Type, and
Desirability of Control as Factors. . . . . . . .

Multivariate Analysis of Variance for the Pre-
Program Measures With Choice, Treatment Type, and
Desirability of Control as Factors. . . . . . . .

Comparisons of Subjects Choosing Exercise Vs.
Subjects Choosing Behavior Modification or
Nutrition Education on Weight Locus of Control. .

Correlations of Weight Locus of Control With
Attendance and Weight Change Measures . . . . . .

Correlations of Frequency and Variety of Previous
Weight Loss Attempts With Outcome Measures. . . .

198

204

205

206

208

208

213

10.

11.

12.

13.

LIST OF FIGURES

Mean Program Attrition by Condition. . . . . . . .

Mean Weight in Pounds at Pre- and Post—Program

for Choice Vs. Assigned Groups. . . . . . . . . .

Mean Weight in Pounds at Pre- and Post-Program

Measures for Treatment Types. . . . . . . . . . .
Mean Attendance by Condition . . . . . . . . . . .

Mean Adherence Ratings by Condition. . . . . . . .

Mean Scores on Perceptions of Treatment Efficacy

Over Time for Choice Vs. Assigned Subjects. . . .

Mean Scores on Perceptions of Treatment Efficacy

Over Time for Treatment Types . . . . . . . . . .

Mean Program Satisfaction Over Time for Choice
Vs. Assigned Subjects . . . . . . . . . . . . . .

Mean Program Satisfaction Over Time for Treatment
Types O O O O O O O O O O O O O O O O O O O O O O

Mean Program Reactance Over Time for Choice Vs.
Assigned Subjects . . . . . . . . . . . . . . . .

Mean Program Reactance Over Time for Treatment
Types O O O O O O O O O O O O O O O O O O O O O O

Mean Self-Efficacy Scores Over Time for Choice
Vs. Assigned Subjects . . . . . . . . . . . . . .

Mean Self-Efficacy Scores Over Time for Treatment
Types O O O O O O O O O O O O O O O O O O O O O O

xi

143

158

167

170

173

178

179

187

188

193

194

199

200

LIST OF NOMENCLATURE AND ABBREVIATIONS

Throughout the text, tables, and figures of this
document several terms are used with sufficient frequency to
warrant their abbreviation. These terms and abbreviations

are presented below to assist the reader.

Choice faotoz, choice oonditioo, fiooto;_A:

The terms "choice factor", "choice condition", and
"factor A” refer to the primary independent variable
of the current study. This factor has two levels:
choice of treatment type and random assignment to
treatment type.

Treatment_fa2f2r. LIEQEEEBE_§YR§: I§E§2£_Ez

The terms "treatment factor", "treatment type" and
"factor B" refer to the secondary independent variable
of the current study. This variable has three levels:
nutritional education, behavior management, and
exercise training.

NE : The abbreviation "NE" refers to level one of factor B,
the nutritional education treatment.

on : The abbreviation ”BM“ refers to level two of factor B,
the behavior management treatment.

The abbreviation "ET" refers to level three of factor
B, the exercise training treatment.

If:

B I : The abbreviation "BMI" refers to Body Mass Index, one
of the dependent measures of body weight.

xii

CHAPTER I

INTRODUCTION

In recent years, psychologists have been playing an
increasing role in the study and treatment of health and
illness within the context of medical settings (Gatchel and
Baum, 1983). In particular, psychologists have devoted
considerable attention to enhancing treatment efficacy for
alcoholism, drug abuse, smoking, and obesity (Gatchel and
Baum, 1983). The present study and discussion is offered as
a contribution to the psychologist's role as service
provider in the treatment of overweight and obesity.

The National Health and Nutrition Examination Survey I
(NHANESI), conducted between 1971 - 1974, found that 18.4%
of men and 24.1% of women between the ages of 20 - 74 were
20% or more above their ideal weights. The National Health
and Nutrition Examination Survey II (NHANESII), conducted
between 1976 - 1980, found that similar proportions of the
United States population were 20% or more above their ideal
weights. Using somewhat different criteria, the National
Center for Health Statistics found that 19.3% of males and
27.6% of females were obese (Abraham, 1983). Similar

findings have been reported by VanItallie and Abraham

(1985), Piziak (1983), Brownell (1982), and Simopoulos
(1985). In general, reported findings suggest that between
30 million and 40 million Americans are overweight (as
defined by the 20% over ideal weight criteria). In response
to these population overweight statistics, the United States
Public Health Service has recommended that by 1990 the
prevalence of overweight in the U.S. adult population should
be decreased to 10% of men and 17% of women, and that 50% of
the current overweight population should have adopted weight
loss regimens (Dwyer, 1985).

Recommendations for weight loss amongst obese persons
are based upon more than a popular image of physical
attractiveness. There are serious health risks associated
with having a greater than "normal" degree of body fat.

Body weight of 120% or more than desirable weight (as
defined by the 1983 Metropolitan Life Insurance Tables) is
significantly associated with hypertension,
hypercholesterolemia, hypertriglyceridemia, and
hyperlipidemia --- all risk factors for coronary heart
disease (Berchtold, Berger, Greiser, Dohse, Irmscher, Gries,
and Zimmerman, 1977; Blackburn and Read, 1984; Ducimetiere,
Richard, and Cambien, 1986; Piziak, 1983; Truswell, 1985).
A body weight greater than 20% over ideal weight increases
the risks for developing adult onset diabetes, pulmonary
problems, renal problems, cerebrovascular disease,

gallstones, arthritis, and even an increased incidence of

accidents (Brownell, 1982; Burton, Foster, and VanItallie,
1985; Larsson and Bjorntorp, 1981; Truswell, 1985; Van
Itallie and Abraham, 1985). Surgical risks, complications of
pregnancy, and increased incidence of cancer of the breast
and colon have also been associated with a body weight of
20% or more over ideal weight (Blackburn and Read, 1984:
Brownell, 1982; Truswell, 1985). Further, Brownell (1982)
and Harris (1983) have documented social hazards associated
with obesity, including discrimination, blame, rejection,
and negative attitudes of others. Finally, the mortality
ratio (number of deaths observed/number of deaths expected)
increases with degree of overweight (Burton et al., 1985;
Sankey, 1984; Bray, 1979; VanItallie and Abraham, 1985).
Sankey (1984), reporting data from the 1979 Build and Blood
Pressure Study, has said that actual mortality in overweight
individuals is significantly more than what should be
expected for the following causes: diabetes mellitus,
cerebrovascular disease, coronary artery disease, and
hypertensive heart disease. Thus, being overweight or obese
is associated with sufficient risks to justify clinical
intervention (Burton et al.,1985).

A related issue is the growing trend towards
consumerism and movement away from the traditional medical
model practice of professionally prescribed treatments
(Krantz, Baum, and Wideman, 1980). Several guidebooks have

been published in the past ten years encouraging patients to

become more active and informed participants in the health
care process generally, and the weight loss process
specifically (Vickery and Fries, 1976; Schaller and Carroll,
1976; Miller and Schildkraut, 1983: Consumer's Union of the
United States, 1980). This trend has both positive and
problematic aspects.

Research has begun to demonstrate that medical outcomes
(such as weight loss) can be determined, in part, by the
patient's degree of participation and perceived control in
the health care process (Krantz et al., 1980; Marston,
1970). For some patients, increased participation and
perceived control enhances treatment outcomes (Krantz et
al., 1980; Averill, 1973; Langer and Rodin, 1976). The
relationship between participation, control, and outcomes
is, however, complex. Research has been initiated to assess
the effects of providing choice and control over some
aspects of treatment in medical settings, yet relatively
little is known about the effects of treatment choice on
treatment outcomes (Krantz and Deckel, 1983; Krantz et al.,
1980).

Additionally, while most weight loss treatments are
generally effective in producing weight loss, all treatments
are not equally effective with all individuals (Bray, 1979).
At the present time there are no recognized methods for a
priori prediction of which overweight individuals will be

most successfully treated with which weight loss strategies.

Bray (1979) has suggested that the highest priority in
weight loss research is developing techniques to predict and
enhance patient-treatment matching so that both
professionals and consumers can be guided in the task of
making accurate differential diagnoses and treatment
selections.

MM

Given the prevalance of overweight in the United
States, the substantial health risks associated with being
overweight, and the inadequate knowledge base regarding
differential prescription of weight loss treatments,
research into the determinants of successful patient-
treatment matching is needed. Further, the growing trend
toward consumerism in the health care industry means that
overweight people are choosing their own treatments. The
aspect of patient choice of weight loss strategy needs to be
investigated.

A number of comprehensive reviews of weight loss
treatments are available in the literature (Brownell, 1982,
1985: Wilson, 1980, 1985: Jeffery, Wing, and Stunkard, 1978:
Stunkard and Penick, 1979: Stunkard, 1975, 1978: Wing and
Jeffery, 1979: Bellack, 1975: Hirsch, 1985a: Dwyer, 1985:
VanItallie, 1978; Blondell, 1984; Porcello, 1985). Based
upon these reviews, several major weight loss treatments are
currently in use in the United States: surgery, jaw-

fixation, medication, psychotherapy, support groups,

exercise, dieting and nutritional management, and behavior
modification. While surgical interventions are still in use
for the severely obese, the use of diet pills has decreased
in recent years due to negative side-effects associated with
the medications used (Bray, 1979). Porcello (1985) reviewed
forty-two professional weight control programs and reported
that all professional programs offer some combination of
nutritional, behavioral, and exercise interventions. The
most common forms of weight reduction treatments are
behavior modification, nutritional management, exercise, and
social support (Brownell, 1982, 1985; Jeffery et al.,1978:
Stunkard, 1975, 1978).

The most significant problems plaguing weight loss
programs have been attrition and adherence (Wing and
Jeffery, 1979: Wilson, 1980, 1985). The attrition rate in
medically based weight loss programs has traditionally been
quite high, often up to 80% (Wilson, 1985). Since
attendance is significantly linked to success at weight
loss, attrition is a significant problem (Jeffery, Bjornson-
Benson, Rosenthal, Lindquist, Kurth, and Johnson, 1984). The
use of behavior modification principles in weight loss
programs has substantially and reliably reduced attrition
rates to an average of 13% or less (Wilson, 1980, 1985:
Stunkard et al., 1978; Wing and Jeffery, 1979).
Contingency-based monetary incentive schemes have proven

useful in increasing attendance in weight loss programs

(Perri, McAdoo, McAllister, Lauer, and Yancey, 1986; Wilson,
1985: Mahoney, 1974: Mavis, 1987). Adherence to prescribed
treatment methods is significantly associated with
successful weight loss (Sandifer and Buchanan, 1983: Wilson,
1985). Unfortunately, little work has been done in the area
of increasing subject adherence to treatment behaviors I
(Wilson, 1985). Wing and Jeffery (1979) have suggested that
finding means to increase adherence in weight control
programs is a matter of high priority.

One method to increase both attendance and adherence
(as well as other outcome variables, such as treatment
satisfaction, self-efficacy, and absolute weight loss) may
be to offer patients a choice of effective treatments when
more than one treatment exists. Enhancing patient
partiCipation and perceived control through offering a
choice of treatment (or some aspect of treatment) has been
found to enhance medical outcomes with stroke and heart
attack victims (Krantz et al., 1980: Krantz and Deckel,
1983), and also in geriatric residential programs (Langer
and Rodin, 1976). Giving patients a choice of treatments
has also been found to be successful in reducing fear levels
in snake phobic Subjects (Devine and Fernald, 1973).
Patient choice of treatment strategy has been most widely
used in alcohol treatment studies, where freedom to choose
treatments has been associated with increased attendance,

compliance, satisfaction with treatment, and reductions in

drinking (Thornton, Gottheil, Gellens, and Alterman, 1977:
Vanicelli, 1979; Sanchez-Craig, 1980; Costello, 1975:
Kissin, Platz, and Su, 1971; Parker, Winstead, and Willi,
1979a, 1979b: Miller and Hester, 1986).

Giving patients a choice of treatment has not been
adequately tested in weight loss programs. Bjorvell, Edman,
Rossner, and Schalling (1985) offered a group of overweight
Swedes a choice between in-patient behavior therapy and
surgical jaw-fixation. Unfortunately, this study did not
involve a control group of overweight subjects who did not
receive a choice of treatments (ie, a randomly assigned
group). Further, the Bjorvell et a1 (1985) study did not
adequately assess differences in weight loss, but rather
focused on personality differences between subjects choosing
the two forms of treatment. Murray (1976) offered a group
of overweight women a choice between self-control training
and a support group as strategies for weight loss and failed
to find a significant effect for choice on weight loss.
Murray's (1976) experimental design was quite poor however:
a close analysis of his methods reveals that subjects were
given a choice of "time" of treatment rather than a choice
of "type" of treatment, and that this difference was
confounded in his analysis and discussion. Further,
Murray's (1976) study included only twelve subjects, a
sample far too small to detect possible significant effects.

No other studies offering patients a choice of treatments in

the area of weight loss could be located. Thus, given the
beneficial effects of choice-of—treatment that are beginning
to appear in other areas of health care research, and the
paucity of such studies in the weight control literature, an
investigation into the role of treatment choice for weight
loss efforts would be worth pursuing.

The present study will contribute knowledge to the
weight loss literature regarding treatment preferences,
participation, choice, and perceived control in the
treatment of the overweight. This study will contribute to
the knowledge base of obesity treatment research by
providing an empirical analysis of the role of choice-of-
treatment in a weight loss program.

W

Since there are a number of weight loss strategies of
apparently equivalent effectiveness available to the
overweight consumer of weight reduction programs, and since
overweight individuals respond differentially (but by
unknown parameters) to these treatments, the purpose of this
study is to investigate the role of giving overweight
participants in a weight reduction program an informed
choice of non-medical treatments. Specifically, the purpose
of this study is to investigate the effect of choice of
treatment type on attrition, weight loss, program

attendance, treatment adherence, program satisfaction,

10

program reactance, perceptions of treatment efficacy, and
perceptions of self-efficacy in weight loss efforts.
Genera; Hypogheses of the Study

The primary hypotheses of this study revolve around the
effects on treatment outcome derived from giving overweight
subjects a choice of either exercise, nutritional education,
or behavior modification as treatment strategies for losing
weight. A second level of hypotheses are exploratory,
regarding the roles of two variables of perceptions of
personal control (desirability of control and locus of
control) in enhancing predictions of treatment outcomes when
choice of treatment type is given or withheld. An
additional exploratory hypothesis regarding the role of past
efforts at losing weight is also examined. The third level
of hypothesis regards differences between the three
treatment types offered. In general terms, the research
hypotheses are as follows:

1. Having a choice of exercise, nutritional education, or
behavior modification as weight loss strategies will affect
the following treatment outcomes in a desirable direction:

a. program attrition

b. weight lost in the program

c. attendance at program meetings

d. adherence to weight loss strategies prescribed
within the chosen treatment type

e. perceptions of the chosen treatment's efficacy

11

f. perceptions of self-efficacy for one's ability to
maintain the use of weight loss strategies in
weight-relevant stressful situations

g. satisfaction with treatment

h. reactance to treatment.

2. There will be no significant differences in outcome
variables between the three treatment strategies offered
other than those attributable to the effects of having a
choice.

3. The locus of control and the desirability of control will
interact meaningfully with the condition of choice to
enhance prediction of outcomes.

11129:!

The regulation of body fat and obesity occurs by two
major influences: 1) the biochemical system of the body
which tends to store fat in adipose tissue, a result of
adaptive evolutionary forces which created a human body
efficient in energy storage, and 2) the psychosocial and
environmental influences which lead us to aquire and store
quantities of energy, including food-seeking behaviors, the
availability and attractiveness of food, eating behaviors,
and all events and behaviors which lead to the expenditure
of energy (Hirsch, 1985a). Obesity or overweight are
conditions which result from a larger than average number of
adipose cells (hyperplasia), or an enlargement of existing

cells (hypertrophy) which may be average in number, or both.

12

Some evidence suggests that juvenile-onset obesity results
from a genetically determined excess of adipose cells, while
adult-onset obesity results from enlargement of adipose
cells in a body with a "normal” number of these cells
(Brownell, 1982). The distinction between hyperplastic and
hypertrophic obesity may be regulated by other factors as
well, and the difference between juvenile-onset and adult-
onset obesity is not always due to hyperplasia (Brownell,
1982; Hirsch, 1985a; Piziak, 1983). It should be noted that
adipose cell size can be increased or decreased, but that
cell number can only be increased and not decreased (the
exception being the surgical removal of adipose tissue).
Further, as Brownell (1982) has discussed, the size of fat
cells may set the biological limit to weight gain and loss,
but the number of fat cells may set the weight at which
these limits occur.

Hirsch (1985b) reviews evidence indicating that body
weight, in the general population, is essentially normally
distributed, but with a marked skewing toward upper levels
of body weight. The explanation for this, Hirsch (1985b)
suggests, is that there exists a subset of people who
control their body weight around a higher weight than the
average weight for an entire population. Keesey (1980) is
most commonly associated with suggesting that body weight,
like other physiological dimensions, is intrinsically

regulated around a "set-point".

13

A body weight set-point is generally considered to be a
product of metabolic adaptive mechanisms, probably governed
by the hypothalamus (Piziak, 1983: Hirsch, 1985a, 1985b).
That is, the hypothalamus modulates appetite and satiety to
preserve a biological body weight reference point in
response to short-term underfeeding or overfeeding (Piziak,
1983). Piziak (1983) has noted, however, that gradual
changes in energy intake/output may alter the biological
set-point for body weight regulation, as in the gradual
increases in body fat associated with normal aging.

Hirsch (1985a) has proposed a theory of body weight
regulation, based upon a model of homeostasis. Weight
change (in either direction) is said to meet resistive and
restorative forces which oppose alteration in body weight.
The human body possesses regulatory systems that work to
maintain an equilibrium. Regulatory forces in this process
might be dietary thermogenesis, carbohydrate tolerance, or
adipose cell size, as well as cognitive, affective, and
social influences. Hirsch (1985a) suggests that obesity
results from regulatory defects which reduce the resistive
and restorative forces. Thus, obesity represents an
adaptive re-equilibration, at a higher weight level, in
compensation for regulatory defects. Weight loss
treatments, then, work to repair regulatory defects,
restoring equilibrium at a lower weight level (Hirsch,

1985a).

14

The question remains as to whether the principle source
of regulatory defects in human obesity is biological or
psychological and environmental. While biological factors
clearly are intimately involved in body weight regulation,
Piziak (1983) suggests that only 1% - 2% of obese patients
are overweight due to endocrine dysfunctions, hypothalamic
tumors, or hereditary disorders. Both Piziak (1983) and
Hirsch (1985a) agree that for obesity to occur there must
ultimately be an imbalance between caloric energy input and
caloric energy output. This suggests that inappropriate
behavior is always a requisite for producing and maintaining
obesity.

The behavioral conceptualization of body weight
regulation rests upon the assumption that energy input and
output are forms of behavior, and as such have been
established over many years, and are under the control of
antecedent and consequent stimuli (Blundell, 1984: Bellack,
1975; Wilson, 1980). That is, levels of eating behavior or
activity levels exist as habitual response patterns and
always occur in the presence of particular stimuli and
consequences. Viewing body weight regulation as a learned
mechanism (at least in part) means that learning based
change strategies can be developed to alter habitual styles
of body weight regulation (Wilson, 1980). Nutritional
management, exercise regimens, and behavior modification are

three primary strategies for altering body weight and its

15

regulation based upon a model incorporating behavioral
assumptions (Stalonas, Johnson, and Crist, 1978: Coates,
1977; Wing and Jeffery, 1979). As Coates (1977) has stated:
"one person's (weight problem) may be maintained by
ignorance of nutrition, another individual may be extremely
sedentary, while the third may be under inappropriate
stimulus control..." (p.99). In a sense, then, there may
well be different obesities that will respond differently to
different types of treatment methods (Wilson, 1980). Within
a broadly behavioral conceptualization of body weight
regulation the issue then becomes one of differential
diagnosis and treatment choice.

In a review of the literature regarding patient
autonomy and choice of treatment in the rehabilitation of
alcoholism, Parker et al. (1979a) found that no single
treatment was uniformly effective and that all treatment
approaches appear to be generally helpful. Parker et al.
(1979a) conclude that a substantial amount of evidence
supports the use of an ecclectic approach to alcohol
rehabilitation. Parker et al. (1979b) suggest that the need
to engage the patient in treatment, the need to provide a
treatment to which the patient can respond, and the need to
reduce attrition (because attendance predicts success), all
indicate the desirability of offering the patient a choice
of effective treatments. Much the same situation is true

for the area of weight control. The general population of

16

obese persons is heterogeneous with regard to the etiology
of obesity, the nature of obesity, personality variables,
and past experience with weight loss efforts (Wilson, 1985;
Weiss, 1977: Bray, 1979). Further, the three major weight
loss treatments (nutritional management, exercise, and
behavior modification) are all more or less equivalently
effective, yet individuals respond differentially to these
treatments (Wing and Jeffery, 1979: Miller and Sims, 1981:
Porcello, 1985). In concert with the Parker et al. (1979a,
1979b) findings from the alcohol rehabilitation literature,
it would stand to reason that in weight control treatment,
offering patients a choice of treatment type would result in
a greater overall success rate and a clearer picture of
important differential diagnostic variables.

In a review of the factor of personal control in
relationship to stress and coping, Averill (1973) identified
three distinct types of control: 1) behavioral (involving
direct action upon the environment), 2) cognitive
(interpretation of events and information), and 3)
decisional (having a choice among alternative courses of
action). While all three forms of control are relevant to
weight regulation, it is the decisional aspects of control
that are pertinent to the present discussion. Decision
theory has a rich tradition within the field of psychology
and health care (Janis and Mann, 1977; Mann and Janis, 1982;

Orford, 1985; Permuter and Monty, 1979: Brehm and Brehm,

17

1981). For an extensive review of this subject the reader
is referred to Permuter and Monty (1979) and Orford (1985).

With regard to weight loss (and other health
behaviors), Janis and Mann's decision theory addresses
primarily the issue of initial decisions to change a
behavior --- for example, the decision to quit smoking, the
decision to quit drinking, the decision to lose weight ---
and focuses on the psychological stress of decisional
conflict and how it influences initial and subsequent
choices to engage in a behavior (Janis and Mann, 1977:
Orford, 1985). As Marlatt and Gordon (1985) have commented,
this makes Janis and Mann's model a theory of decision
making rather than of choice behavior. In other words,
Janis and Mann's decision theory relates more directly to
initial decisions to change a problem behavior than it does
to choices regarding which change strategy to use ("should I
try to lose weight" versus "should I choose dieting,
exercise or behavior modification to lose weight").

Mann and Janis (1982) do, however, discuss cognitive
processes within the context of choice that may be applied
to choices between treatments --- specifically, cognitive
reattributions following choice. Mann and Janis (1982) have
identified five decisional coping patterns: 1) unconflicted
adherence (continuing a present course of action), 2)
unconflicted change (making an uncritical change to a new

course of action), 3) defensive avoidance (a form of escape

18

via procrastination, rationalizations, or shifting
responsibility to others), 4) hypervigilance (impulsive
searching for, and seizure of, a new alternative), and 5)
vigilance (a careful appraisal of alternatives followed by a
rational choice). Of particular interest to the purposes of
the present study are the mechanisms of defensive avoidance:
procrastination, projecting responsibility, and rationalized
"bolstering" of alternatives. Bolstering is a cognitive
dissonance-relieving process that involves distorting the
value of a decision's outcome, or the probability of the
outcome (Mann and Janis, 1982). Specifically, bolstering
can involve exaggeration (playing up the value of a choice),
minimization (playing down the losses associated with a
choice), or denying the long range negative outcomes of a
choice while focusing on the immediate positive outcomes
(Mann and Janis, 1982). This conceptualization of
decisional process is highly similar to the concept of
reactance, as proposed by Brehm and Brehm (1981).

In discussing choice and perceived control in behavior
therapy, Perlmuter and Monty (1979) suggested that providing
patients a choice among alternative therapeutic
prescriptions may act to reduce the psychological reactance
associated with treatment. The theory of psychological
reactance holds that a loss of (or threat to) freedom will
motivate the person to attempt to regain that freedom

through behavioral or cognitive mechanisms (Brehm and Brehm,

19

1981). An integration of Mann and Janis (1982) and Brehm
and Brehm (1981) yields hypotheses relevant to the
decisional situation contained within the present study.
When an individual chooses or is assigned to one treatment
condition, and he/she is aware that alternative treatments
exist, psychological reactance may occur if the individual
perceives a loss of freedom. An individual may reactively
respond to the perceived loss of freedom by 1) attempting to
engage in an eliminated behavior (eg, a member of a
nutrional education group may start exercising), 2)
experiencing reduced satisfaction with the treatment
strategy he/she is being trained in and an increased desire
for lost options, or 3) experiencing negative feelings
toward the person perceived as responsible for the loss of
freedom (Brehm and Brehm, 1981: Mann and Janis, 1982).

Brehm and Brehm (1981) suggest that reactance is aroused
maximally when freedom is eliminated altogether, indicating
that reactance should be stronger, in the present study, for
subjects randomly assigned to treatments. However, three
other factors may mediate this conclusion. Firstly,
reactance may well be stronger for individuals who choose a
treatment once they begin to experience the consequences of
their choice (eg, "exercise is harder than I thought, I wish
I had chosen nutritional education"). Secondly, other
cognitive mechanisms, particularly bolstering, may

counteract reactance effects (Mann and Janis, 1982).

20

Thirdly, individuals may differ in their relative desire for
control over situations (Burger and Cooper, 1979). Burger
and Cooper (1979) have, in fact, demonstrated that
individuals low in the desire for control actually prefer
that decisions are made for them by others, while
individuals high in the desire for control prefer to make
their own decisions. Thus, while reactance is a relevant
variable in a study of treatment choice as a mediator of
outcomes, a priori prediction of decisional reactance based
upon theory is less than straightforward. The present study
will investigate the relationship between choice, reactance,
and the desirability of control in relation to treatment
outcomes.

Rosenstock's (1966) Health Belief Model represents an
attempt to incorporate decision theory into the study of
health and illness behaviors. Both Janis and Mann's (1977)
decision theory and the Health Belief Model assume that
people base decisions upon values and subjective
probabilities (Sjoberg, 1985). Value domains relevant to
health care decisions might include ideas about health,
illness, and the consequences of accepting or refusing
treatment: relevant areas of subjective probabilities would
include perceptions of treatment efficacy and perceptions of
one's ability to comply with treatment (Sjoberg, 1985). The
Health Belief Model holds that compliance/adherence with

treatment is affected by two factors: 1) the patient's

21

perceptions of whether or not he/she is ill, and the degree
and severity of the illness, and 2) the patient's
perceptions of the efficacy of the treatment (Rosenstock,
1966: Gatchel and Baum, 1983). The present study of
treatment choice in weight reduction may provide an
assessment of the second component of the Health Belief
Model. That is, having a choice of weight loss strategy may
affect perceptions of the efficacy of the treatment which
may, in turn, affect adherence with treatment prescriptions.
A final theoretical issue pertinent to a study on the
effects of treatment choice in weight reduction programs is
whether short-term or long-term weight losses are of most
importance (Brownell, 1982). Some major reviews of weight
loss treatments suggest that it is the long-term maintenance
of weight loss that we should be attending to in current
empirical studies (Jeffery, Wing, and Stunkard, 1978: Wing
and Jeffery, 1979: Stunkard and Penick, 1979). Clearly, the
long-term maintenance of weight loss is of the utmost
importance to the overweight patient, and from a practical
standpoint it is also highly important to health care
professionals. Brownell (1982), Colvin and Crist (1983),
and Coates (1977) have all questioned, however, whether the
maintenance of weight loss is really the most important
theoretical issue at the present stage of obesity treatment
research. Colvin and Crist (1983) strongly state that long-

term follow-up studies on obesity treatments are almost

22

useless, for two important reasons: 1) we don't yet have a
truly powerful weight loss treatment to begin with, and 2)
we have inadequate data collection devices to draw causal
relationships over periods of time ranging from six months
to five years (which are reasonable time periods for
asessing weight loss maintenance). Colvin and Crist (1983),
as well as Coates (1977), suggest that current research
should be directed toward subject-treatment assignments and
flexibility in program design. Brownell (1982) makes
perhaps the strongest case for devoting effort toward short-
term studies that investigate new innovations. Brownell
(1982) lists six problems with requiring researchers to
conduct long-term follow-ups in weight loss research: 1)
long-term studies ignore the liklihood that very different
processes influence initial change and the maintenance of
change, 2) innovative and effective short-term approaches
will be ignored if maintenance does not occur, 3) causal
inference simply may not be possible after long periods of
time due to numerous uncontrolled intervening variables, 4)
researchers should not be required to do long and costly
follow-ups of new procedures if short-term results are poor,
5) mandatory long-term study may discourage researchers from
studying obesity, and 6) a focus on maintenance assumes that
the treatment losses obtained are worth maintaining, and

often times they are not. Brownell (1982) suggests that

23

research energies can be well spent in studies that attempt
to maximize the effectiveness of current treatment methods.
ovozyiew of Remaining ghaooegs

In Chapter II, the relevant literature is reviewed in
the following areas: non-medical weight loss treatments,
choice of treatments, the locus of control, the desirability
of control, self-efficacy, attendance and adherence issues,
personal and demographic variables associated with weight
reduction efforts, and measures and definitions of body
weight and weight change. The research design and
procedures are presented in Chapter III. In Chapter IV, the
results and analyses of results are presented. Summary and
conclusions of the research are presented in Chapter V,

along with recommendations for future research.

CHAPTER II

REVIEW OF THE LITERATURE

In this chapter, the following areas of relevant theory
and research are reviewed: definitions and measurements of
body weight and weight change, personal and demographic
variables associated with body weight and weight reduction,
non-medical treatments for weight reduction, attendance and
adherence issues in weight reduction programs, patient
choice of treatments, locus of control in weight loss
research, desirability of control, and self-efficacy
expectations in weight reduction.

W ht

The terms overweight and obese refer to the situation
when fat tissues make up a greater than ”normal" fraction of
total body weight (Bray, 1978). In the average (and normal
weight) American male, 15% - 20% of body weight is fat
tissue; in the average American female, 20% - 25% of body
weight is fat tissue (Bray, 1978: Blundell, 1984). While
there is a normative increase in body fat with age, obesity
or overweight is generally defined as a body fat content

(ie, percent of body consisting of fat) greater than 25% for

24

25

men and greater than 30% for women (Bray, 1978; Blundell,
1984). It should be noted that obtaining a physiologically
precise measure of percent body fat is quite difficult
without sophisticated biophysical or biochemical equipment.
A number of more easily attainable indices of body weight
have been developed, and will be discussed below. A
technical distinction has been made between the terms
"obesity" and "overweight" (Bray, 1978, 1979). Obesity
technically refers to a surplus of body fat, whereas
overweight refers to an excess of body weight relative to
standards of height as indexed in actuarial tables (Bray,
1978, 1979). For practical purposes, the two terms may be
used synonymously and interchangeably.

Colliver, Frank, and Frank (1983) conducted a
correlational and factor analytic analysis of six weight
indices used in the obesity literature. Based upon measures
of 951 adults who were 20% or more above their ideal weights
(relative to the 1959 Metropolitan Life Tables), Colliver et
al. (1983) examined the following measures of body weight:

1)relative weight - ratio of observed weight to ideal

weight

2)weight-height ratio - weight divided by height

3)body mass index - weight divided by height-squared

4)ponderal index - weight divided by height-cubed
5)Sheldon's ponderal index - height divided by the
cubed root of weight

26

6)Benn's index - weight divided by height, where height
is raised to a power based upon a
regresion coefficient of weight and
height relative to mean population
values.
Colliver et al. (1983) found that the mean intercorrelation
between these weight indices was r = .96, with a range of
.87 to 1.0. Factor analyses revealed that all six indices
measure the same factor (obesity), with this factor
accounting for 97% of the variance between the six indices.
Colliver et al. (1983) concluded that while the first three
indices were the easiest to construct and interpret, it
makes little difference from an empirical standpoint which
of the indices is used.

In a comprehensive review of weight reduction treatment
studies, Brownell (1982) discussed six commonly reported
measures of weight change. Brownell (1982) concluded that
absolute weight (measured in pounds or kilograms) was easy
to understand, but did not account for individual
differences in height, frame size, or degree of obesity:
body mass index relates more closely to actual degree of
body fat: the ponderal index is not easily understood: a
weight reduction quotient is generally confusing to
interpret: categorical weight loss (ie, listing subjects who
have lost 20, 30, 40 pounds, etc.) is subject to the same

limitations as absolute weight measures; and percent-

27

overweight is easily understood and accounts for frame size,
but does not reflect percentage of body fat. Brownell
(1982) recommends using several indices of weight change in
reported results, specifically absolute weight, percent
overweight, and body mass index.

The body mass index (weight divided by height-squared)
is considered to be the best single indicator of body fat
and is widely recommended as a useful measure (Burton,
Foster, VanItallie, 1985; Bray, 1979: Annual report of the
National Institutes of Health program in biomedical and
behavioral nutrition research and training, FY1982, 1985).
Burton et al. (1985) provide an excellent discussion of the
body mass index and a useful nomogram for determining body
mass index (BMI) based upon the 1983 Metropolitan tables.
Adult American men have an ideal weight (relative to height,
based upon the 1983 Metropolitan tables) that is represented
by a BMI of 22.7. A BMI of 27.2 represents an adult male
who is 20% over his ideal weight, whereas a BMI of 31.8
represents a body that is 40% over its ideal weight. By the
same criteria, adult American women have an ideal weight
represented by a BMI of 22.4, a 20% overweight level
represented by a BMI of 26.9, and 40% over ideal weight
represented by a BMI of 31.4. Thus, Blundell (1984) has
recommended that we use a BMI of 25 -30 to define mild

obesity and a BMI of 30 - 40 to define moderate obesity.

28

Johnson, Stalonas, and Christ (1978) reported results
of a weight reduction program using "percent of excess
weight lost". While not a widely used measure, this index
gives a useful indicator of the direct benefits to the
program participant.

The health risks associated with excess body weight (as
reviewed in Chapter I) are generally considered to become
significant when an individual reaches a body weight 20% or
more over his/her ideal weight. Burton et al. (1985)
suggest that the 20%-or-more over ideal weight figure is
arbitrary, but useful in prescribing clinical intervention.
Bray (1978) suggests that 25%-or-more over ideal weight is
the point where significant health risks begin (BMI 8 25).

In summary, the most widely used and recommended
indices of body weight and weight change are absolute
weight, body mass index, and percent overweight. An
individual may be considered overweight or obese when a more
than normal percentage of body weight is fat. For men,
obesity and its risks begin somewhere between 20% -25% over
ideal weight. For women, obesity and its risks begin at
about 30% over ideal weight.

3- —o;. - 1,)- ; r- 'e -e . 9 W- ., ;-- '-¢
In this section, literature pertaining to non-program

variables associated with obesity and weight reduction will

be reviewed. Specifically, literature will be reviewed

29

regarding personality factors, demographic variables,
genetic factors, and historical events.

Several authors, reviewing large numbers of studies
on obese individuals, have reported that there is no
evidence to support the existence of a specific pattern of
personality in overweight individuals (Storlie, 1984: Weiss,
1977: Wilson, 1985; Stunkard, 1978). Similar results have
been reported regarding overweight individuals who have been
successful in weight reduction programs: global personality
patterns do not predict who will be successful in weight
reduction efforts (Weiss, 1977: Stunkard, 1978: Wilson,
1985). Two studies with converging results, however,
contradict this general finding (Bjorvell, Edman, Rossner,
and Schalling, 1985: Lauer, Wampler, Lantz, and Romine,
1979).

In Sweden, Bjorvell et al. (1985) administered the
Karolinska Scales of Personality to 107 obese subjects and a
randomly sampled control group. Bjorvell et al. (1985)
found that the obese subjects evidenced a distinct
Impulsiveness Syndrome, characterized by irresponsibility,
acting-out, and failure to use insight. This pattern was
noted to be highly similar to personality patterns found in‘
alcoholics and drug addicts. In the U.S., Lauer et al.
(1979) found similar results administering a battery of
personality tests to 58 adult women averaging 92%

overweight. Lauer et al. (1979) administered the Minnesota

30

Multiphasic Personality Inventory, the Tennessee Self-
Concept Scale, and the Edwards Personal Preference Schedule
to their subjects and found that the obese women had a
significant (p <.05) personality pattern of addictive
potential similar to alcoholics. Lauer et al.'s (1979)
obese women were characterized by having a low endurance
drive, a need to dominate others, a tendency to avoid
closeness in relationships, and traditional sex-role
orientations. Thus, in contrast to the majority of research
indicating that no reliable personality pattern for obese
individuals exists, two studies have found a pattern of
addictive potential in overweight persons.

Lowe and Fisher (1983) and Strain and Strain (1979)
have suggested that overweight individuals are more
emotionally reactive and more likely to engage in
emotionally-cued eating than normal weight people. Lowe and
Fisher (1983) assessed this empirically with 17 individuals
who averaged 31% overweight and 30 individuals of normal
weight. Their results suggested that emotional eating was
associated with the degree of overweight (; a .46, o <.05).
Emotional binge eating has been associated with relapse from
weight control efforts (Wilson, 1985; Stunkard, 1978; Herman
and Polivy, 1975). Herman and Polivy (1975) have identified
"restrained eaters" and "unrestrained eaters" with the use
of their Restrained Eating Scale. Individuals scoring high

on restrained eating are more emotionally reactive, show

31

signs of a stress-deprivation syndrome, are more
distractible, and more likely to cease weight loss
strategies and eat large quantities of food in the face of
emotional stress (Coates, 1977: Harris, 1983: Wilson, 1985:
Herman and Polivy, 1975). Unrestrained eaters are more
likely to persist in their weight loss efforts under
emotional stress. Coates (1977) has characterized the
"restrained eater" as prone to experience breakdowns in
self-control due to poorly learned and haphazardly applied
self-regulatory strategies. The Restrained Eating Scale is
providing robust findings, but the interpretation of these
findings and the meaning of the restrained eating concept
are not as yet clear in the weight loss literature (Wilson,
1985).

Rotter's (1966) locus of control construct is a
personality variable that has been associated with weight
loss efforts. The research pertaining to the locus of
control construct in weight loss will be reviewed in a
separate section.

Demographic variables have also been studied with
regard to obesity and weight control. A greater percentage
of women than men are overweight (Bray, 1979: Annual
Report/NIH, 1985). Women are four times more likely to
participate in a weight reduction program than men (Wilson,
1985), but tend to lose less weight and reduce at slower

rates than men (Weiss, 1977; Wilson, 1985). Gender

 

32

differences in amount and rate of weight loss are, however,
highly equivocal findings. When percent body fat, initial
weight, and age are statistically controlled, gender
differences in weight loss success disappear (Weiss, 1977;
Wilson, 1985).

Socioeconomic status, race, and ethnicity have been
associated with obesity but not with success in weight
reduction efforts (Weiss, 1977: Brownell, 1985: Bray, 1979).
Obesity has been shown to follow socioeconomic status
gradients, with obesity being most common for low income
women and median income men (Annual Report/NIH, 1985).

While Bray (1979) found that black women were more likely to
be overweight than white women, other evidence suggests that
racial differences in body weight disappear when income and
education have been controlled (Annual Report/NIH, 1985).

No evidence indicates that race, SES, or ethnicity are
associated with success or failure in weight reduction
programs (Weiss, 1977).

Evidence drawn from a large number of studies indicates
that the age-at-onset of obesity may be a factor relevant to
success/failure in weight reduction efforts (Wilson, 1985:
Weiss, 1977: Hoiberg et al., 1984). Weiss (1977) and
Hoiberg et al. (1984) suggest that the evidence, though not
unequivocal, supports the statement that juvenile-onset
obesity is more resistant to treatment than adult-onset

obesity. Stunkard (1978), reviewing 21 studies of

33

behavioral treatments for obesity, concluded that adult-
onset and juvenile-onset overweight individuals respond
equivalently to treatment. Wilson (1985) has more recently
concluded that the evidence supporting a better treatment
response for adult-onset overweight persons is not very
strong, and that age-at-onset may be more significant in
long-term success of weight maintenance than short term
ability to produce a weight loss. Research on the age-at-
onset hypothesis has used widely varying criteria and
measures for success, making conclusions tenuous (Wilson,
1985).

Based on evidence derived from social epidemiological
research, it is difficult to separate genetic contributions
from social learning. However, a child of two obese
parents has a 300% chance of being an obese adult, as well
as of being 300% fatter than a comparable child of lean
parents (Annual Report/NIH, 1985). Further, one overweight
sibling is likely to have an overweight sibling 40% of the
time (Annual Report/NIH, 1985). While individuals with
weight-relevant hereditary disorders constitute only a very
small percentage of the overweight population (Piziak,
1983), many obese individuals have a genetically based
surplus of fat cells, or hyperplasia (Brownell, 1982).
Hyperplastically obese individuals may also be those that
develop overweight problems early in life but, independent

of age-at-onset, there is no evidence to support the

34

contention that hyperplastically obese persons have more
difficulty losing weight. Number of fat cells, however, may
determine how much weight can ultimately be lost (Brownell,
1982).

Personal history of weight loss efforts is another
factor that has been studied with relation to success in
weight loss programs. Gormally et al. (1980) studied a
sample of 40 adult women who averaged 23.8% over their ideal
weights and found that a previous history of dieting and
weight loss was significantly correlated with success in
initial weight loss efforts in a behavioral program (2 =
.42, o <.05). In contrast, however, Jeffery et al. (1984),
Jeffery et al. (1985), and Hoiberg et al. (1984) found an
inverse correlation between previous weight loss attempts
and success in weight loss programs. «Weiss (1977) and
Wilson (1985) suggest that the issue of previous efforts to
lose weight is not clearly related to treatment outcome.
Weiss (1977) reviewed four studies reporting results on
previous dieting attempts and found one study relating a
negative relationship to weight loss and three studies
revealing no relationship. Wilson (1985) has concluded that
weight loss history has not been proven effective as a
predictor of future weight loss success. Wilson (1985)
notes that there are two contradictory arguments on this
issue. One argument is that each time a new diet is begun,

the biological adaptation to restricted energy occurs at a

35

faster pace, making it harder to lose weight. The other
argument is that each time a person tries to accomplish a
weight-relevant behavior change, the greater are his/her
chances of success. Currently, no conclusive statements can
be made regarding previous history of weight loss efforts
and chances for success in subsequent efforts.

Wilson (1985) has reviewed evidence indicating that
even when the best six predictors of weight loss treatment
outcome are combined, only 25% - 50% of the variance is
accounted for, and only a small proportion of this is due to
non-program variables. As Stunkard (1978) has noted, there
is a great deal of variability in individual responses to
weight loss efforts that has not been accounted for.

W . l! E i !' I ! !

Porcello (1985) surveyed forty-two professional weight
reduction programs to determine the characteristics of
weight loss treatments available to consumers in the
marketplace. The sample included thirteen nationally based
programs, eleven regionally based programs, twelve hospital
weight loss programs, and six university based research
programs. Porcello (1985) found that the average duration
of a commercially available weight loss program was between
8 and 16 weeks. The average weekly weight loss, for both
men and women, was 2 - 3 pounds: the average cumulative
weight loss was 36 - 40 pounds for men and 26 - 30 pounds

for women (Porcello, 1985). The mean weekly cost of a

36

professional weight loss program was found to be $35.50; the
range of weekly costs was from zero dollars to $200.
Porcello's (1985) survey revealed that nutritional
counseling, behavioral counseling, and exercise regimens
were offered in all professional programs.

Several authors have noted the use of nutritional
education, behavior modification, and exercise, alone or in
combination, as primary treatment strategies for weight
reduction (Brownell, 1982; Stalonas et al., 1978: Miller and
Sims, 1981; Wing and Jeffery, 1979; Coates, 1977: Frankle,
1985). In a review of 145 weight loss research studies
published between 1966 and 1977, Wing and Jeffery (1979)
concluded that nutritional education, behavior modification,
and exercise were the prevailing non-medical weight loss
treatments. Wing and Jeffery (1979) discovered that the
average weekly weight loss in behavior therapy treatments
was 1.04 pounds, and that subjects lost an average of 11.2
pounds during the course of behavioral weight loss programs.
Similarly, subjects in nutritional education treatments lost
an average of 18.4 pounds, at the rate of 1.87 pounds per
week. Subjects in exercise programs lost an average of 8.3
pounds at a rate of 0.50 pounds per week. Wing and Jeffery
(1979) noted that the small apparent differences in these
three treatment strategies disappear when gender and length
of treatment are controlled. Exercise, as a weight

reduction strategy, tends to produce a slower rate of weight

37

loss in the initial weeks of treatment, whereas behavioral
and nutritional management approaches tend to produce faster
weight losses early in treatment; yet, over time, these
differences equalize (Wing and Jeffery, 1979).

Since each of the three major weight reduction
treatment strategies --- nutritional education, behavior
modification, and exercise --— generally occur together in
some combination (and to varying degrees of emphasis),
separating the relative contributions of each treatment is
difficult. Nevertheless, each treatment will be described
and reviewed separately below in so far as is possible.

Hall, Veale, Horne, and Watts (1984) have reviewed
evidence suggesting that severely obese individuals possess
a poor level of knowledge regarding nutrition and dietary
management. While this finding is open to question,
nutritional education and dieting have been traditional
approaches to weight loss (as reviewed above). In a
comprehensive review of nutritional approaches to weight
loss, VanItallie (1978) distinguished two primary types of
diets: 1) diets that overtly restrict calories, and 2) diets
that manipulate the circumstances of eating and attempt to
induce spontaneous changes in calorie intake. Diets that
overtly restrict caloric intake can be balanced (providing a
conventional distribution of protein, carbohydrate, and fat)
or unbalanced (eg, low-carbohydrate, high-protein, etc.).

Diets that manipulate eating or food characteristics may

 

38

attempt to alter the novelty, palatability, or taste of
foods, or may make use of behavioral principles to narrow
the range of food choices or eating times.

The median recommended dietary allowance for
maintaining body weight in adults between the ages of 23 and
50 years of age is 2700 kcal/day for men and 2000 kcal/day
for women (Dwyer, 1985). The amount of calorie restriction
per day necessary for weight reduction varies depending upon
how fast a rate of weight loss is desired (Dwyer, 1985;
James, 1984; Frankle, 1985; Snetselaar, 1983). James (1984)
suggests that a reduction of caloric intake of about 1000
kcal per day is necessary to produce a 2 lb./week rate of
weight loss. Dwyer (1985) suggests that it is not necessary
to reduce daily calorie intake below about 1200 kcal for
weight reduction.

Dwyer (1985) reviewed 39 popular diets available in the
marketplace and currently in use by American consumers.

Many diet plans do not make use of the reasonable rate-of-
weight-loss approach described above. Specifically, Very
Low Calorie Diets (VLC) recommending 500 or fewer calories
per day have produced dramatic weight losses, but are
associated with serious risks to the dieter's health (Dwyer,
1985; Hirsch, 1985b; Vertes, 1984; VanItallie and Abraham,
1985; Annual Report/NIH, 1985). VLC diets have been noted
to produce gastrointestinal disturbances, fatigue,

dizziness, hair loss, muscle cramps, insomnia, mood

39

disturbances, amenorrhea, and even sudden death (VanItallie
and Abraham, 1985: Vertes, 1984). Further, weight loss
maintenance has been noted to be very poor with VLC diets
(Vertes, 1984). Vertes (1984) and the National Institute of
Health (Annual Report/NIH, 1985) have recommended that VLC
diets be used only under the close supervision of a
physician and that they be followed with longer-term
behavioral and exercise interventions.

Hirsch (1985b) also has reviewed popular diets and
concluded that such diets select foods and eating principles
arbitrarily to provide simplicity and monotony in the hope
of spontaneously lowering food intake and producing rapid
(but unstable) weight losses. Hirsch (1985b) suggests that
the popularity of fad diets is based upon misinformation
about obesity and nutrition. Dwyer (1985), Frankle (1985),
Hirsch (1985b), Snetselaar (1983), and the National Dairy
Council (1985) have recommended the use of weight reduction
diets that are nutritionally balanced and provide for a slow
rate of stable weight loss. Frankle (1985), Snetselaar
(1983), and the National Dairy Council (1985) provide
excellent discussions of reasonable nutritional approaches
to weight loss and offer a variety of dieting plans that can
be tailored to individual needs. Frankle, the director of
nutrition for Weight Watchers International, suggests that a

nutritional management approach to weight loss alone can

40

produce clinically significant reductions in body weight
(Frankle, 1985).

Behavior modification is another widely used method to
induce reductions in body weight. Miller and Sims (1981),
Wilson (1980), Bellack (1975), and Blundell (1984) have
discussed the assumptions underlying a behavioral approach
to weight control. Within a behavioral paradigm, calorie
intake (eating) and calorie expenditure (activity) are
considered to be classes of behavior. A person's weight
control behaviors have been established over many years and
exist as habitual response patterns (Blundell, 1984).
Further, these weight control behaviors always take place
within the context of antecedent and consequent stimuli
(Blundell, 1984; Bellack, 1975; Miller and Sims, 1981:
Wilson, 1980). Individuals with weight problems persist in
habitual weight control behaviors despite the many aversive
consequences of such behaviors (Bellack, 1975). Bellack
(1975) suggests that there are two factors relevant to this
situation: 1) the immediate positive reinforcing
consequences of weight-relevant behaviors have greater
control over the behavior than delayed aversive
consequences, and 2) weight control behaviors are also under
the control of external antecedent stimuli. The task of
behavior therapy for obesity, then, is to analyze the
network of antecedent and consequent stimuli and adjust the

components to eliminate maladaptive responses and develop

 

41

productive behaviors for weight control (Blundell, 1984:
Miller and Sims, 1981).

The primary strategies for affecting consequent control
over weight relevant behaviors are initiating aversive
consequences for inappropriate weight control behaviors or
adding positive consequences for appropriate behaviors
(Bellack, 1975). The primary strategy for antecedent
control has been to bring weight control behaviors under the
control of conscious internal cues and to disrupt the
stimulus-response sequences in habitual behavior patterns
(Bellack, 1975). Thus, in contrast to a nutritional
education approach to overeating, for instance, a behavior
modification strategy would not focus on what a person eats
but rather upon how eating occurs (Blundell, 1984).

The primary "targets" in behavior therapy for obesity
are eating behaviors and activity behaviors (Miller and
Sims, 1981: Wilson, 1980; Sandifer and Buchanan, 1983). The
methods used to achieve habit changes in these target
behaviors are adapted from standard behavior therapy
approaches, and include stimulus control training, self-
monitoring, self-reward schedules, contingency contracting,
graphing, social skills training, stress management
training, and cognitive restructuring (Mahoney, 1974:
Sandifer and Buchanan, 1983; Bellack, 1975; Miller and Sims,

1981). Bellack (1975) provides a review of 22 stimulus

42

control techniques used in behavioral weight reduction
programs.

Sandifer and Buchanan (1983) trained 21 subjects
averaging 57% overweight in behavioral methods for weight
reduction and found that 11 behaviors predicted 87.5% of the
variance in weight lost during their program. Examples of
behavioral methods (and their correlations with the weight
reduction index) employed by Sandifer and Buchanan (1983)
include: taking a two-minute pause during a meal (3 s .78,

o <.001), eating only in a designated food area (;=.45,

P <.01), monitoring calorie intake (; = .62, o <.005),
limiting intake to one serving of each food (; = .67, o <
.001), refraining from other activities during a meal (; =
.60, o <.005), recording all food intake (1 = .57,p <.01),
and inhibiting urges to snack with aversive imagery (; =
.59, 2 <.005).

Miller and Sims (1983) trained 67 subjects averaging
48.1% overweight in behavior strategies that included
methods for stimulus control while shopping in supermarkets,
use of a daily self-monitoring record booklet, modification
of eating behaviors, and social skills training to increase
social support. The average weight loss in Miller and Sims'
(1983) study was 17.2 pounds (over a four week period).
Using stringent criteria for "success”, Miller and Sims

(1983) found that changes in eating behaviors and social

43

skills training were among the variables that predicted
success at a one year follow-up.

Jeffery, Wing, and Stunkard (1978) have used behavior
therapies for weight reduction at the Stanford Eating
Disorders Clinic. Their therapy package includes self-
monitoring of weight and other selected target behaviors,
stimulus control training, alternate response training,
dietary planning, and problem-solving skill development. A
sample of 125 adult subjects at the Stanford Clinic were
treated for overweight problems with Jeffery et al.'s (1978)
behavioral method. The female subjects averaged 67%
overweight, while the male subjects averaged 78% overweight
at the beginning of the program; across subjects, the
average initial weight was 215.9 pounds. Thirty-two
subjects participated in a ten week program and lost an
average of 7.2 pounds, while 93 subjects participated in a
20 week program and lost an average of 12.4 pounds. The
average weight loss using behavior modification at the
Stanford Eating Disorders Clinic has typically been between
ten and fifteen pounds (Jeffery, et al., 1978).

Stalonas, Johnson, and Christ (1978) have also found
that behavior therapy is effective in weight reduction
efforts. Stalonas et a1. (1978) trained adult subjects who
averaged 40.2% overweight in a basic behavioral package that
included stimulus control training, monitoring and graphing

of target behaviors, and nutritional advice. The subjects,

44

initially weighing an average of 181.1 pounds (range: 130-
275 lbs.), lost an average of 10 pounds by the tenth week of
the program (9 <.001). Stalonas et al. (1978) also found
that adding a contingency training component to the
behavioral package did not significantly increase the
effectiveness of the program, but that encouraging subjects
to exercise was effective (2 <.001) in producing additional
weight loss and superior maintenance of weight loss at one
year.

Currey, Malcolm, Riddle, and Schachte (1977) used a
behavioral weight loss treatment with 165 adult women
(average initial weight = 185 pounds) who had been
overweight for more than 20 years. In addition, 76% of the
subjects in Currey et al.'s (1977) study were receiving
treatment for medical conditions, including hypertension,
diabetes, cardiovascular disease and degenerative joint
disease. Even with this medically ill and chronically obese
group of subjects, behavior therapy was effective in
producing significant (p <.01) weight loss. Currey et al
(1977) report that 28% of their subjects lost between 11 and
19.8 pounds in the ten week program, and 23% lost more than
19.8 pounds (including 6% who lost more than 39.6 lbs.).
Given that Currey et al. (1977) worked with an unusually
"tough" population, these results indicate that behavior

therapy for weight reduction is an effective treatment.

45

Several major reviews of behavioral weight loss
treatments suggest similar results: behavior modification
methods produce a consistent I - 2 pound per week weight
loss, and an average loss of 11.5 pounds over the course of
a ten week program (Brownell, 1982; Wilson, 1980: Stunkard
and Penick, 1979; Blundell, 1984: Stunkard, 1975; Stunkard,
1978). It is also worth noting that behavior therapy has
been associated with the lowest attrition rates of all
weight reduction methods (Brownell, 1982; Stunkard, 1975,
1978; Blundell, 1984). Based upon reviews of more than 100
studies of behavioral weight loss programs, Brownell (1982)
and Wilson (1980) conclude that behavior modification is
effective in producing weight loss on a short-term basis,
that subjects often do not continue to lose weight after the
termination of the program, and that maintenance of weight
loss at one year follow-up is quite good.

Physical exercise is another method that has proven
effective in producing weight loss. Exercise leads to a
reduction of body weight by increasing caloric expenditure
and mobilizing the metabolism of fat (Brownell, 1982:
Blundell, 1984: Perri et al., 1986; Piziak, 1983). Physical
exercise works directly to decrease the size of adipose
cells (Blundell, 1984; Holm, Jacobsson, Holm, Bjorntorp, and
Smith, 1977).

In addition to providing a reduction in fat cell size

and percent body fat, exercise yields other benefits.

46

Physical exercise has been found to increase cardiovascular
fitness and produce a reduction in the risk factors
associated with obesity (Brownell, 1982; Blundell, 1984:
Stalonas et al., 1978;Perri et al., 1986: Bjorntorp, 1978).
Most importantly, physical exercise counteracts some of the
negative aspects of dieting. Specifically, exercise works
to minimize the loss of lean body tissue that occurs with
dieting (Brownell, 1982; Blundell, 1984; Perri et al.,
1986), and increases the basal metabolic rate which
ordinarily decrease during dieting (Brownell, 1982: Holm et
al. ,1977; Perri et al., 1986; Lennon, Nagle, Stratman,
Shrago, and Dennis, 1985; Tremblay, Fontaine, Poehlman,
Mitchell, Perron, and Bouchard, 1986). The increase in
basal metabolic rate, lasting for hours after exercise is
terminated, provides the major source of weight reduction in
this weight loss strategy (Brownell, 1982; Perri et al.,
1986; Tremblay et al., 1986). Physical exercise has also
been shown to produce beneficial effects in self-concept,
mood, and sense of well-being (Blundell, 1984).

Bjorntorp (1978) has suggested that physical exercise
may be a more effective weight loss strategy for mildly and
moderately obese individuals than severely obese
individuals, noting that severely obese persons may
experience more discomfort while exercising. Bjorntorp
(1978) also suggests, however, that exercise may be a better

weight loss strategy for persons unwilling to deprive

47

themselves with dietary restrictions. This concept
parallels Blundell's (1984) suggestion that it is easier to
introduce new active components into a person's behavioral
repetoire than to displace longstanding habits.

A variety of studies have found that physical exercise
can significantly reduce body weight even in the absence of
any dietary changes in women (Dahlkoetter, Callahan, and
Linton, 1979; Duddleston and Bennion, 1970; Franklin,
Buskirk, Hodgson, Gahagan, Kollias, and Mendez, 1979;
Gwinup, 1975; Zuti and Goldberg, 1976) and in men (Dempsey,
1964; Leon, Conrad, Hunninghake, and Serfass, 1979; Oscai
and Williams, 1968). The amount of weight lost, however,
depends upon the intensity, frequency, and duration of the
exercise regimen (Blundell, 1984).

Though Bjorntorp et al. (1973) failed to produce a
significant weight loss in five of eight severely and
hyperplastically obese persons using exercise only, the
majority of studies using exercise as a weight reduction
method have reported successful results. Leon et al. (1979)
trained young adult men whose BMI was 1-2 standard
deviations above the population mean in a walking program.
Leon et al.'s (1979) subjects walked for 90 minutes per day
on a treadmill at a slow pace; they walked five days per
week for 16 weeks. The subjects lost an average of 12.5

pounds, reducing their body fat content from 23.3% to 17.4%.

48

This reduction brought Leon et al.'s (1979) subjects below
the range at which obesity related health risks occur.

Oscai and Williams (1968) used a graduated exercise
program with five adult men who averaged 194 pounds and were
at least 15% over ideal weight (a matched control group of
sedentary subjects was also employed). Oscai and Williams'
(1968) subjects engaged in light walking/jogging for 15-30
minutes per day, three days per week. The subjects
experienced a weight loss of 10 pounds over the 16 week
program (o <.05). No dieting strategies were employed by
the subjects. Dempsey (1964) reported similar results using
a program of vigorous exercise with seven obese men.

Franklin et al. (1979) made successful use of exercise
(without dieting) to reduce weight in 23 obese adult women
whose mean initial weight was 167 +/- 20 pounds. Franklin
et al.'s (1979) subjects walked or jogged at 75% capacity
four days per week for twelve weeks, generally averaging 15-
20 minutes per day. The subjects lost an average of 5.6
pounds and reduced percent of body fat significantly (p
<.05). Similar results for a comparable program were
reported by Gwinup (1975).

Several studies have compared exercise with dieting as
weight loss strategies used alone or in combination.
Duddleston and Bennion (1970) assigned 12 adult women who
averaged 198 pounds and were at least 40% over ideal weight

to one of four groups: dieting (1200 kcal/day), exercise

49

(1 hour treadmill walking or bicycling 4 days per week),
dieting plus exercising, and a no-treatment control group.
Over the course of the six week program, exercisers lost an
average of 3.5 pounds: when combined with dieting,
exercisers lost 15.3 pounds. Zuti and Goldberg (1976) also
compared dieting, exercise, and a combination of the two
treatments on 25 adult women who were 20-40 pounds over
their ideal weights. Over the 16 week program, dieters lost
11.7 pounds, exercisers lost 10.6 pounds, and the combined
treatment subjects lost 12 pounds (all significant at o
<.01). The differences between the Duddleston and Bennion
(1970) exercise subjects and the Zuti and Goldberg (1976)
subjects in terms of weight lost during the programs is most
likely due to the length of the two programs. The rate of
weight loss through exercise strategies is somewhat slower
during initial weeks of training, increasing over time
(Bjorntorp, 1978).

Pi-Sunyer (1985), as well as Zuti and Goldberg (1976)
kept close measures of caloric intake on overweight subjects
who were exercising to lose weight. Results indicate that
exercise does not increase appetite or calorie intake, and
in many cases actually suppresses appetite. Thus, Pi-Sunyer
(1985) concludes that exercise alone can induce a loss of
body weight even in the absence of dieting. Also of note
are studies reporting significant weight losses with the use

of quite moderate, or submaximal, exercise regimens

50

(Tremblay et al., 1986; Gwinup, 1975; McKeen et al., 1983;
Lennon et al., 1985; Franklin et al., 1979: Holm et al.,
1977). Moderate exercise, such as walking, walk-jogging,
stationary cycling, or slow swimming produce significant and
stable weight losses in overweight persons. Walking as
little as 20 minutes per day on an every-other-day basis, at
only a 50%-75% level of intensity, produces weight losses
comparable to intense high-frequency aerobic programs when
results are examined over periods of time sufficient to
factor-out short-term weight loss due to depletion of body
fluids (Tremblay et al., 1986: MacKeen et al., 1983: Lennon
et al., 1985; Gwinup, 1975; Holm et al., 1977).

A major problem with exercise programs (for any group
of subjects, but particularly the obese) is the high rate of
attrition (Bjorntorp, 1978; Brownell, 1982: Dishman, 1982).
Brownell (1982), reviewing a large number of weight loss
studies, concluded that up to 30% of obese persons who enter
exercise programs will drop-out. Bjorntorp (1978) suggests
the average attrition rate for exercise programs designed
for overweight persons is 25% or more. Since exercise
produces a slower rate of weight loss than other weight
reduction strategies, obese subjects may get discouraged
with the lack of apparent pay-off for their exercise efforts
during the intial weeks of a program (Bjorntorp, 1978).
Based upon a review of ten studies examining exercise

treatments for obese persons, Bjorntorp (1978) sugggests the

 

51

following strategies for reducing attrition and increasing
adherence in exercise programs: start at a slow pace, use a
skillful and energetic group leader, avoid injuries,
exercise in pleasant surroundings, give physiological
feedback to participants, and encourage making exercise a
part of routine daily living. Dishman (1982) and Franklin
(1984), reviewing the exercise compliance literature, also
suggest that exercise programs should use realistic and
flexible goals that are set by participants, offer a variety
of exercise activities, demand only moderately intense
energy output, and offer the programs at convenient times
and locations.

Perri et al. (1986) reported impressive results in a
combined behavior therapy and exercise weight loss program,
with 80% of their subjects exceeding their target weight
loss goals. Perri et al. (1986) conclude that the success
of their program was due to the use of an exercise regimen
that consisted of simple activities, included a warm-up
routine, did not involve excessive pain or stress, and used
regular heart rate monitoring to allow participants to gauge
their own progress. Stalonas et al. (1978) also recommend
using simple exercise activities with obese subjects,
starting at a slow pace and increasing gradually the
intensity of exercise, and making ample use of self-
monitoring. Dahlkoetter et al. (1979) attributed their 0%

attrition rate to the use of a buddy-system in exercise

52

regimens for overweight people, a finding also reported by
Stalonas et al. (1978).

In summary, there are three widely used non-medical
approaches to the reduction of body weight: nutritional
management, behavior modification, and physical exercise.
These weight loss strategies have been used separately or in
combination. Nutritional management and behavior
modification produce a minimum average rate of weight loss
equal to approximately one pound per week while physical
exercise produces a somewhat slower minimum average rate of
weight loss of approximately 0.60 - 0.75 pounds per week.
Over time, the three weight loss methods produce similar
reductions in body weight if the treatment-specific
strategies are adhered to. Behavior modification and
exercise are associated with greater long-term maintenance
of weight loss. Exercise is associated with continued
weight loss following program termination. Exercise
programs generally show the highest attrition rate while
behavior modification programs have demonstrated low
attrition rates. The average length of a weight loss
program (not including maintenance programs) is about 10
weeks, ranging from 8 - 16 weeks. The average weight loss
obtained in a non-medical weight reduction program ranges

between 10 and 15 pounds per person.

 

53

t d a e

Subject attendance at program meetings and subject
adherence to strategies prescribed in weight loss treatments
are issues that have plagued weight reduction programs, as
has been the case in many health related treatments (Wilson,
1985). Since attendance at program meetings and adherence
to prescribed behaviors are strongly associated with success
in weight loss efforts (Jeffery et al., 1984: Wilson, 1985),
methods for increasing attendance and adherence in weight
loss programs have been a priority issue for behavioral
scientists working in the area of weight loss (Wing and
Jeffery, 1979; Wilson, 1985).

Several studies have reported a strong association
between attendance at program meetings and success in weight
loss efforts (Holmes et al., 1984; Jeffery et al., 1984;
Stuart and Guire, 1978; Dahlkoetter et al., 1979; Eufemia
and Wesolowski, 1985; Perri et al., 1984). Medically based
weight loss programs (eg, medication) have been shown to
have attrition rates as high as 80%, greatly compromising
the utility of such programs (Wilson, 1985). The major
contribution of behavioral strategies to weight reduction
programs has been to reduce attrition rates to 10%-15%
(Stunkard, 1975; Stunkard, 1978; Wilson, 1985: Brownell,
1982). Stunkard (1978), in a review of 21 representative

studies of behavioral approaches to weight reduction, found

54

that the average drop-out rate for subjects in these
programs was less than 10%. In other reviews of weight loss
treatments, Brownell (1982) and Wilson (1985) reported that
behavioral weight loss programs have a mean attrition rate
of 13.5%. The exact mechanism by which behavioral weight
loss strategies reduce attrition has not been clearly
specified, though the use of contingent monetary incentives
appears to play an important role (Brownell, 1982; Wilson,
1985).

In a review of 17 studies using attendance-contingent
monetary incentives, Wilson (1985) reported that the average
attrition rate in weight loss programs using monetary
incentives was 9.5%, compared with an average attrition rate
of 19.3% in behavioral treatments not using monetary
contingencies. Brownell (1982) reported similar results.
The ”schemes" for contingencies and pay-offs have varied
considerably in the literature. Stalonas et al. (1978) had
subjects deposit $10.00 and returned $1.00 each week
contingent upon attendance. Mahoney (1974) required a
$35.00 deposit and fined his subjects $5.00 for each absence
from a program meeting. In a meta-analytic review of
attrition in 97 behavioral weight loss studies conducted
between 1967 and 1984, Eufemia and Wesolowski (1985) found
strong support for the contention that monetary deposits
reduce subject attrition. Eufemia and Wesolowski (1985)

found that deposits of less than $23.50 resulted in

55

significantly higher attrition (26.11%) than did deposits of
more than $23.50 (14.12%).

In an unpublished doctoral dissertation, Mavis (1987)
assessed the relative efficacy of various monetary incentive
conditions in a behavioral weight reduction program. Mavis'
(1987) subjects were 118 adults, averaging 35.9% overweight
(mean initial weight = 205.6 pounds). Each participant
deposited $40.00, which was matched with equal funds from a
research grant, and was assigned to one of six incentive
conditions offering equivalent behavioral weight reduction
treatments. Members of the weight-contingent monetary
reward group were credited with $8.00 each week they reached
their weekly weight loss goal. The weight-contingent
monetary response cost group fined members each week that
their goals were not met. Two lottery payback conditions,
parallel to the monetary conditions (ie, credit vs. fine),
rewarded members with a chance in a lottery each week that
they met their weight loss goals. Money was pooled in the
lottery groups, and a drawing was held at the last treatment
session for a $1000 first prize, $400 second prize, and $200
third prize. The attendance-contingent group awarded
members $8.00 each week that they attended the program,
regardless of weight. Finally, there was a no-incentive
group which was free of charge or deposit.

Mavis' (1987) results indicatethat the use of monetary

incentives is effective in reducing attrition and in

56

increasing perceptions of treatment effectiveness in weight
reduction programs. Using the no-incentive group as a
baseline for attrition (65%), the weight-contingent
incentives (excluding the monetary response cost group)
reduced attrition by 52%. The weight-contingent monetary
response cost condition produced a high rate of attrition
(55%). Attendance-based incentives reduced attrition to
20%. Subjects in the no-incentive group rated the weight
loss program as less effective than subjects in the other
groups, suggesting that the use of incentives is associated
with perceptions of program efficacy. When asked to rate
the appeal of the various incentive schemes, 46% of the
subjects preferred the monetary reward incentive, 25%
preferred the lottery scheme, and 17% liked the attendance-
contingent reward. Only 12% of the subjects chose one of
the response cost procedures. Finally, while all of the
incentive conditions produced significantly greater weight
loss than the no-incentive condition, the weight-contingent
incentives were superior to the attendance-contingent
condition in promoting weight loss (Mavis, 1987).

Wing and Jeffery (1979), reviewing 145 studies of
weight reduction programs, concluded that exercise
approaches to weight loss produce the highest attrition
rates. This finding is consistent with the exercise
compliance literature generally, which suggests that

exercise programs have high drop-out rates (Dishman, 1982;

 

 

57

Franklin, 1984; Perri et al., 1986). Brownell (1982)
reports that exercise programs for obese persons will
typically yield at least a 30% attrition rate. Several
reasons for the high attrition rate in exercise programs
have been offered. Franklin (1984) found that simply being
overweight was a predictor of drop-out in exercise programs.
Wilson (1985) found that individuals who experience a slow
rate of weight loss are more likely to drop out of programs
and, as reviewed above, exercise-based weight reduction
programs tend to yield slower rates of weight loss. Wilson
(1985) found that up to 57% of subjects who dropped out of
weight reduction programs reported that their reason for
doing so was dissatisfaction with the slow rate of weight
loss.

In addition to monetary incentive schemes, other
factors helpful in reducing attrition and increasing
attendance in exercise-based weight loss programs have been
suggested. Stalonas et al. (1978) suggest that designing
exercise programs for the overweight which start at low
levels of physical exertion and build gradually to higher
levels of intensity will decrease the aversiveness of
exercise and thus reduce attrition. Perri et al. (1986)
suggest that their relatively low attrition rate in an
exercise program for obese persons was due to the use of
simple activities that did not involve excessive pain or

stress and the use of regular heart rate monitoring as a

 

 

58

feedback device for subjects to gauge their own progress.
Dahlkoetter et al. (1979) attribute their 0% attrition in an
exercise program for weight loss to the use of a buddy-
system, where participants exercised in pairs.

Pekarik et al. (1984) analysed the attrition patterns
of subjects in a behavioral weight reduction program,
separating subjects who dropped out early in the program
from those who dropped out late in the program. Pekarik et
al. (1984) used 52 adult subjects (50 were female) who
averaged 38.8% over ideal weight: 83% had participated in
weight reduction programs prior to Pekarik et al.'s program.
There were no differences between early drop-outs, late
drop-outs, and program completers on socioeconomic status,
education, marital status, age of obesity onset, prior
participation in weight loss programs, or evaluations of
prior weight loss programs. A key finding, however, was
that early drop-outs differed from late drop-outs and
program completers on the degree of personal responsibility
assumed for their weight loss efforts (Pekarik et al.,
1984). Methods that enhance attributions of personal
responsibility in weight loss efforts might reduce attrition
in weight loss programs and are in need of development.

Adherence to prescribed program methods and behaviors
has also been associated with success in weight loss
(Sandifer and Buchanan, 1983; Wilson, 1985; Dubbert and

Wilson, 1984: Holmes et al., 1984). In a study of 21

 

 

 

59

adults, averaging 57% overweight, who participated in a
behavioral weight loss program, Sandifer and Buchanan (1983)
found that adherence to prescribed behaviors predicted the
amount of weight lost in the program. In a multicomponent
study using behavior modification, exercise, and dieting,
Stalonas et al. (1978) reported an 82% rate of adherence to
program behaviors and a corresponding significant reduction
in body weight (mean weight loss = 10.7 pounds in ten weeks,
9 <.001).

Wilson (1985) notes that very little research has been
conducted assessing the reasons why overweight individuals
adhere to, or fail to adhere to, treatment prescriptions.
Similarly, little research has been conducted to assess
methods for increasing adherence in weight loss programs.
Monetary incentives, contingent upon adherence to prescribed
behaviors, represent the only scientific effort to increase
adherence in weight loss programs. A number of studies have
used monetary incentive schemes contingent upon adherence in
weight loss programs (Dahlkoetter et al., 1979; Perri et
al., 1986; Perri et al., 1984; Jeffery et al., 1976). These
studies have all demonstrated that monetary incentives,
contingent upon adherence to program behaviors, are
effective in increasing adherence and promoting weight loss.

Methods for measuring adherence in weight loss programs
have varied considerably, with self-report being the most

common approach. Stalonas et al. (1978) and Perri et al.

 

 

60

(1986) have reported a technique for evaluating adherence
that allows for at least some objectivity in adherence
ratings, though the approach still represents a self-report
method. The technique used by Stalonas et al. (1978) and
Perri et al. (1986) involves assigning an adherence score to
subjects on a weekly basis, using a three-point scale of
adherence. At each program meeting subjects present a self-
monitoring record to the therapist at the weigh-in. The
record kept by the overweight subject is based upon the
behaviors prescribed in the weight loss method he/she is
learning (eg, graphs of eating behaviors for behavior
modification subjects, record of number of minutes or days
of exercise for subjects in exercise conditions, caloric
intake record for subjects in a nutritional management
program, etc.). The therapist then assigns a score to rate
the degree of adherence: two-points representing full
adherence, one-point representing partial adherence, and
zero points representing nonadherence. While far from
perfect, this method of assessing the degree of subject
adherence to prescribed program behaviors appears to be
acceptable in the current status of weight loss research
(Wilson,1985).

In summary, both attendance and adherence are important
issues in weight reduction programs and related research.
Both attendance at program meetings and adherence with

prescribed program behaviors are associated with success in

61

weight loss efforts. Apart from the use of monetary
incentives to increase attendance and adherence, little
attention has been given to developing methods for enhancing
program conditions that might increase attendance and
adherence in weight loss programs.
ea e

The central hypotheses in the present study, regarding
the effects of choice of treatment on treatment outcomes,
has moderate support in the health care literature (Krantz
et al., 1980; Marston, 1970). While only two studies could
be located that directly assessed the role of treatment
choice in weight reduction (Bjorvell et al., 1985: Murray,
1976), a review of the alcohol treatment literature revealed
six studies assessing the role of choice of treatment in
alcohol rehabilitation programs (Sanchez-Craig, 1980;
Thornton et al., 1977; Vannicelli, 1979; Kissin et al.,1971;
Costello, 1975; Parker et al., 1979b) and four discussions
of the role of choice of treatments in alcohol
rehabilitation (Appel, 1986; Ewing, 1977; Miller and Hester,
1986: Parker et al., 1979a). In addition, four studies were
located that assessed the role of treatment choice in
recovery from myocardial infarction (Krantz and Deckel,
1983), academic performance (Perlmuter and Monty, 1979),
adjustment to nursing home mileu (Langer and Rodin, 1976),

and fear reduction in snake phobias (Devine and Fernald,

62

1973). These studies, and their implications for weight
reduction research, are reviewed below.

Kissin et al. (1971) assigned alcoholic patients at
random to be offered three, two, one, or no alternative
treatments for alcoholism. At a twelve month follow-up,
results indicated that giving patients a choice of alcohol
rehabilitation treatment resulted in greater acceptance of
the treatment and superior recovery rates. In addition,
success in treatment increased directly with the number of
alternatives available to the subject. In a study of 100
alcoholic adults, Vannicelli (1979) found that giving
patients a choice regarding the goals and nature of
treatment led to improved outcomes and greater program
satisfaction. Parker et al. (1979b) review an alcohol
rehabilitation program in a U.S. Army hospital that used a
choice-based "smorgasboard" approach to treatment. Parker
et al. (1979b) report that this program evidences a 0%
attrition rate, though factors unique to the military
culture may account for this finding.

The treatment choices available to alcohol
rehabilitation patients are not necessarily equivalent to
the types of choices available to weight loss subjects,
however. The choices Sanchez-Craig (1980) and Thornton et
al. (1977) made available to alcoholic patients were between
total abstinence and controlled-drinking. While patients in

'these two studies were assessed as having higher rates of

63

compliance and improved treatment outcomes (relative to no-
choice controls) the results were somewhat weak. Further,
in Sanchez-Craig's (1980) study, since all of the subjects
ended up engaging in some alcohol use, the stronger
compliance rating given to subjects who were allowed to set
their own controlled-drinking goals may not reflect a
clinically significant improvement. Parker et al. (1979b)
report the use of a wider range of alcohol treatment
choices, including individual counseling, group counseling,
didactic classes, physical conditioning, relaxation
training, disulfiram, Alcoholics Anonymous, and significant
other involvement.

In a review of 58 alcohol treatment program evaluations
conducted between 1951 and 1973 (that included 11,022
patients), Costello (1975) concluded that offering patients
a choice of alcohol rehabilitation treatments has
significant benefits, including: reducing reactance or
resistance to treatment, increasing the motivation for
attendance and adherence, enhancing stable self-management,
and increasing the success rate with high-risk/poor-
prognostic cases. In another large scale alcohol treatment
review, Parker et al. (1979a) conclude that giving patients
a choice of treatments enhances outcomes in the following
ways: 1) it decreases the aversive stigma of treatments
associated with a loss of control by giving some control to

the patient, 2) as treatment is seen as less aversive,

64

referrals increase, 3) it reduces attrition, 4) it reduces
resistance and reactance, and 5) facilitates more global
life style changes by enhancing intrinsic motivation.

Parker et al. (1979a, 1979b) and Ewing (1977), noting that
the etiological factors in alcoholism are still basically
unknown, suggest that there is no justification in
advocating a monotherapeutic approach to alcohol
rehabilitation. Several authors, therefore, recommend
giving alcoholic patients a choice of treatments to increase
treatment compliance, attendance, satisfaction, and recovery
from alcoholism (Miller and Hester, 1986; Appel, 1986:
Ewing, 1977; Parker et al., 1979a). Miller and Hester
(1986) even suggest that, given adequate information
regarding alternative treatments, patients may be better
than professionally trained therapists at selecting optimal
treatments for themselves.

The role of treatment choice has also been assessed in
other areas. Permulter and Monty (1979) provide a review of
choice of instructional procedures in academic settings.
Though the data are difficult to interpet directly,
Perlmuter and Monty (1979) conclude that giving students a
choice of instructional procedures leads to a situation
where students work harder, faster, and have a more positive
reaction to assigned tasks. Langer and Rodin (1976) allowed
some aged nursing home residents greater choice and

decision-making power in their institutional milieu and

65

found that these subjects showed greater happiness and a
longer life span than the subjects who were not given choice
and decision-making power.

Devine and Fernald (1973) gave 48 snake-phobic
undergraduate students a choice between four standard phobia
treatments. The subjects, who had extreme fear of snakes,
were shown a videotape demonstrating and describing four
treatments, including systematic desensitization, encounter
therapy, rational emotive therapy, and a combination of
modeling and rehearsal. The subjects were then asked to
rate the treatments in order of their preferences. Four
treatment groups were devised, each containing four subjects
who were receiving their preferred treatment, four subjects
who were randomly assigned, and four subjects who were
receiving a treatment they had indicated they strongly did
not want. Devine and Fernald (1973) found that subjects who
received their preferred form of treatment experienced
significantly more fear reduction than did subjects who were
randomly assigned or who received their non-preferred
treatment (p <.01). Devine and Fernald (1973) offered some
possible explanations for their results: 1) preferred
treatments might have been more effective because subjects
expected that it would be so (ie, enhanced treatment-
efficacy expectancies), 2) subjects intuitively know what
the best treatment "fit" is for themselves and their

preference reflected this, or 3) cognitive bolstering

66

amongst subjects worked to enhance effort in subjects who
received their preferred treatment, while subjects who
received non-preferred treatments derogated their treatments
and put forth less effort.

In a review of literature regarding the role of
perceived control in recovery from myocardial infarction,
Krantz and Deckel (1983) suggest that general medical
outcomes improve as patients perceive control over treatment
when given choices in their health care. Krantz and Deckel
(1983) discuss variables that might confound this situation,
however. Perceived control, and hence outcomes, may
fluctuate depending upon the personal meaning a patient
gives to various treatment choices. Unkown individual
difference factors may mediate the relative effect of
perceived control. Some individuals do not desire control
over treatments or outcomes, thus reducing (or even
reversing) the effects of choice for those individuals.
Thus, the relationship between treatment choice and
treatment outcomes is not necessarily straightforward, and
may be mediated by personal factors (Krantz and Deckel,
1983).

Two studies using treatment choice have been reported
in the weight loss literature, and both studies are plagued
with methodological problems. Bjorvell et al. (1985)
offered 107 obese Swedes a choice between in-patient

behavioral treatment and outpatient jaw-fixation. Several

67

problems exist with this study. First, the two treatment
alternatives each involve a possibly great loss of freedom,
thus compromising the value of having a choice and reducing
any perceived freedom or control that might accrue from
having a choice. Second, no control group of overweight
subjects not receiving a treatment choice was used.
Finally, since the principal purpose of the study was to
assess personality patterns in obese persons, adequate data
on weight losses were not reported. Thus, the Bjorvell et
al. (1985) study does not provide an assessment of the role
of treatment choice on weight reduction program outcomes.
Murray (1976) offers the only direct assessment of the
role of treatment choice in a weight loss program. A
confounding of choice for time of treatment and choice for
type of treatment compromises Murray's (1976) study,
however. Of the 27 overweight female subjects that Murray
(1976) initially selected, only 12 remained after scheduling
conflicts appeared. Murray's (1976) remaining subjects were
between 23 and 43 years old (mean = 34.5 yrs.) and weighed
an average of 213.4 pounds. Murray's (1976) original plan
was to assign half of the subjects to their preferred
treatment and randomly assign the other half. This strategy
was compromised, however, by Murray's effort to give
subjects their preferred day/time of treatment. The two
treatments offered were 1) "self-control training", using a

popular approach to changing eating habits, and 2)

68

"determination-raising", a support group with the goal of
raising subjects' determination to change eating behaviors.

Six subjects preferred determination-raising and were
paired on initial weight by Murray (1976). One member of
each pair had a preferred day/time for treatment and was
given his/her preferred treatment at the preferred time; the
other pair members were assigned to the non-preferred
treatment. The six subjects preferring self-control
training were split on the basis of availability for one of
the time slots and on the basis of their initial weight in
an effort to keep equivalent weight levels between the two
preference conditions and the two treatments.

Over the course of the nine week (10 session) treatment
subjects who received their preferred treatment lost an
average of 11.6 pounds and subjects who received their non-
preferred treatment lost an average of 11.4 pounds (Murray,
1976). While both preferred treatment and non-preferred
treatment subjects lost significant amounts of weight (o
<.01), there was not a significant difference between the
conditions. The results at a three-month follow-up were
similar, with the preferred treatment subjects regaining
less weight than the non-preferred treatment subjects, but
not significantly so. Subjects receiving the determination-
raising treatment lost more weight (mean = -13.3 pounds)
than subjects receiving self-control training (mean = -9.7

pounds), but not significantly. There were no interactions

69

between the type of treatment received and preference for
the treatment received.

Murray's (1976) failure to find a significant outcome
effect for offering overweight subjects a choice of
treatment type may result from defects in his experimental
design. In addition to confounding choice of treatment type
with choice of treatment time, Murray's (1976) study
included only 12 subjects. This small sample size may have
masked the magnitude of the effects of treatment choice in
the statistical analysis. Further, the choice of treatments
offered to Murray's (1976) subjects does not adequately
reflect the types of weight reduction strategies currently
in use in weight loss programs. A wider range of choices
more representative of the currently used weight loss
methods may provide results different from those reported by
Murray (1976).

Krantz and Deckel's (1983) findings that personal
individual difference factors mediate the role of treatment
choice in affecting alcohol rehabilitation outcomes may also
apply to the role of treatment choice in weight reduction.
For instance, O'Leary (1985), in a review of self-efficacy
and health care, noted that judgements of perceived self-
efficacy will determine choice behavior. Self-efficacy
judgements may work to determine which activities will be
attempted and which will be avoided, as well as affecting

the amount of effort and persistance devoted to health care

70

activities (O'Leary, 1985; Bandura, 1977, 1982). Similarly,
Goldney and Cameron (1981), reviewing the role of locus of
control in weight management, suggest that an overweight
person's locus of control beliefs may determine the type of
weight loss strategy chosen. It is suggested that internal
locus of control beliefs will more likely lead to a choice
of exercise as a weight loss strategy (an active, self-
directed activity) and that external locus of control
beliefs will more likely lead to a choice of passive,
therapist-directed weight loss strategies such as social
support groups, medication, or nutritional education
(Goldney and Cameron, 1981). Further, the degree to which
someone desires having control may mediate the role of
treatment choice on weight reduction outcomes (cf. Burger
and Cooper, 1979). The relative desirability of having
control may impact the personal value of any perceived
increase in personal control accrued through having a choice
of treatment type. Thus, there exist relevant individual
difference factors that may interact with choice of
treatment in a weight reduction program. These factors are
reviewed in more detail in subsequent sections.

In summary, offering patients a choice of treatment
‘types has moderate support as an outcome-enhancing technique
in the health care literature. Treatment choice has been
associated with increases in treatment satisfaction,

compliance, and recovery rates in alcohol rehabilitation.

71

Treatment choice has also been associated with improved
outcomes in the treatment of snake phobias, recovery from
myocardial infarction, academic performance, and adjustment
to nursing home milieu. Two studies of the role of
treatment choice in weight loss programs were so compromised
with methodological problems that they did not provide an
adequate assessment of the thesis. Finally, individual
difference factors such as self-efficacy, locus of control,
and the desirability of control may mediate the role of
treatment choice in improving outcomes of weight reduction
programs.

The effort to reduce one's body weight represents a
self-regulation process that may be mediated, in part, by
generalized expectancies concerning the nature and locus of
personal control over outcomes (Bellack, 1975). Further, to
the extent that choice of treatment alternatives influences
perceived control, beliefs about control may also mediate
the choice process. Literature regarding the locus of
control construct in weight loss is reviewed in this
section.

Rotter (1966) introduced the construct of locus of
«control as a theoretical device to describe individual
«differences in beliefs regarding generalized outcome
expectancies. When a person perceives that events (eg,

reinforcements, outcomes) are contingent upon his/her own

72

behavior or personal characteristics, he/she is said to have
a belief that control is internally located: when outcomes
are perceived as resulting from factors not related to one's
behavior (eg, luck, chance, fate, the behavior of others),
the individual is said to have a belief in an external locus
of control (Rotter, 1966, 1975). Rotter (1966) defines
locus of control beliefs as expectancies that particular
behaviors will be followed by reinforcements. A person's
history of reinforcement will determine his/her locus of
control expectancies.

Bellack (1975) has suggested that external locus of
control beliefs represent a self-regulation deficit. He
suggests that individuals with an external locus of control
belief system are unable to evaluate their own behavior
adequately in the absence of external input, and thus do not
make effective use of self-reinforcement in behavior change
efforts (Bellack, 1975). Piziak (1983) and Hirsch (1985a)
have suggested that obesity represents an inadequate self-
regulatory system. By extension, locus of control beliefs
may describe problems with body weight regulation.

Garner et al. (1976) studied body-image disturbances in
16 adults with juvenile-onset obesity who ranged from 25% to
75% overweight. Garner et al. (1976) found that the obese
group was significantly more external on Rotter's (1966)
locus of control scale (p <.025) than were three groups of

normal weight controls. Garner et al. (1976) speculated

73

that obese persons experience little self-control and have
an overall sense of ineffectiveness with regard to efforts
to master their bodies. While this finding has not been
replicated, evidence does suggest that locus of control
beliefs are related to weight loss efforts.

Using an ill-defined sample of mildly overweight
adults, Weinberg et al. (1984) found that internality
correlated with weight loss in a comprehensive weight loss
program (3 = .44, o <.01). Weiss (1977) reviewed seven
studies that assessed locus of control in relation to weight
loss and found that persons with an internal locus of
control lose more weight in weight reduction programs than
do external locus of control overweight persons. It may be
more accurate to say, however, that overweight persons who
are more external in locus of control beliefs are more
likely to participate in weight reduction programs, but of
this population it is the more internal locus of control
subjects who will lose the most weight (Wallston and
Wallston, 1978; Weiss, 1977; Goldney and Cameron, 1981).
Based upon a closer examination of available data, it
appears that the type of treatment offered in a weight
reduction program has a differential effect on internal and
external locus of control individuals. Specifically,
persons with an internal locus of control lose more weight
in self-directed programs while persons with an external

locus of control lose more weight in therapist-directed

74

group programs and programs that use monetary incentives
(Weiss, 1977; Wallston and Wallston, 1978; Saltzer, 1978).

Several authors have suggested that a locus of control
measure may provide a diagnostic tool for prescribing weight
loss treatments congruent with a person's cognitive syle
(Weiss, 1977; Saltzer, 1978; Balch and Ross, 1975; Wallston
et al., 1976; Wallston and Wallston, 1978). Wallston et
al.(1976) matched subjects for locus of control scores and
type of weight reduction treatment to assess whether the
locus of control was a useful differential diagnosis
construct. Thirty-four adult women, averaging 32.4 pounds
overweight, were paired on the basis of their locus of
control scores, using the Health Locus of Control Inventory
of Wallston et al. (1976). Subjects were then randomly
assigned to either a self-directed weight control treatment
or a weight loss support group for eight weeks. As
predicted, internals who received the self-directed
treatment lost more weight and were more satisfied with
their treatment than were internals who received the support
group treatment (9 <.01). Similarly, externals who received
the support group treatment lost more weight and were more
satisfied with their treatment than were externals who
received the self-directed treatment. In other words,
subjects who were matched with treatments congruent with
their cognitive styles lost more weight and were more

satisfied with treatment (Wallston et al., 1976). Balch and

75

Ross (1975) found similar results with 34 adult women,
averaging 35% overweight, who participated in a nine session
self-control weight loss program: there were significant
correlations between internal locus of control beliefs and
both completion and success in the program.

Goldney and Cameron (1981), reviewing the literature on
locus of control and weight loss, suggest that locus of
control may predict the type of weight loss strategies
chosen by overweight people when they select treatments.
Specifically, Goldney and Cameron (1981) predict that only
those overweight persons with an internal locus of control
are likely to freely choose exercise as a weight loss
strategy. Obese persons with more external locus of control
attributions are predicted to be more likely to select
weight loss strategies that they believe will require less
self-direction, such as educational and supportive programs
or medical treatments. This hypothesis has not been
directly assessed.

Research regarding the locus of control construct and
weight loss is not unequivocal. Using Rotter's (1966)
scale, several authors have failed to find support for the
locus of control construct as a predictor of response to
weight reduction programs (Rodin, Bray, Atkinson, Dahms,
Greenway, Hamilton, and Molich, 1977: Lauer et al., 1979;
wallston et al., 1976), while others have found support for

the use of Rotter's (1966) scale as a diagnostic and

76

prescriptive tool (Balch and Ross, 1975; Weinberg et al.,
1984; Chambliss and Murray, 1979). This situation may
represent a lack of specificity in Rotter's (1966) scale for
measuring locus of control beliefs relevant to weight loss
efforts. Rotter (1975) recommended that researchers use
situation—specific expectancy measures when attempting to
predict behaviors in specific situations, noting that his
1966 Locus of Control Scale measures only generalized
expectancies. At least three efforts have been made to
develop locus of control scales specific to the weight
reduction situation (Wallston et al., 1976; Saltzer, 1982;
Tobias and MacDonald, 1977).

Wallston et al. (1976) developed the Health Locus of
Control (HLC) Scale for use in diagnostic and predictive
studies on health related behaviors. As discussed above,
Wallston et al. (1976) used the HLC inventory in an
investigation of matching overweight persons with weight
loss treatments on the basis of cognitive style. In a
review of three studies that used the Wallston et al. (1976)
HLC scale, Winefield (1982) concluded that the HLC scale was
inadequate as a predictor of health behaviors. Saltzer
(1978, 1982) also presents evidence suggesting that Wallston
et al.'s (1976) scale is not the best available locus of
control scale for weight loss research.

Tobias and MacDonald (1977) developed an internal-

external control of weight scale. On a sample of 27 normal

77

weight undergraduate students, Tobias and MacDonald found
that their five item locus of control scale (consisting of
ten items in five forced-choice pairs) had a test-retest
reliability of .52; Rotter's (1966) scale had a test-retest
reliability of .76 with the same sample. With a group of
100 undergraduate women who averaged 33.1% overweight,
Tobias and MacDonald (1977) found that locus of control
atributions changed over the course of a weight reduction
treatment (9 <.05). More specifically, subjects who
participated in a self-determination raising group became
more internal in their locus of control attributions while
subjects in behavioral contracting group became more
external in their control attributions (Tobias and
MacDonald, 1977).

Saltzer (1978, 1982) developed a four item Weight Locus
of Control (WLOC) Scale. Saltzer (1978) reports that the
WLOC is a better predictor of intention to lose weight than
either Rotter's (1966) I-E scale or Wallston et al.'s (1976)
HLC scale, supporting Rotter's (1975) suggestion that a
behavior-specific expectancy scale will prove more useful in
practical applications of the locus of control concept.
Using 110 undergraduates, Saltzer (1982) administered the
Rotter (1966) I-E scale, two versions of Wallston et al.'s
(1976) HLC scale, the WLOC scale, and the Crowne-Marlow
social desirability scale, in an effort to establish

psychometric properties of the WLOC. After a 24 day post-

78

test, the WLOC had a test-retest reliability of .67 (o
<.001). Using Cronbach's alpha, the WLOC was found to have
an internal consistency of .58. The WLOC correlated
moderately with Rotter's (1966) I-E scale (; = .32, o <.001)
and correlated mildly with Wallston et al.'s (1976) HLC
scale (I = .21, o <.02), suggesting that the WLOC is related
but not identical to these scales and posseses some
convergent validity. The WLOC was not significantly
correlated with the social desirability scale (; = -.03,
n.s.), suggesting that the WLOC is not biased by a social
desirability response set. While not a psychometrically
strong scale, the WLOC appears to be a measure of locus of
control more relevant to weight loss research than Rotter's
(1966) I-E scale, Wallston et al.'s (1976) HLC scale, or
Tobias and MacDonald's (1977) I-E scale of weight control.
Returning to Bellack's (1975) suggestion that external
locus of control individuals fail to adequately reinforce
themselves, several authors have suggested that there is a
relationship between locus of control and self-evaluation
(Chambliss and Murray, 1979; Kaplan, Atkins, and Reinsch,
1984; Weiss, 1977). In a study of locus of control, weight
loss, and self-efficacy attributions, Chambliss and Murray
(1979) found that while higher efficacy attributions for
weight loss appeared to increase a person's ability to lose
weight, this was only true for individuals who were internal

in locus of control. Kaplan et al. (1984) found similar

79

results in an exercise program for patients with chronic
obstructive pulmonary disease: correlations between self-
efficacy judgments and exercise compliance criteria were
significant for persons with internal locus of control but
not for externals. This apparent interaction between self-
efficacy attributions and locus of control expectancies
should not be surprising. Self-efficacy attributions refer
to a person's perceived ability to produce a behavior
(Bandura, 1977), while locus of control expectancies refer
to a perceived relationship between a behavior and an
outcome (Rotter, 1966). Self-efficacy and locus of control
are, then, related constructs. If an overweight person has
confidence that he/she can produce a certain weight
reduction behavior (high self-efficacy), but no confidence
that such a behavior will lead to weight loss (external
locus of control), then locus of control will be a better
predictor of weight loss than self-efficacy. On the other
hand, if a person has confidence that certain weight
reduction behaviors will lead to weight loss (internal locus
of control), but no confidence in his/her ability to produce
the behaviors (low self-efficacy), then self-efficacy will
be a better predictor of weight loss.

In summary, available evidence suggests that locus of
control attributions are relevant to weight loss. While
weight loss programs tend to attract individuals with

relatively external locus of control orientations,

80

individuals within programs that are more internal tend to
lose more weight. Overweight persons with internal locus of
control attributions have been found to respond more
favorably to self-directed programs, while overweight
persons with external locus of control attributions have
been found to respond more favorably to therapist-directed
group programs. Assertions regarding locus of control and
choice of treatments have not been directly tested. Four
locus of control scales have been used in the weight loss
literature, though none are fully acceptable. Saltzer's
(1982) Weight Locus of Control Scale appears to provide the
most direct assessment of weight-relevant locus of control
at the present time.
Ths_De§irabilitx_9f_92ntrel

In a review of coping processes following coronary
heart disease and stroke, Krantz and Deckel (1983) conclude
that general treatment outcomes improve when patients
perceive that they have some control over treatment. Krantz
and Deckel (1983) note, however, that several factors may
mediate this situation, including the possibility that some
individuals may not desire to have control over treatment.
In the present study regarding the effect of treatment
choice on weight reduction outcomes, the relative
desirability of control may also play a role in patient
perceptions, though this hypothesis has never been directly

assessed.

81

Burger and Cooper (1979) presented a scale to measure
the desirability of control. The Desirability of Control
Scale is a 20-item measure of the degree to which the
respondent desires having personal control, and is scored on
a seven point likert scale (Burger and Cooper, 1979). Using
a norming sample of 453 undergraduate psychology students,
Burger and Cooper (1979) found the desirability of control
scale to possess an internal consistency coefficient of .80:
an internal consistency of .81 was found on a second sample
of 98 undergraduates as well. Thirty-one of the 453
undergraduates were readministered the scale at a six week
re-testing, and a test-retest reliability coefficient of .75
was obtained. The Desirability of Control Scale had a weak
relationship with Rotter's (1966) I-E scale (; = -.19),
indicating that while internality and desirability of
control are related, the constructs are independent. The
Desirability of Control Scale was not significantly
correlated with responses on the Marlowe-Crowne social
desirability index (3 = .11, n.s.), indicating that
desirability of control does not represent a response set of
social desirability.

Dembroski, MacDougall, and Musante (1984) also found
that the Desirability of Control Scale was not significantly
related to the Rotter (1966) I-E scale in a group of 67 male
undergraduate students (1 = -.11, n.s.). Demobroski et al.

(1984) state that despite conceptual similarity, locus of

82

control and desirabilty of control have little overlap,
noting that the latter may tap actual preferences for
control while the former may reflect an ideology regarding
the nature of control.

Burger and Cooper (1979) have characterized the person
with a high desirability for control as being assertive,
decisive, and active. The high desirability of control
individual is a leader, seeks to influence others, and
avoids unpleasantness with manipulation. In contrast, the
person with a low desirability of control is characterized
as nonassertive, passive, indecisive, and not likely to try
to influence others. The low desirability of control
individual prefers that decisions be made by others. Burger
and Cooper (1979) suggest that differences in the desire to
control events should help account for variations in
observed behavior.

Burger and Cooper (1979) speculate that a desire for
control may be related to learned helplessness, and that a
person with a high desire for control may be more reactive
to uncontrollable situations. This hypothesis has not been
directly assessed. If true, however, this hypothesis may
imply that overweight individuals who are high in the
desirability of control, and who do not receive a choice of
weight loss treatments, may exhibit more reactance.

Further, there may be some relationship between the

83

desirability for control and a person's past history of
weight loss efforts in a treatment choice study.

In summary, the desirability of control is an
individual difference factor that may prove relevant to a
study of outcomes following treatment choice, though this
hypothesis has not been directly assessed. A person with a
relatively high degree of desire for control will prefer to
make his/her own choices and decisions, whereas a person low
in desire for control will prefer that decisions be made for
him/her by others. Burger and Cooper (1979) have presented
a scale for measuring the desirability of control that
possesses acceptable psychometric properties. The
desirability of control construct is distinct from the locus
of control construct.

W

Bandura (1977, 1982) postulates that a critical
determinant of behavior is not so much the person's
perception of the relationship between a behavior and an
outcome, but the individual's expectancies regarding his/her
ability to successfully implement the behavior. This
relates to the results of Gormally et al. (1980) which
indicate that subjects who sustain weight losses report
greater confidence in their ability to maintain weight loss
strategies than weight-regainers. O'Leary (1985) and
Sternberg (1985) have reviewed evidence that reveals a

direct relationship between a person's internal self-

84

statements and his/her ability to carry-out weight reduction
procedures and produce a weight loss. Judgements about
one's ability to initiate and sustain weight loss efforts
are referred to as self-efficacy expectations (O'Leary,
1985: Bandura, 1977, 1982).

Bandura (1977, 1982) has discussed the concept of self-
efficacy, a cognitive mediator of behavior. Self-efficacy
refers to the individual's perceived ability to perform a
coping response, an active behavior to deal effectively with
a specific situation. Bandura (1977) has differentiated
efficacy expectations from outcome expectations. An outcome
expectancy is defined as a person's estimate that a given
behavior will lead to certain outcomes. An efficacy
expectation is the person's conviction that he/she can
successfully execute the behavior. Self-efficacy, then, is
a form of self-referent thought that mediates the
relationship between knowledge and action.

Bandura (1977, 1982) also differentiates self-efficacy
from related constructs such as self-esteem and self-
concept, noting that self-efficacy refers to perceived
performance competency in specific situations, whereas self-
concept and self—esteem refer to global self-image across a
wide variety of situations.

Self-efficacy judgements, whether accurate or
inaccurate, influence the individual's choice of activities

and environmental settings (Bandura, 1977, 1982).

85

Judgements of self-efficacy also determine how much effort a
person will expend, and how long he/she will persist, in the
face of obstacles or aversive experiences.

With regard to weight loss efforts, self-efficacy
relates to the overweight person's sense of control or
subjective sense of mastery over temptations or urges in
weight loss relevant high-risk situations (Marlatt and
Gordon, 1985; Sternberg, 1985; O'Leary, 1985). According to
both Marlatt and Gordon's (1985) Relapse Prevention Model
and Bandura's (1977) self-efficacy theory, it is assumed
that when an overweight person is following a set of
treatment strategies governing nutrition, exercise, or
behavior modification, he/she experiences a sense of
personal control (self-efficacy) over these behaviors. The
perception of self-efficacy continues until the overweight
person encounters what Marlatt and Gordon (1985) call a
high-risk situation. If the individual persists with weight
loss strategies in the high-risk situation, his/her sense of
self-efficacy is enhanced; alternatively, a strong sense of
self-efficacy in the high-risk situation will yield
persistance behaviors (Sternberg, 1985). Conversely,
relapse in high-risk situations is associated with lower
self-efficacy attributions. The demoralizing effects of
relapse and reduced self-efficacy, and the consequent

decrease in future persistance in high-risk situations,

86

represents what Marlatt and Gordon (1985) have referred to
as the Abstinence Violation Effect.

The probability that an obese person will fail to
maintain the use of his/her weight loss strategies decreases
significantly, then, when the individual has a high level of
self-efficacy for performing the specific weight loss
strategies in the specific high-risk situation (Sternberg,
1985; O'Leary, 1985). The probability of relapse increases
when the overweight person has low self-efficacy
expectations of his/her ability to perform weight loss
strategies in specific high-risk situations. Past
experiences with weight loss efforts will play a role in the
overweight person's self-efficacy judgements regarding
his/her ability to perform specific weight control coping
responses. Past success, according to Bandura's (1977,
1982) theory, will enhance self-efficacy attributions, while
past failures will tend to lower self-efficacy attributions.
Thus, the overweight person's previous experiences with
weight control efforts may well impact his/her self-efficacy
judgements regarding which types of weight control behaviors
can be successfully used in specific high-risk situations.
The individual's past history of weight control efforts may
also determine which type of weight loss strategy an
overweight person will choose to learn if given a choice.

The relationship between past weight loss efforts, self-

87

efficacy judgments, and preferred weight loss strategies has
not been directly examined.

In a review of self-efficacy and health behaviors,
O'Leary (1985) concluded that the obese person's self-
efficacy judgments regarding ability to manage body weight
are typically low. O'Leary (1985) suggests that low self-
efficacy attributions may explain why overweight persons
fail to utilize effective self-regulatory measures in the
control of body weight. Unfortunately, few studies have
been conducted applying the self-efficacy construct to
weight control efforts. Chambliss and Murray (1979)
manipulated self-efficacy attributions in obese subjects
during a weight reduction program. Sixty-eight adult women,
averaging 31% overweight, were given a placebo medication
described as a metabolic stimulant. After two weeks, half
the subjects were told that the drug was inert and were
encouraged to attribute success in weight loss to their own
efforts and ability. Results indicated that for individuals
who were internal in locus of control, attributions of self-
efficacy increased ability to lose weight (g <.01). It
should be noted that self-efficacy was not measured in this
study, and its manipulation was assumed. Further, the
results suggest an interaction between self-efficacy and
locus of control (Chambliss and Murray, 1979).

Weinberg et al. (1984) divided 44 overweight adults

into high and low self-efficacy groups based upon pre-

88

existing levels of self-efficacy. Self-efficacy was also
manipulated in this study by telling a portion of the
subjects that they had been specially selected for their
self-control capacities; these subjects were encouraged to
make attributions to self-control efforts for their
successes. Weinberg et al. (1984) found that individuals
with high pre-existing levels of self-efficacy lost more
weight than individuals with low pre-existing levels of
self-efficacy (mean loss of 8 pounds vs. mean loss of 3
pounds, o <.01). Further, individuals with high levels of
manipulated self-efficacy lost more weight than individuals
who were not subject to self-efficacy manipulation (mean
loss of 7 pounds vs. mean loss of 2 pounds, o <.01).
Hartigan, Baker-Strauch, and Morris (1982) found
support for the utility of self-efficacy theory in weight
reduction research. Hartigan et al. (1982) examined
perceptions of the causes of obesity and treatment outcome
in the context of Weiner's attribution theory. Four areas
of causal attribution for goal attainment in weight loss
efforts were assessed: personal ability, expended effort,
task difficulty, and random factors (eg, luck). Hartigan et
al. (1982) randomly assigned 27 adult subjects, averaging
36% overweight, to one of three treatment conditions:
standard behavior therapy plus a nutritionally sound diet,
dieting alone, and a delayed-treatment control. After seven

weeks of treatment, only the behavior therapy group had lost

89

a significant amount of weight (mean=-10.55 pounds, p<.01).
Of Weiner's four attributional areas, only perceptions of a
lack of ability were associated with pre-treatment weight
(r=.43,p<.05). During the course of the study, only those
subjects who received a treatment increased their ability
attributions. Subjects who attributed treatment outcomes to
task difficulty (ie, "it's too difficult to lose weight")
were less likely to attribute their weight status to
personal ability factors (r=-.38, p<.02). Subjects who felt
personally or socially victimized in their obesity were less
likely to attribute treatment outcomes to ability (;=-
.32,o<.05). Hartigan et al. (1982) conclude that successful
weight loss is more a function of perceived ability than
effort expended. The authors recommend that weight loss
programs be designed to enhance ability attributions in
subjects, increase a sense of personal responsibility,
decrease feelings of vicitimization, and offer opportunity
for simple task-mastery.

A dissertation study by VanKoten-Chappell (1982) found
evidence contradictory to the majority of self-efficacy and
weight loss research. In VanKoten-Chappell's (1982) study,
individuals with low self-efficacy and unsupportive families
lost more weight than subjects with high self-efficacy
attributions and ample family support. The results may have
been due to an interaction between self-efficacy, other

subject personality characteristics, and the treatment

90

offered. Subjects received a hospital-based fasting
treatment with protein supplements. Subjects who were low
in self-efficacy were also measured as having dependent
personalities. VanKoten-Chappell (1982) argues that
dependent subjects with low self-efficacy may normally
respond better under medical structure and that the high
self-efficacy subjects may have been so supported (and
unchallenged) by their families that they were unmotivated
to lose weight.

While there is some empirical support for the utility
of self-efficacy theory in weight loss research, there are
difficulties associated with the measurement of self-
efficacy. Self-efficacy is not a global, cross-situational
construct like locus of control, but rather refers to a
person's judgments about his/her ability to cope with a
specific situation (Bandura, 1977). Thus, self-efficacy is
a "state" measure, not a "trait" measure, making its
assessment somewhat more difficult than more global
contructs. Further, the assessment of self-efficacy is a
relatively new endeavor, and the development of specific
measures for self-efficacy expectancies is an ongoing effort
in the psychological literature (Marlatt and Gordon, 1985).

The goal of assessing self-efficacy is to provide the
client with a list of potentially stressful or high-risk
situations that are likely to be encountered, and to

evaluate the client's judgments about his/her ability to

91

emit the necessary coping responses in those situations. In
the field of addictive behaviors, self-efficacy measures
have been used in smoking cessation programs, alcohol
treatment programs, and weight reduction programs (Marlatt
and Gordon, 1985; O'Leary, 1985; DiClemente, 1981).

According to Bandura (1977, 1982), self-efficacy can be
assessed along three dimensions: level, strength, and
generality. Level refers to a discrete yes/no judgment made
by the client regarding whether he/she has the ability to
perform the target behavior. Strength refers to the degree
of confidence the client has regarding his/her judgment.
Generality refers to the strength of the judgment across
varying situations. Self-efficacy questionaires used in
weight loss research typically involve presenting the
subject with a list of high-risk situations relevant to
weight control, and asking the subject to rate how much
confidence he/she has in his/her ability to perform weight
reduction strategies in those specific situations. Bandura
(1977, 1982) has recommended using a rating scale from 0% to
100% confidence, expressed in 10-point intervals.

Kaplan et al. (1984) have noted that there are serious
problems associated with establishing the reliability of
self-efficacy scales. The use of test-retest reliability
coefficients may not be apprOpriate because self-efficacy is
conceptualized as being a dynamic construct, changing over

time. Kaplan et al. (1984) also suggest that, since

92

efficacy attributions pertain only to specific situations,
internal consistency measures of reliability may not be
appropriate either. That is, each specific efficacy scale
is a single item, and a reliability estimate based on the
average intercorrelation between items may be an
inappropriate use of the psychometric model. Kaplan et al.
(1984) suggest that indirect evidence for the reliability of
self-efficacy scales can be obtained from correlations
between self-efficacy measures and other variables.

Few of the studies regarding self-efficacy and weight
loss reviewed above presented the measure of self-efficacy
employed. While Weinberg et al. (1984) did report their
self-efficacy measure, it apparently consisted of only one
item. VanKoten-Chappell (1982) reported the measure used in
her dissertation study, but failed to conduct any
psychometric analyses on the self-efficacy scale. Mavis
(1987) used VanKoten-Chappell's scale in his dissertation
and, following extensive psychometric analysis, found the
scale to possess acceptable psychometric properties.

Mavis (1987) reduced VanKoten-Chappell's (1982) self-
efficacy scale from 49 items to 30 items, and used the scale
in a weight reduction program with 118 overweight adults.
Cluster analyses of the self-efficacy scale revealed four
factors. The factors and their respective internal
consistency coefficients are as follows: emotional cues

(alpha=.92), situational cues (alpha=.83), social anxiety

93

(alpha=.82), and appetitive (alpha=.63). Low internal
consistency for the appetitive factor is undoubtedly due to
the fact that the cluster consisted of only two items. A
composite internal consistency coefficient for the entire 30
item scale was alpha = .91. Test-retest reliability
coefficients for the self-efficacy scale factors, based upon
the responses of 15 subjects retested after two weeks, were
also established: emotional cues (.91), situational (.71),
appetitive (.80), and social anxiety (.55). Thus, Mavis'
(1987) revision of VanKoten-Chappell's (1982) self-efficacy
scale possesses acceptable reliability for use in weight
loss research.

In summary, there is some support for the application
of Bandura's (1977) self-efficacy construct to weight loss
research. Self-efficacy is defined as an individual's
perceptions of his/her ability to execute a coping response
in a stressful situation. A person's self-efficacy
expectations work to determine the amount of effort and
persistance that will be put forth in the face of obstacles.
Higher levels of self-efficacy have been associated with
success in weight loss efforts and with the ability to
maintain weight loss strategies in weight relevant high-risk
situations. Self-efficacy attributions may also determine a
person's choice of activities. An individual's past history
of situation-specific outcomes may determine future self-

efficacy expectancies in similar situations. The assessment

94

of self-efficacy is a relatively new endeavor in the field
of psychology and problems exist with creating reliable
standardized measures. Mavis' (1987) revision of VanKoten-
Chappell's (1982) self-efficacy scale possesses acceptable
psychometric properties for use in weight loss research
until such time that a better scale is developed.

SQEEQIY

Relevant theory and research was reviewed in the
following areas: definitions and measurements of body weight
and weight change, personal and demographic variables
associated with body weight and weight reduction, non-
medical treatments for weight reduction, attendance and
adherence issues in weight reduction programs, patient
choice of treatment type, locus of control in weight loss
research, desirability of control, and self-efficacy
expectations in weight reduction efforts.

An individual may be considered overweight or obese
when a greater than normal percentage of body weight is fat.
For men, obesity and its risks begin when body weight
reaches a level somewhere between 20%-25% over ideal weight
(Body Mass Index a 27.2) as indexed by the 1983 Metropolitan
Life Insurance Tables. For women, obesity and its risks
begin at about 30% over ideal weight (BMI = 29.15). The
most widely used and recommended indices of body weight and
weight change are absolute weight (in pounds or kilograms),

body mass index, and percent overweight.

95

Global personality patterns have not been found to
predict obesity or response to weight reduction programs.
Two studies were located that reported a personality pattern
of addictive potential in overweight persons, similar to
that found in alcoholics. These studies stand alone in the
weight loss literature.

A greater percentage of women than men are overweight,
and women outnumber men in weight reduction programs by a
ratio of 4:1. Apparent gender differences in the rate of
weight loss during the course of weight reduction programs,
favoring men, disappear when age, percent body fat, and
initial weight are controlled. Obesity does appear to
follow socioeconomic gradients, with lower socioeconomic
status being associated with higher levels of body weight.
Apparent racial differences in body weight disappear when
income and education are controlled.

Evidence suggests, though not unequivocally, that
juvenile-onset obesity is more resistant to treatment than
adult-onset obesity. Age-at-onset may be a more important
factor in the maintenance of weight loss than in short-term
treatment response. Genetic factors may determine
hyperplastic obesity, but no evidence exists indicating that
hyperplastically obese individuals have more difficulty
losing weight.

Personal history of weight loss efforts has not proven

effective as a predictor of future success in weight loss

96

efforts. Two contradictory arguments exist on this issue:
one suggesting that past weight loss attempts make future
success less likely because of biological adaptations, the
other suggesting that past weight loss attempts make future
success more likely due to learning factors.

There are three widely used non-medical approaches to
the reduction of body weight: nutritional management,
behavior modification, and physical exercise. These weight
loss strategies have been used separately or in combination,
and produce roughly equivalent results in body weight
reduction. Nutritional management and behavior modification
produce a minimum average rate of weight loss equal to
approximately one pound per week while exercise produces a
somewhat slower minimum average rate of weight loss of
approximately 0.60 - 0.70 pounds per week. Over time, the
three weight loss methods produce equivalent reductions in
body weight if the treatment-specific strategies are adhered
to. Behavior modification and exercise are associated with
greater long-term maintenance of weight loss. Exercise is
associated with continued weight loss following program
termination. Exercise programs generally show the highest
attrition rate while behavior modification programs have
demonstrated low attrition rates. The average length of a
weight reduction program (not including maintenance
programs) is about ten weeks, ranging from 8 - 16 weeks.

The average weight loss obtained in non-medical weight

97

reduction programs ranges between 10 and 15 pounds per
person.

Attendance and adherence are important issues in weight
reduction programs and related research. Both attendance at
program meetings and adherence with prescribed program
behaviors are associated with success in weight loss
efforts. The use of monetary incentives, contingent upon
weight loss or attendance, has been found effective in
reducing attrition. Apart from the use of monetary
incentives, little attention has been given to developing
methods for enhancing program conditions that might increase
attendance and adherence in weight loss programs.

Offering patients a choice of treatment types has
moderate support as an outcome-enhancing technique in the
health care literature. Treatment choice has been
associated with increases in treatment satisfaction,
compliance, and recovery rates in alcohol rehabilitation.
Treatment choice has also been associated with improved
outcomes in the treatment of snake phobias, recovery from
myocardial infarction, academic performance, and adjustment
to nursing home milieu. Two studies of the role of
treatment choice in weight loss programs were so compromised
with methodological problems that they did not provide an
adequate assessment of the thesis. Individual difference

factors such as self-efficacy, locus of control, and the

98

desirability of control may mediate the role of treatment
choice in improving outcomes of weight reduction programs.

Available evidence suggests that locus of control
attributions are relevant to weight loss efforts. While
weight loss programs tend to attract individuals with
relatively external locus of control orientations,
individuals within programs that make relatively more
internal locus of control attributions will lose more
weight. Overweight persons with internal locus of control
attributions have been found to respond more favorably to
self-directed programs, while overweight persons with
external locus of control attributions have been found to
respond more favorably to therapist-directed group programs.
Assertions regarding locus of control and choice of
treatments have not been directly tested. Four locus of
control scales have been used in the weight loss literature.
Though none of the locus of control measures have fully
acceptable psychometric properties, Saltzer's (1982) Weight
Locus of Control Scale appears to be the most useful at the
present time.

The desirability of control is an individual difference
factor that may prove relevant to a study of outcomes
following treatment choice, though this hypothesis has not
been directly tested. A person with a relatively high
degree of desire for control will prefer to make his/her own

choices and decisions, whereas a person low in the desire

99

for control will prefer that decisions be made for him/her
by others. Burger and Cooper (1979) have presented a scale
for measuring the desirability of control that possesses
acceptable psychometric properties.

There is some support for the application of Bandura's
(1977) self-efficacy construct to weight loss research.
Self-efficacy is defined as an individual's perceptions of
his/her ability to execute a coping response in a stressful
situation. A person's self-efficacy expectations work to
determine the amount of effort and persistance that will be
put forth in the face of obstacles. Higher levels of self-
efficacy have been associated with success in weight loss
efforts and with the ability to maintain weight loss
strategies in weight-relevant high-risk situations. Self-
efficacy attributions may also determine a person's choice
of activities. An individual's past history of situation-
specific outcomes may determine future self-efficacy
expectancies in similar situations. The assessment of self-
efficacy is a relatively new endeavor in the field of
psychology and problems exist with creating reliable
standardized measures. Mavis' (1987) revision of VanKoten-
Chappell's (1982) self-efficacy scale possesses acceptable
psychometric properties for use in weight loss research

until such time that a better scale is developed.

CHAPTER III

METHODOLOGY

A plan for the design and implementation of the
research procedures to investigate the hypotheses generated
in Chapter I is presented in this chapter. The population
of interest is defined and sampling techniques are
described. The measures used in the study are described and
procedures for scoring the assessment devices are presented.
The design of the study and testable hypotheses are
delineated, and procedures for analytic treatment of the
data are outlined.

Regulation

The population of interest for this study consists of
overweight adults who freely respond to weight control
program advertisements. One possible limitation to the
generalization of research findings based on a sample of
this population is that not all overweight adults who engage
in efforts to lose weight participate in organized weight
reduction programs. That is, many overweight adults attempt
to lose weight on their own, without the aid of a structured
group program. This population of individuals has not been

adequately studied and, hence, results derived from an

100

101

organized group weight reduction program may not apply to
this population.
518213

The study sample consisted of 118 overweight male and
female adults who responded to a newspaper advertisement for
a weight reduction program placed in Tho_Loo§iog_§toto
Joorooi during February, 1987. Procedures for the selection
of subjects and their assignment to treatment groups is
described in the "Procedures" section below. Attrition is
described in Chapter IV. Based upon data obtained from the
initial research questionnaire administered at the first
program meeting, the sample can be characterized as follows:

1. The age ranged from 24 to 65 years of age, with a mean
of 42.2 years and a standard deviation of 9.65 years.

2. 15.8% of the subjects were male; 84.2% were female.

3. Marital status indicated that 11.6% were single, 63.2%
were married, 21.1% were divorced, and 4.2% were
widowed.

4. 86.4% of the subjects were White, 9.5% were Black,

4.2% were Hispanic.

5. The mean number of years of education was 15.2, with
a range of 11 to 22 years.

6. 71.6% of the subjects were employed full-time outside
their home, 13.7% were employed part-time outside their
home, 8.4% classified themselves as homemakers, 4.2%

were unemployed, and 2.1% were full-time students.

102

7. Based upon self-reported age-at-onset of obesity, the
range was from infancy to 50 years, with a mean of 21.4
years and a median of 22.5 years. 38.9% of the sample
was obese before the age of 18 years.

8. 33.7% of the subjects had made between 1 and 5 previous
attempts at losing weight, 27.4% had made between 6 and
10 previous attempts, 9.5% had made between 11 and 15
previous attempts, and 29.5% had made more than 15
prior attempts at losing weight. The average subject
had used about four different methods for losing
weight, with a range of 1 to 11.

9. At the first weigh-in, the subjects' weights ranged
from 134 to 343 pounds, with a mean of 197.6 and a
standard deviation of 41.5. The subjects ranged from
1.1 to 145.0 percent overweight, with a mean of 44.36
and standard deviation of 27.35. Body Mass Index
ranged from 22.5 to 53.5, with a mean of 32.4 and a
standard deviation of 6.02.

oce e
Soojoot_§oioo§ioo. The sample under study in the

present investigation was drawn from a population of males
and females residing in the Lansing, Michigan metropolitan
area. The research subjects were recruited through an
advertisement placed in the Lagging_§toto_goo;noi (February,
1987) for Mavis' (1987) weight loss research study. As

Mavis' advertisement yielded far more respondents than were

103

necessary for his study, a waiting list of potential weight
loss program participants was created and these individuals
were notified that they would be eligible for the next
program offered. In June, 1987, three hundred (300) wait-
list subjects were contacted through the mail to 1) ensure
their willingness to participate in the entire ten week
program, 2) to provide information regarding further contact
and assignment to a weight reduction treatment condition,
and 3) to begin the physical risk factor screening process.

Through a form letter, the 300 subjects were asked to
refrain from participating in the weight loss program if
they were pregnant, had diabetes, or were under medical care
for hypertension. In addition, the Physical Activity
Readiness Questionaire (PARQ), used by the American Heart
Association (1984a), was included in the mailing to identify
and screen out individuals who might be at high risk for
participation in this study. The PARQ can be found in
Appendix A.

The rationale for screening out potential subjects who
may have possessed physical health risks for participating
in exercise prior to actual selection of subjects was to
help prevent differential exclusion or attrition after
randomization. Individuals who, on the PARQ, indicated the
existence of potential risk factors for participating in
physical exercise were excluded from participation in the

study. The only exceptions to this were individuals who

104

marked only item #7 on the PARQ, indicating they were men
over the age of 45 or women over the age of 50 and had no
other potential health risk factors. These individuals were
asked to provide written consent from their personal
physicians prior to being invited to participate in the
study. Individuals who did not indicate risk factors on the
PARQ, and who were men under the age of 45 or women under
the age of 50, were considered to be at minimal risk and
were allowed to participate in the study without further
screening. Note that subjects in the exercise training
conditions received an additional assessment for physical
fitness and potential health risks (see outline of Exercise
Training treatments in Appendix D).

fiobjoo§_o§§ionmoot. Of the 300 potential subjects
contacted by mail, about 200 returned the PARQ and expressed
interest in participating in the study. One hundred forty
(140) individuals were eligible for participation in the
study.

One hundred twenty (120) subjects were selected and
randomly assigned to two groups of sixty subjects,
comprising the two levels of factor A (choice of treatment
type versus random assignment to treatment type).

The sixty subjects who were to receive a choice of
treatment type (Group A.1) were contacted through the mail
and asked to attend a meeting where the three weight loss

methods (ie, nutritional education, behavior management, and

105

exercise) would be described. The sixty subjects in group
A.2 were randomly assigned to one of the three treatment
types and were mailed instructions regarding the date, time,
and place of the first meeting for their assigned treatment.
One subject from each level of factor A failed to show up
for the first scheduled meeting (and could not be reached by
phone), leaving 59 subjects in each level of factor A (118
subjects overall).

Following a description of the treatments and the
overall program at the pre-program meeting, the subjects in
group A.1 were asked to choose the one treatment method in
which they desired to receive training during the course of
the ten week program. Subjects in group A.1 were then
assigned to their treatment of choice and informed of the
date, time, and place of the first program meetings. No
changes in choice of treatment were allowed at that time.
Subjects in group A.2 were not offered a choice of treatment
type and were not informed that subjects in group A.1 were
offered a treatment choice until the end of the program.

The description of treatment alternatives offered to
subjects in group A.1 was constructed to reflect the
available research data on the three major weight loss
strategies. The treatments were described as roughly
equivalent in their effectiveness for promoting weight loss,
though known treatment differences were highlighted. Each

subject was encouraged to select the treatment strategy that

106

he/she evaluated as appropriate for him/her at the present
time. The description of treatments and instructions to
subjects can be found in Appendix 8, along with the results
of a pilot study conducted to demonstrate equivalence of the
treatment descriptions.

The description of treatments was offered to subjects
in group A.2 at the first program meeting, though the
opportunity for choice of treatment type was not addressed
at that time. Subjects in group A.2 were informed of the
experimental design at the last treatment session during the
debriefing discussion, as were subjects in group A.1.

All subjects were encouraged to view the weight
reduction strategy assigned as a means to initiate weight
reduction. All subjects were informed that they would be
offered a weight loss maintenance program following the ten
week program and were strongly encouraged to participate in
the maintenance program. Subjects were instructed that
weight loss (and its maintenance) is a long term process and
that the present program was merely a means to begin the
process of weight reduction with thorough training in a
proven method for weight control.

0 et ' s o t . All subjects were
required to deposit $45.00 with the experimenter at the
beginning of the program. Five dollars from each subject's
deposit was used to defray costs of the program associated

with photocopying expenses for program handouts and

107

treatment manuals. The remaining forty dollars from each
participant's deposit was placed in an account with the
Department of Psychiatry at Michigan State University. The
pool of deposited money was used in a monetary incentive
scheme for weight loss program participants.

The monetary incentive scheme used in this study was
based upon a similar strategy used by Mavis (1987). Each
subject received a raffle credit, contingent upon
attendance, at each program meeting. At the tenth program
meeting, a raffle for monetary prizes was held in each
treatment group. The prize amounts were based upon a total
raffle fund of $3750.00 deposited by the 95 active program
participants. The pooled monetary deposits were divided
evenly among the six treatment groups ($625.00 per group) to
ensure equivalent incentive strength for all participants.
Each treatment group raffle offered three prizes: lst prize
a $350, 2nd prize = $175, 3rd prize = $100. It should be
noted that a $45.00 "fee” for a ten week weight loss program
is substantially less than the average cost of $385.00 for
similar programs in the commercial marketplace (Porcello,
1985).

IEIQIEQQ_§QD§§E§- Only those subjects who provided
informed consent were accepted into the study. Subjects
were provided with an Informed Consent Form (see Appendix C)
after they received a description of the three treatment

conditions and had knowledge of which particular treatment

108

strategy they had been assigned to. For subjects in group
A.1, distribution of the Informed Consent Form occured at
the organizational meeting where treatments were described
and a choice of treatment type was offered. For subjects in
group A.2, distribution of Informed Consent Forms occured in
the first program meeting following a description of the
treatments. Participants in group A.2 were told that they
had been randomly assigned to a treatment condition. All
participants were told prior to providing informed consent
that a full disclosure of the specific nature of the
research would be explained in the last treatment session.
ss eat t o d't' ns. Three weight

reduction strategies were offered to subjects in this study:
nutritional education, behavior management, and exercise
training. Each treatment was based upon an established
weight reduction method drawn from the weight reduction
literature. Identical programs were conducted for each of
the major treatment conditions: one each for subjects in
group A.1 receiving a choice of treatment and one each for
subjects in group A.2 randomly assigned to treatments.
Outlines for each ten week treatment program are presented
in Appendix D.

The nutritional education and nutritional management
treatment program offered to subjects in this study was

based upon the weight management plan published by the
National Dairy Council (1985). Lifestoos:You; Porsooai Elan

109

f2r_fleight_uanagement (National Dairy Council. 1985)

contains extensive information regarding caloric values in
common foods, shopping and meal planning, suggested menus,
guidelines for restaurant eating, and educational materials
regarding the nutritional requirements of the human body.
Lifioogoog provides a method for personalizing dietary plans
to fit individual needs. Behavior modification methods
contained within the National Dairy Council program were
included in the present program, but not emphasized.

The behavior management treatment offered to subjects
in this study was based upon behavioral treatment methods
widely reported in the weight loss literature (Bellack,
1975; Stalonas et al., 1978; Sandifer and Buchanan, 1983:
Brownell, 1979). Subjects were trained to identify weight-
relevant behaviors, to specify and set personal behavioral
goals, to monitor personal behaviors, to graph target
behaviors, to develop alternative coping behaviors, to
develop contingency contracts with significant others, to
identify behavioral relapse cues, to develop stimulus
control strategies, and to monitor and alter internal self-
dialogue. Subjects were also offered training in
assertiveness and stress management. The subjects in the
behavior modification treatments were given a modified
version of Brownell's (1979) weight control manual. The
manual was supplemented with additional written handouts on

behavior management methods drawn from various sources in

110

the weight reduction research literature. The Brownell
(1979) manual contains information on nutrition and exercise
that was included in the program but not emphasized.

The exercise training program offered to subjects in
this study was based upon the work of Perri et al. (1986),
the American Heart Association (1984a, 1984b, 1984c, 1984d),
and Bjorntorp (1978). Subjects were educated regarding the
role of exercise in weight reduction and trained in safe
aerobic physical activities. Subjects received written
handouts on aerobic exercise, caloric expenditure associated
with various physical activities, and forms to aid in
monitoring progress in physical conditioning. The exercise
program consisted of submaximal physical exercise, which has
been shown to promote significant weight loss and is
associated with minimal injury (MacKeen et al., 1983:
Tremblay et al., 1986: Lennon et al., 1985; Holm et al.,
1977; Franklin et al., 1979; Gwinup, 1975). Submaximal
physical exercise generally consists of physical activity at
50%-75% of aerobic capacity. Subjects received information
on a variety of methods of moderate physical exercise,
though brisk walking was the primary exercise strategy
emphasized. Each program meeting involved therapist-led
demonstrations, a warm-up routine, an exercise routine, and
a cool-down period. Exercise intensity began at low levels
and was gradually increased to moderate levels. Individual

exercise between program meetings was strongly encouraged.

111

Exercise training subjects also received the Brownell (1979)
manual as an informational resource to supplement their
exercise training, though the behavioral and nutritional
aspects of the manual were not emphasized.

Thus, each of three treatment conditions offered to
subjects in the present study was based upon a rational and
documented strategy of weight loss. Each weight reduction
method was taught as a separate strategy, though some
overlap existed between the methods. Two identical programs
were conducted for each treatment condition: one each for
subjects in group A.1 receiving a choice of treatment and
one each for subjects in group A.2 randomly assigned to
treatments. All subjects were assigned weekly homework
assignments, relevant to each treatment condition, which
were monitored by the program leader. Specific information
pertaining to each treatment program can be found in
Appendix D and in the treatment manuals referenced above.

e t t ov' e . Each treatment strategy was
taught by an individual with expertise specific to the
intervention, as described below. The principal
investigator was present at each treatment session for all
groups and acted as program leader. The principal
investigator also collected data at the appropriate times.

The two nutritional education treatments were taught by
a nutritional consultant employed by the Dairy Council of

Michigan, volunteering his time to this study in exchange

112

for advanced training credit with his employer. The
nutrition education group leader had just over four years of
experience providing nutritional education, and was familiar
with the treatment manual used in this study. The nutrition
management group leader had a master's degree in human
nutrition and was a registered dietician.

The exercise training groups were led by an exercise
physiologist. She held a master's degree in exercise
physiology, and had just over four years of experience
leading exercise and health promotion programs. She had
previous experience developing and leading exercise programs
for weight loss. The exercise instructor was paid $350.00
for her involvement in the study: $300.00 was contributed by
the Department of Psychiatry at Michigan State University
and the investigator contributed the remaining $50.00.

The two behavioral management treatments were led by
the principal investigator, a doctoral candidate in
counseling psychology. The behavioral management instructor
holds a master's degree in clinical mental health counseling
and had over four years of experience in the provision of
behavior management interventions.

Thus, all three treatment providers in the present
study held master's degrees in their respective fields and
had just over four years of experience in their specialty

areas. While the three treatment providers were not

113

"equated" for education and experience, they possessed
roughly equivalent levels of expertise.
ugooozomoot_§ohooolo. Appendix E contains the measures
used in this study; specific information about the measures
is contained below in the section entitled Instrumentation.

Body weight, attendance, and adherence were measured at
each program meeting. A demographic questionaire was
administered at the first program meeting, as were the
Desirability of Control Scale and the Weight Locus of
Control Scale. Measures of perceptions of treatment
efficacy and self-efficacy were administered at the first
treatment session, the fifth treatment session (mid-program
measure), and at the last treatment session. A measure of
treatment satisfaction and reactance was administered at the
fifth program meeting and the final program meeting.

A separate questionaire was mailed to subjects who
dropped out of the study during the course of the program,
assessing their reasons for dropping out. A 62% return rate
was obtained on the first mailing of this questionaire, and
a 0% return rate was obtained on the second mailing.
Subjects who dropped out at the first program meeting were
called on the telephone, read the drop-out questionaire,
and asked to supply the reason for their decision to refrain
from participation in the program. As attrition is an
outcome of this study, results relating to drop-out are

contained in Chapter IV. Longer term follow-up assessments

114

were not planned as a part of the study, as the primary
research questions were about short-term outcomes.

Deb;iofing_ond_§oiiog;oo. Full disclosure of the
research design, fundamental research questions explored in
this study, and available results were explained to subjects
at the final program meeting.

All subjects were offered a weight loss maintenance
program immediately following the study. Subjects were
informed of this program at the beginning of the study and
strongly encouraged to participate in the maintenance
program throughout the ten week training program.
mimics

Eleven areas of measurement are included in this study.
Body weight, attendance, attrition, adherence, program
satisfaction, program reactance, perceptions of self-
efficacy, and perceptions of treatment efficacy are outcome
variables. The desirability of control, locus of control,
and a variety of demographic factors were measured as
potential predictor variables. The assessment devices used
in this study can be found in Appendix D.

Booy_goigh§. Body weight was measured at weekly
program meetings for all subjects using the same scale.
Body weight was measured in pounds. Body weight data will
also be discussed in terms of Body Mass Index (BMI) and

percent over ideal weight (as defined in Chapter II).

115

AEEEEQQQQQ- Attendance records for each subject were
kept, using the weekly weight records as the data source.
Attendance data are reported as total attendance figures for
for each subject for the entire program (10 sessions).

Aoggioioo. Data regarding attrition from the program
were drawn from the weekly weight records and the mid- and
post-program questionaires. Individuals who did not
complete the pre-test questionaires were considered drop-
outs; individuals who did not complete the final program
questionaire were also considered to be drop outs.

AQBQIQEQQ- Adherence to prescribed program behaviors
was assessed for all subjects using the method described by
Perri et al. (1986) and Stalonas et al. (1978). Each
subject in each treatment condition received weekly homework
assignments relevant to his/her specific treatment
condition. At each weekly weigh-in, the experimenter
inquired of each subject whether the homework was completed.
The experimenter assigned an adherence score to each
subject's homework self-report. A score of two points was
used to indicate full adherence, a score of one point
indicated partial adherence, and a score of zero points
indicated nonadherence. Adherence data are analyzed and
reported in terms of total adherence scores for the entire

program (possible range = 0 to 18, for nine sessions).

116

Subjects were informed that the experimenter was
evaluating adherence, though no contingencies were
associated with subject adherence ratings.

-ef ' a . Perceptions of self—efficacy for one's
ability to maintain weight loss strategies in the face of
weight control relevant stress were assessed using the
"Self-Confidence" scale. The self-efficacy scale used in
this study was developed by VanKoten-Chappell (1982) and
revised by Mavis (1987). Using 118 obese subjects in a
weight loss program, Mavis found the self-efficacy scale to
possess an internal consistency reliability coefficient
(alpha) of .91 (Mavis, 1987). The internal consistency of
the self-efficacy scale obtained in the present study was
.93 (Cronbach's alpha).

The thirty-item measure is scored by the testee on an
eleven point response gradient (0 - 10), reflecting the
strength of the subject's degree of self-confidence for
his/her ability to maintain weight loss strategies in each
situation described. Data are reported in terms of each
subject's average score for each test period.

'o s me t ' a . Subjects'
perceptions of the efficacy of their assigned treatment
method was assessed using the "Perceptions of Your Weight
Loss Method" questionaire. This measure consists of three

items, scored on a five point scale (0 -4) for agreement

117

with the item statement. Data are reported in terms of the
average score of the three items.

Items #2 and #3 were drawn from Mavis' (1987) weight
reduction program questionaire. They possessed an internal
consistency coefficient of .79 for Mavis' (1987) sample.
Item #1 was taken from Borkovec and Nau (1972), and did not
have a reported reliability coefficient. The internal
consistency reliability estimate for the three item
perceptions of treatment efficacy scale used in the present
study, obtained from the sample in the present study, was
alpha = .61.

Item four, asking which of the treatment methods
offered in the present study was believed to be the most
effective, is analyzed separately. A score of one (1) was
assigned to item #4 if the treatment checked was the same as
the subject's treatment assignment; a score of zero was
assigned if the response differed from the subject's
treatment assignment. Item #4 responses were included in
the analyses in various ways, but were not included as part
of the perceptions of treatment efficacy measure's analysis.

£1oo;om_§o§i§joo§iog. Satisfaction with the program
was assessed using the "Program Acceptance" questionaire.
Items #1 - #11 were designed to measure program
satisfaction. The items were drawn from Mavis (1987). Only
those items with a cluster analysis coefficient of .40 or

above in Mavis' (1987) analysis were chosen for the present

118

study. The eleven items were found by Mavis to have an
internal consistency coefficient of alpha = .86. The
internal consistency of the program satisfaction scale
obtained with the present sample was alpha = .83.

The eleven program satisfaction items were scored on a
five point response scale (0 - 4). Data are reported as
average responses to the eleven items.

2:og;om_3ooo§onoo. Psychological reactance to the
study was assessed using items #12 - #15 on the "Program
Acceptance" questionaire. These items were drawn from Mavis
(1987), having met the criteria of possessing a cluster
analysis coefficient of .40 or above in Mavis' analysis.
The four items were found to have an internal consistency of
.79 with Mavis' (1987) sample. Analysis of the reactance
scale with the sample in the present study yielded an
internal consistency alpha of .57.

The reactance items were scored on a five point
response scale. Data are reported as average responses to
the four items.

Dooirobiiity_of_goo§:_i. The relative desire for
control was assessed in this study using the "Personal
Preferences" questionaire. This questionaire is the
Desirability of Control Scale of Burger and Cooper (1979).
Using 453 undergraduate students, Burger and Cooper (1979)
found the Desirability of Control Scale to possess an

internal consistency of .80. Thirty-one subjects were

119

retested at six weeks, yielding a test-restest reliability
coefficient of .75. The responses of Burger and Cooper's
(1979) norming group subjects had a low (r = .11) and
insignificant correlation with a social desirability
response set. The internal consistency estimate of
reliability for the Desirability of Control Scale obtained
with the sample in the present study was alpha = .81.

The twenty item questionnaire is scored on a seven-
point Likert type response scale (1 = doesn't apply to me at
all to 7 2 always applies to me). The questionnaire is
scored in the direction of desire for control. Items #7,
#10, #16, #19, and #20 are reversed for scoring. Data are
reported as total scores for the entire scale.

Woioh§_Looo§_oj_§og§;ol. Weight loss relevant locus of
control attributions were assessed using the ”Beliefs about
Weight Loss“ questionaire. This questionaire is the Weight
Locus of Control Scale presented by Saltzer (1982). Using a
group of 110 college undergraduates, Saltzer (1982) found a
test-restest reliability of .67 (p < .01) for the WLOC scale
at a twenty-four day posttest. The internal consistency of
the WLOC scale was measured by Saltzer (1982) as alpha =
.58. The WLOC scale has a small and insignificant
correlation (r = -.03) with a social desirability response
set (Saltzer, 1982). Using the sample in the present study,
the WLOC scale was found to possess an internal consistency

estimate of reliability of alpha = .69.

120

Items #1 and #4 are internally worded while items #2
and #3 are externally worded. The scale is scored in the
external direction with each item ranging from 1 (strongly
disagree) to 6 (strongly agree) for the externally worded
items and reverse scored for the internally worded items.
Data are presented in terms of the total scale score.

e ' esc ' v ed ctor Var'a es.
Demographic data were assessed in the present study using
the "Personal Information" questionaire. Items #1 through
#6 assessed the subject's age, sex, marital status,
educational level, occupational status, and ethnicity. Item
#8 assessed the age-at-onset for obesity. Item #10 assessed
the subjects' family weight history; each question within
item #10 was scored as either 0 = not overweight or 1 =
overweight.

Item #11 was designed to assess perceived social
support. The items within question #11 were drawn from
Mavis (1987) and selected for meeting the criteria of having
a cluster analysis coefficient of .40 or above.

Items #9 and #12 were designed to assess the subjects'
prior history of weight loss efforts. Item #9 asked the
subject about the frequency of past attempts at weight
reduction while item #12 asked the subject about the types

of weight reduction methods used in his/her past efforts.

121

t es'

Overall, the present study represents a 2 X 3 Analysis
of Variance Design. Factor A has two levels: choice of
treatment (choice) versus random assignment to treatment
(assigned). Factor B has three levels: nutritional
education (NE), behavior management (BM), and exercise
training (ET). Subjects were randomly assigned to levels of
factor A. Within level 1 of factor A, subjects were allowed
to choose the level of factor B for their weight loss
treatment. Within level 2 of factor A, subjects were
randomly assigned to levels of factor B.

The primary research hypotheses (1 - 8), regarding the
role of choice of treatments, reflect the 2 X 3 ANOVA design
in a straightforward manner. The secondary level hypotheses
are highly exploratory in nature and represent deviations
from the basic model. Hypothesis nine adds a third factor
to the model: desirability of control is added in a
categorical fashion (high vs. low scores, split at the
median), rendering a 2 X 3 X 2 design. Hypothesis ten,
concerned only with subjects within level one of factor A,
represents a single factor design with three treatment
levels. Finally, hypotheses eleven and thirteen are not
concerned with the effects of either factor A or factor B,
but represent correlational analyses across all subjects

regardless of experimental condition.

122

W

1- Attrition

Null: There will be no difference in attrition rates
between subjects in level one of Factor A and level two of
Factor A.

Alternative: There will be a difference in attrition
rates between subjects in levels one and two of factor A.

Symbolically: H0: Zri. - zrz. = 0

“1‘ Tr1. " T7‘2. 9" 0
where 1. = the proportion of subjects dropping out of
level one of factor A and 2. = the proportion of subjects
dropping out of level two of factor B.

2- W-

Null: There will be no difference in weight loss
between subjects in level one of factor A and level two of
factor A at the end of the program.

Alternative: There will be a mean difference in weight
loss between subjects in levels of factor A at the end of
the program.

Symbolically: Ho: “1. - u2. = 0

H1: ul. - “2. # 0
where u1_ = mean weight loss of subjects in level one of
factor A, and “2. = mean weight loss of subjects in level

two of factor A.

123

3- Attonoonoe-

Null: There will be no difference between subjects in
level one and level two of factor A on attendance.

Alternative: There will be a difference in attendance
between subjects in level one and level two of factor A.

Symbolically: Ho: u1 - “2. = 0

H1: “1. - “2. s 0
where “1. = mean attendance for subjects in level one of
factor A, and “2. = mean attendance for subjects in level
two of factor A.

4. meromo-

Null: There will be no difference on post-program
adherence scores between subjects in level one of factor A
and level two of factor A.

Alternative: There will be a difference on adherence
between subjects in levels one and two of factor A.

Symbolically: Ho: “1' - “2. = 0

H1: “1. - “2. # 0
where “1. = mean adherence score for subjects in level one
of factor A, and “2. = mean adherence score for subjects in
level two of factor A.
5. WWW-
Null: There will be no difference on measures of

perceptions of treatment efficacy between subjects in level

one of factor A and subjects in level two of factor A.

124

Alternative: There will be a difference on the measure
of perceptions of treatment efficacy between subjects in
levels one and two of factor A.

Symbolically: Ho: “1. - “2. = 0

H1: “1. - “2. ¢ 0
where “1. = mean perceptions of treatment efficacy score for
subjects in level one of factor A, and “2. = mean
perceptions of treatment efficacy score for subjects in
level two of factor A.

6. Ezoogom satisfaotioo.

Null: There will be no difference on measures of
program satisfaction between subjects in level one of factor
A and level two of factor A.

Alternative: There will be a difference in program
satisfaction between subjects in levels one and two of
factor A.

Symbolically: Ho: u1_ - ”2. = 0

H1: “1. - “2. ¢ 0
where u1.= mean program satisfaction scores for subjects on
level one of factor A, and u2.= mean program satisfaction
for subjects in level two of factor A.

7- Boootontt-

Null: There will be no difference on the measure of
reactance between subjects in level one of factor A and

level two of factor A.

125

Alternative: There will be a difference in program
reactance scores between subjects in levels one and two of
factor A.

Symbolically: Ho: “1. - “2. = 0

H1: “1. - “2. + 0
where “1. = the mean program reactance score for subjects in
level one of factor A, and “2. = the mean program reactance
score for subjects in level two of factor A.

8. W1-

Null: There will be no differences on self-efficacy
scores between subjects in level one of factor A and level
two of factor A.

Alternative: There will be a difference in self-
efficacy scores between subjects in levels one and two of
factor A.

Symbolically: Ho: “1. - “2. = 0

H1: “1. - u2_ ¢ 0
where “1. = the mean self-efficacy score for subjects in
level one of factor A, and u2.= the mean self-efficacy score
for subjects in level two of factor A.

9- WW-

Null: There will be no interaction between scores on
the Desirability of Control Scale and membership in levels
of factor A for the outcome variables assessed in hypotheses

two through eight.

126

Alternative: There will be an interaction between
Desirability of Control and levels of factor A for the
outcomes assessed in hypotheses two through eight.

Symbolically: Ho: (os7)j1 = 0

H1: «x37)j1 ¢ 0
where j = levels of factor A, and 1 = levels of the
Desirability of Control factor.

10. WW.

Null: For subjects in level one of factor A, there will
be no difference on locus of control scores between subjects
in the three levels of factor B.

Alternative: Locus of control will interact with the
type of weight loss strategy chosen for subjects in level
one of factor A. Specifically, subjects who choose exercise
training will be significantly more internal than subjects
who choose other treatments.

Symbolically: Ho: 1/2(u11 + “12) - u13 = 0

H1: 1/2(u11 + “12) - u13 ¢ 0
where “11 - mean locus of control score for nutritional
education subjects in level one of factor A, u12 = mean
locus of control score for behavior management subjects in
level one of factor A, and u13 = mean locus of control score

for exercise training subjects in level one of factor A.

127

11. Woight Loouo ot goo roi.

Null: There will be no relationship between weight
locus of control scores and outcome measures of attendance
and weight loss.

Alternative: Internality on the weight locus of control
measure will be positively related to attendance and body
weight change across all subjects.

Symbolically: Ho: p = 0

H1: p ¢ 0
where p = the correlations of weight locus of control with
attendance and the body weight change measures.

12. IIEQIEEEE_IYE§-

Null: There will be no significant differences in
weight loss between subjects in levels of factor B.

Alternative: There will be differences in weight loss
between subjects in levels of factor B.

Symbolically: Ho: u.1 - u.2 - u.3 = 0

31‘ “.1 ' “.2 ‘ “.3 * 0
where 11.1 = mean weight loss for subjects in the nutritional
education treatments, 11.2 = mean weight loss for subjects in
the behavior management treatments, and 11.3 = mean weight
loss for subjects in the exercise training treatments.

13. History of woioht loso etfiorts.

Null: Past experience with weight loss efforts will not

be related to scores on treatment outcome measures.

128

Alternative: The frequency of past efforts in weight
reduction will be related to scores on treatment outcome
measures across all subjects.

Symbolically: Ho: p = 0

H1: ‘p ¢ 0
where p = the correlations between frequency of past weight

loss efforts and the outcome measures.

Ahalysis or the Qato

Categorical outcome data associated with hypothesis one
are analyzed with chi-square tests of association and
Marascuilo post-hoc multiple comparisons (Glass and Hopkins,
1984).

The 2 X 3 factorial nature of the research design for
hypotheses two through eight, with multiple outcome
measures, lends itself readily to a multivariate analysis of
variance (MANOVA) analytic approach (Hand and Taylor, 1987:
Bray and Maxwell, 1982). As Bray and Maxwell (1982)
indicate, when the research questions relate to the effects
of treatments on several criterion variables, significant
MANOVA tests are followed by univariate analyses of variance
(ANOVAs) which, in turn, are followed by multiple comparison
tests. Thus, the overall strategy for analytic treatment of
the data for hypotheses two through eight and hypothesis
twelve is as follows: (1) MANOVA tests, followed by (2)

ANOVA tests, and (3) multiple comparison tests.

129

MANOVA tests are grouped by the time of assessment:
post-program measures, mid-program measures, and pre-program
measures are analyzed separately. A similar strategy is
used for testing hypothesis nine on the same group of
outcomes, within the 2 X 3 X 2 model of that hypothesis.
While other strategies for grouping or clustering measures
may be possible, there is no a priori theory or evidence
that would guide such alternative analytic strategies.

Hypotheses regarding variables with repeated measures
are subsequently analyzed with the repeated measures
analysis of variance procedure for the univariate tests
(Glass and Hopkins, 1984). Multiple comparisons between
treatment types are analyzed with the Scheffe post-hoc
comparison method (Glass and Hopkins, 1984). Since the
primary questions of interest in the present study relate to
factor A, the lack of random assignment to levels of factor
B for subjects in A.1 will not affect the analysis of the
data for hypotheses regarding the role of choice of
treatment type. Comparisons between levels of factor B will
be compromised by the lack of random assignment for subjects
in level #1 of factor A.

The use of correlational and multiple regression
procedures are used sparingly in the analysis of the data
from the current study to illuminate relationships among the
data that could not otherwise be seen with the analysis of

variance procedure. Specifically, a simple multiple

130

regression model is used in the analysis of hypothesis two,
and correlational data are reported for hypotheses eleven
and thirteen.

The significance level of alpha = .05 is used for all
hypothesis tests in the data analysis. As all the
hypotheses are highly speculative, two-tailed tests of

significance were used.

CHAPTER IV

RESULTS

The results of the current research are presented in
two sections: (a) a determination of the comparability of
the groups at the beginning of the experiment and, (b) tests
of the hypotheses stated in Chapter III.

Com b' 't t e G o

The initial research assessment questionaire,
administered at the first program meeting, was used to
determine the similarity among experimental groups. The
questionaire gathered background and demographic
information, data regarding prior attempts at losing weight,
perceived self-efficacy, perceptions of treatment efficacy,
weight locus of control, and the desirability of control. A
summary of group comparisons based on analysis of variance
(ANOVA) are presented in Table 1: similar comparisons based
on Chi-square analyses are shown in Table 2.

Although full random assignment was constrained due to
the nature of the experimental design, only four significant
differences were discovered. A significant difference
between subjects in the three treatment types was found for

number of overweight family members (F [2,89] = 3.09, o <

131

132

_c. v a ««

 

 

mo. v a a
b_.o om._ co.c o~.n ao.m am.~ ~_.m oc.n co.n soeouuum
ucoaueouh no occuuaoooma
mm.o «amc.m ¢~.o oo.m mm.e mm.q m~.n om.¢ __.e socoacem C_om
m~.o ac.o qo.c m.aq _._e ~._e _.cq b.¢¢ _.~< uzmuozuo>o Devotee
-.c no.o No.c ~.mm m._n _.~m ~.~n c.~n ~.nn xoozu mam: soon
n~.o ~_.o c_.c m.mo~ c.~m_ ~._o~ o.na_ c.oo_ _.oa_ Ana—v Lsmuoz zoom
-._ mm.~ ~m.~ mm.n mn.n oc.< ~m.n mo.m co.q monouum mac; uzwqoz
o:o«>oua to xuouum>
mm.o «-.n cm.c m~.~ oc.~ sa.~ mm.~ an.~ so.~ aboaanm amuoom
eo.o «so.n -.c ee.n 58.. oo._ 35.. ~_._ ns.. mousse: knosar
utmqoluo>3 .o Logan:
a~._ «n.o .o.o n.m. o._~ n.s~ m.¢~ e.o_ ~.- suauoao to amaze on use
mn._ ~¢._ o~._ n.o_ ~.m_ n.o_ ~.m_ ~.c~ o.n_ Aoo>v acqoaoaem
me._ oe.o ~n.o ~.o¢ s.n¢ «.me «.mc ~.c¢ o.qc om<
cauuosuuucu unusuaoue oouogo PM In mz Hm z: a: o_;a_tc>

nouucm uocwgmm< oouozo

moduoz m couuupcoo

 

 

ceauupcoo unusuauuomxm an acqumz m use acme: omouooa

. m4m<8

133

 

co.o

qm.~

 

cauuocuoucu

UCSUQOHP CUdO—nu

o~.i no.6 on.~ «6.8 souucoo to undo; name»:
~..c_ o._o_ n.oo n.no douucoo mo Asa—«bouuaon
mz Pm In NZ odomuum>

cowoso
couuutcoo

 

 

couuuvcou ~a»:o§«uoaxm an nouuaa m use memo: amououa

Aamszaezooo _ mee<s

134

 

 

 

 

 

Nam N5 N5 Nb- um. um
n _ _ n c _ oo.;3ncoz
use Nna «am new Nwm N_a
o- c— c. c— w— c— sous:
ANN. I m..n I Mm .sm.~ I mmv tcsouwxocm o‘cssm
Nam nan NNN nan NQ_ «an
n m c __ q 0 counts: ooz
Mme «so ans ch Nmm was
a o_ - m a. n mouths:
2:. .. m .m: we .2.o n «.5 253m 13:62
an NM~ NON NG— nm_ N¢
m ~ n N e _ 0.x:
«55 Nun now use Nnm Nqa
o. m— N. s— m. o. oncsom
3a. .. a .n .- one .3; u wwv .253
Pm tn m2 PM In mz mabmuum>
vocmuea< couczo
coda—tcoo

 

 

cauuuvcou aqucos«uoaxm as nausea»: Icudmsmnoum so canoe ouczgmu_:o

N mqa<b

135

 

 

 

 

NA. Nna NNN «no use use
N a. a N. .N N am»
Now NN "Ne NNn NA Nen
__ . o. e _ G oz
Aooo. . m..m - mm .N_.nn . va scoscmamm< to Nm>ocaa<
N.» NNN NoN NeN NNN Nos
3 c N m N A m. cm>o
um NoN N3. N6 NS. No
_ N N o N o m. - __
use NNG Nam neN NS No.
a N m n _ N o. - 9
Nn_ NN Nan Nee Now Non
N _ m a __ s n - _
ANmo. a m..n_ . mm .mo.NN . mmv age-u.u< «no; Samoa: asod>mta to coasaz
em :a mz em as m2 «Seasca>
vocmuua< acqozo
co_ooucoo

 

 

acoNqucoo Naucosuuomxmnwm.uouaoeoz Iouuoumaoum co ounce ouoawmuuzo

Namazaszooo N mnm<s

136

0.05), degree of social support (3 [2,89] = 3.22, o <
0.05), and self-efficacy (3 [2,89] = 5.63, o < 0.01).
Subjects in the nutritional education groups had a slightly
lower number of overweight family members (mean = 1.08) than
did subjects in the behavior management groups (mean = 1.73)
or the exercise training groups (mean = 1.63). Subjects in
the behavior management groups had a mean social support
rating of 2.54, which was significantly lower than the
rating for subjects in the nutritional education groups (M =
2.73) and subjects in the exercise training groups (M =
2.81). The nutritional education groups had an initial mean
self-efficacy score of 4.19, which was significantly lower
than the scores for the exercise training groups (M = 5.54)
or the behavior management groups (M = 4.89).

An additional pre-test difference between the
experimental groups was found on subjects' approval of their
treatment type assignment. Sixty-four percent of the
subjects who chose nutritional education and 68% of the
subjects who chose exercise training thought that their
treatment type was the most effective means for them to lose
weight. These percentages were nearly reversed for subjects
who were assigned to treatment type: 67% of the assigned
nutritional education subjects and 85% of the assigned
exercise subjects did not think their assigned treatment was
the most effective means for them to lose weight. This

pattern of reversal on approval of treatment assignment did

137

not hold true for subjects in the behavior management
treatments. Ninety-five percent of the subjects choosing
behavior management and 93% of the subjects assigned to
behavior management thought their treatment type was the
most effective means for them to‘lose weight.

The likelihood of obtaining four significant results
over a series of 44 comparisons at a probability level of
.05 is approximately 20% (Sakoda, Cohen, and Beall, 1954).
Given that the chances of these findings did not reach the
.05 level of significance, there is some likelihood that the
differences identified between experimental groups at the
first program measure are due to chance. Further, the
differences found on number of overweight family members,
social support, and self-efficacy were between treatment
types and not between the levels of factor A (choice
condition); hence, they do not affect the primary research
hypotheses. The differences found regarding approval of
treatment type assignment may have more interpretive
significance and are discussed elsewhere.

0 es

A probability level of .05 was used as the criteria for
significance for each of the hypotheses tested. Statistics
and statistical tests were calculated using SPSS programs.
Sample sizes used in the hypothesis tests vary according to
the number of subjects remaining in the program at the time

relevant assessments were made.

138

Tests relevant to attrition are presented first as
attrition patterns affect the remaining hypothesis tests,
and because these tests are Chi-square analyses while the
majority of remaining tests are ANOVAs. Following an
analysis of the attrition data, MANOVA results pertaining to
hypotheses two through eight are presented as a group.
Subsequent follow-up ANOVA and multiple comparison tests, as
well as other pertinent analyses, are presented for each
hypothesis separately in order of their listing in Chapter
III.

hypothesis One

The first hypothesis stated that subjects who receive a
choice of weight loss treatment types will differ
significantly from subjects who were assigned to a type of
weight loss treatment on rate of attrition.

Attrition in the current study occured at two levels or
times: subjects who dropped out of the program at the first
program meeting and subjects who dropped out sometime after
the first program meeting during the course of the ten week
program. Subjects who dropped out at the first program
meeting did not fill out the initial research questionaire,
but were contacted by phone and verbally administered the
questionaire regarding reasons for dropping out. Subjects
who remained in the program following the first meeting were
defined as drop outs if they did not attend the last program

meeting and complete the final research assessment

139

questionnaire. These subjects were contacted by mail and
asked to return the questionnaire assessing reasons for
dropping out.

Analyses are presented separately for attrition at the
first program meeting and attrition during the course of the
program, and then combined for an analysis of total program-
related attrition.

The distribution of drop outs among the experimental
groups is shown in Table 3. A total of 23 subjects dropped
out of the program at the first program meeting, for an
attrition rate of 19.5%. A total of 22 subjects dropped out
during the course of the program, or 23.2% of the 95
subjects remaining in the program after the first program
meeting. The total program-related attrition was 45
subjects, for an overall attrition rate of 38.1%. To
determine if the attrition patterns were related to the
independent variables Chi-square tests were used.

Attrition rates at the first program meeting differed
significantly between subjects receiving a choice of
treatments and subjects assigned to treatments (M2 = 4.34,
ofi=1, p < .05). The attrition rate was 11.9% for the choice
condition and 27.1% for the assigned condition. Attrition
rates did not significantly differ at the first program

meeting for the three treatment types (M2 = 0.587, ot=2, o >

140

 

 

 

 

«.mc c.nm c.oo n.- o._N ~.oc u N
a N N. o m b u z
A. I we .cmo.n I .mv noouam Damascus?
A. I ma .Imn.¢~I wmv acuoum moaoru czuofitss< _:.:e
_.m~ m.m_ m.ee w.m~ o.n_ c.om u N
n N u n n e u 2
Aw I mu .a~o.~ I my acuoam ucoaumous
A. I ud .~O._NI mwv scuoom 004050 amsmczi oz. quoza
o._n o.m~ o.m~ o.m~ n.m v.n~ u N
o m n n a w . u z
Aw I mu .smn. I mnv scuooh ucoauaoue
A. a we .«sn.e a may tongue ouNoao manna»: .mc_r 3<
am to NZ km a: mz u.:a..a>
vocwfimn< OONOEU
cofiuutcoo

 

 

lol'l’l'

sawuwuuu< poumdom I Ecwmwta .3329

 

van unannoum on» No onusoo orb w:«~:n.wur«uooz Ecuuoua uauwm «so on couuuuuu< mo m«am.m:< oumzrmnozu

a. 3.34:.

141

.05). The attrition rates for the nutritional education,
behavior management, and exercise training treatments were
21.2%, 15.9%, and 22.0%, respectively.

Rates of attrition during the course of the program did
not differ significantly between subjects receiving a choice
of treatments and subjects assigned to treatments (M2 =
1.01, disl, p > .05). The attrition rate during the course
of the program was 19.2% for the choice condition and 27.9%
for the assigned condition. Rates of attrition during the
course of the program did significantly differ between the
three treatment types (32 = 7.67, ofi=2, o < .05). The
attrition rates for the nutritional education, behavior
management, and exercise training treatments during the
course of the program were 42.3%, 13.5%, and 18.8%,
respectively. The nutritional education treatment clearly
lost the greater percentage of subjects.

Total program-related attrition was significantly
different between subjects receiving a choice of treatment
and subjects assigned to treatment (M2 = 4.38, QIPI. p <
.05), as well as between the three treatment types (32 =
5.98, ot=2, p < .05). The total (combined) attrition rate
for subjects receiving a choice of treatments was 28.8%,
whereas the total attrition rate for subjects assigned to
treatments was 47.5%. The total attrition rates for the
nutritional education, behavior management, and exercise

training treatments were 54.5%, 27.3%, and 36.6%

142

respectively. A graphic presentation of total attrition in
the experiment can be found in Figure 1.

Comparisons between treatment types on attrition during
the course of the program and total program-related
attrition are presented in Table 4. The statistical method
for the multiple comparisons regarding attrition was the
Marascuilo method, as suggested by Glass and Hopkins (1984).
The nutritional education treatment had a significantly
greater attrition rate than did behavior management both
during the program (M2 = 6.58, orez, o <.05) and for total
attrition (32 = 6.17, ot=2, o <.05). The nutritional
education treatment's attrition was also greater than the
average of the other treatments during the course of the
program (M2 = 6.00, orez, p <.05). The attrition rate of
the nutritional education treatment was not significantly
different from the exercise training treatment, nor was the
difference between behavior management and exercise training
significant.

Table 5 presents a summary of the reasons given for
dropping out of the program at the first meeting (N = 23).
While 56.5% of the program attrition at the first meeting
was due to schedule conflicts, 84.6% of the subjects
dropping out because of schedule conflicts were in the
randomly assigned groups. 34.8% of the subjects dropping
out at the first program meeting did so because of

unhappiness with the monetary incentive scheme.

62200 3 .5553 80.696 com: 3 v.50:

mob Eocboot.

 

143 ‘

.D Jm M2
020:0 w... ale 0
pocm_mm< n ale
10?
1ON
10m
Tog
Tom
100

 

ION.

U0!l!JllD queued

TABLE 4

144

Multiple Comparisons of Attrition Between Weight Loss Methods During

the Course of the Program and Total Attrition

 

Chi-Square Value

 

Contrast During the Program Total Attrition
NE - ET 3.89 2.43

NE - BM 6.58* 6.17*

BM - ET 0.35 0.86

NE - % BM - % ET 6.00* 5.09

 

*p<.05

TABLE 5

145

Reasons for Attrition at the First Program Meeting

 

Percent of

 

 

 

 

 

Reason Choice Assigned Attrition
Did not like type of program 0 2 8.7
Schedule conflicts 2 11 56.5
Monetary incentive scheme 5 3 34.8
TABLE 6
Reasons for Attrition During the Program

Percent of
Reason Choice Assigned Attrition
Did not like type of program 0 O 0
Did not like program leader 0 0 0
Program was not helping l 0 4.8
Schedule problems 2 5 33.3
Unanticipated life events 6 2 38.1
Not ready to lose weight 2 0 9.5
Other 1 2 14.3

 

146

After two mailings, 62% of the subjects dropping out
during the course of the program returned their assessments:
70% of the subjects receiving a choice of treatments and 50%
of the subjects who were assigned to treatments returned the
questionaire. Table 6 presents a summary of the reasons
given for dropping out during the course of the program.
Over 70% of the respondents indicated that schedule problems
or unanticipated life events precipitated their dropping
out. Two individuals felt they were not personally ready to
lose weight and one individual felt the program was not
helping her to lose weight. Three individuals used the
"other" category to indicate their reasons for dropping out.
Two of these subjects dropped out because they moved out of
town, and one subject said she "was picking up the bad
habits people were talking about." A comparison of all
subjects dropping out of the program at any time because of
schedule conflicts versus subjects dropping out for all
other reasons combined was not significant (32 = 4.85, g;=2,
p >.05).

Subjects who remained as participants in the program
were contrasted with the subjects who dropped out during the
program on available demographic and outcome measures. A
summary of these comparisons can be found in Table 7 and
Table 8. Subjects who remained as participants in the

program had significantly more overweight family members

TABLE 7
Means, Standard Deviations,

During the Program

147

and T-Tests for Participants Vs. Drop:0uts

 

Means (S.D.)

 

Drop-Outs Participants Sig.
Variable N - 22 N - 73 T* of T
Age 42.6 ( 8.8 ) 42.6 ( 8.9 ) 0.02 .98
Education (yrs) 16.1 ( 2.7 ) 14.9 ( 2.6 ) -1.78 .08
Age at Onset 20.5 (12.1 ) 21.7 (12.4 ) 0.37 .71
Number of Overweight
Family Members 0.68 ( 0.95) 1.8 ( 0.91) 4.88 .00
Social Support 2.74 ( 0.35) 2.66 ( 0.50) -0.68 .50
Variety of Previous
Weight Loss Attempts 4.18 ( 2.9 ) 4.19 ( 2.4 ) 0.02 .99
Attendance 3.1 ( 1.5 ) 9.0 ( 1.1 ) 19.87 .00
Adherence 2.9 ( 2.4 ) 11.6 ( 3.8 ) 10.07 .00
Self-Efficacy Pre 4.2 ( 1.4 ) 5.1 ( 1.6 ) 2.41 .02
Desirability of Control 102.9 (14.7 ) 100.0 (13.6 ) -0.87 .39
Weight Locus of Control 7.8 ( 3.2 ) 7.15 ( 3.4 ) -0.82 .41
Perceptions of Treatment
Efficacy: Pre 3.1 ( 0.59) 3.1 ( 0.46) 0.05 .96
Program Satisfaction: MID** 3.35 ( 0.47) 3.43 ( 0.46) 0.44 .66
Program Reactance: MID** 1.32 ( 1.3 ) 0.69 ( 0.68) -2.14 .04

 

* df 8 93

** Drop-out N 8 7, df = 78

148
TABLE 8

Chi-Square Analyses of Participants Vs. Drgp-Outs

 

 

 

Percent of Percent of
Variable Participants Drop-Outs
Gender (32 =1.83, g -1, p :- .18)
Female 87.7 72.7
Male 12.3 27.3

Marital Status (32 = 1.46, _j = 1,.2 = .23)

 

Married 67.1 50.0
Not Married 32.9 50.0

Ethnic Group (g? = 1.11, dj 8 1,‘p a .29)

 

White 89.0 77.3
Non-White 11.0 22.7

Number of Previogs Weight
Loss Attempts (__ 8 3. 79, df = 3,‘2 = .29)

 

1 - 5 35.6 27.3
6 — 10 28.8 22.7
11 - 15 11.0 4.5
Over 15 24.7 45.5

Approval of Treatment Assignment (12 - .00, g; = 1, p = 1.0)
No 34.2 36.4

Yes 65.8 63.6

 

149

than did drop outs (i = 4.88, g;=93, p <.01). In addition,
participants had higher scores on the self-efficacy pre-test
(I = 2.41, g;=93, p =.02) and lower scores on the program
reactance mid-program measure (I = -2.14, g;=73, p =.04).
The participants also had higher means on attendance and
adherence measures, however these differences represent a
measurement artifact. It is also important to note that
there were no differences between participants and drop outs
on whether or not they approved of their particular
treatment method (X2 = 0.00, g;=1, p = 1.0).

In conclusion, with regard to total program-related
attrition, subjects who received a choice of treatment type
were less likely to drop out of the program than were
subjects randomly assigned to treatment type. The
differences were larger than would be expected by chance.
Thus, hypothesis one, regarding the effect of the choice
condition in affecting attrition rates, was supported.

'v ' s s

The dependent variables associated with hypotheses two
through eight were tested simultaneously with the
multivariate analysis of variance (MANOVA) procedure prior
to univariate analyses (Hand and Taylor, 1987: Bray and
Maxwell, 1982). Separate MANOVA tests were conducted for
(1) the seven post-test dependent measures (body weight
change, attendance, adherence, self-efficacy, perceptions of

treatment efficacy, program satisfaction, and program

150

reactance), (2) the four mid-program dependent measures
(self-efficacy, perceptions of treatment efficacy, program
satisfaction, and program reactance), and (3) the two pre-
test dependent measures (self-efficacy and perceptions of
treatment efficacy). Only one of the three body weight
change indices was used in the MANOVA test (pounds lost) to
reduce redundancy and artifactual clustering in the data.

The MANOVA procedure aggregates properties of the
multiple dependent measures by combining them into a linear
composite. MANOVA is similar to an ANOVA, but analyzes
multiple outcome measures simultaneously as a group
composite. Four common test statistics are used for
reporting MANOVA results: the Pillai-Bartlett trace, Wilk's
lambda, the Hotelling trace, and Roy's largest eigenvalue
(Hand and Taylor, 1987). The four test statistics are
highly interrelated and there is insufficient evidence to
make statements about their relative power (Hand and Taylor,
1987). The Hotelling E-ratio is used to present the results
of the current analyses, though it should be noted that the
probability values for the other test statistics did not
differ significantly from those associated with the
Hotelling E.

Table 9 presents a summary of the MANOVA tests for the
seven post-test outcome measures. There was a significant
difference on the group of post-test outcome measures

between subjects receiving a choice of treatment type and

151

TABLE 9

Multivariate Analysis of Variance on Post—Program Measures

 

 

 

Source df Hotellings F Sig. of F
Choice Factor (A) 7,60 6.90 .000
Treatment Factor (8) 14,118 1.20 .288

A x 8 Interaction 14,118 1.21 .277
TABLE 10

Multivariate Analysis of Variance on Mid-Program Measures

 

 

Source df Hotelling: F Sig. of F
Choice Factor (A) 4,71 2.01 .102
Treatment Factor (3) 8,140 ' 1.46 .178

A x 8 Interaction 8,140 1.58 .137

 

152

 

 

mam. om.o q~_.¢ couuomuoucw m x <
coo. _a.m q-.q Ame tongue acmeumoue
coo. Nq.o mm.~ A<V wouomm oouoso
m we .wwm m mwcfiaamuoz up mousom

 

 

monommoz Emuwoumamum co oocmfium> mo mumzamc< osmium>auasz

a" mqm<h

153

subjects assigned to treatment type (EH = 6.90 [7,60],
p<.001). The MANOVA was not significant for treatment type
differences (p =.288) or a treatment type by choice
condition interaction (9 = .277).

Table 10 presents a summary of the MANOVA tests for the
four mid-program outcome measures. No significant
differences were found on the group of mid-program measures
for the choice factor (9 =‘.102), the treatment type factor
(Q = .178), or their interaction (p = .137). Table 11
presents a summary of the MANOVA tests for the two pre-test
outcome measures. While there were no significant
differences on the group of pre-test measures between levels
of factor A (p = .660) or for the factor A by factor B
interaction (3 = .839), the treatment types did differ
significantly from each other on this group of measures
(EH 2 3.71 [4,174], 9 = .006). As can be seen in Table 1,
the treatment groups differed in their initial self-efficacy
scores.

The results of the MANOVA analyses warrant further
exploration of the differences between levels of the choice
versus random assigned factor as well as between levels of
the treatment type factor.

mm

The second hypothesis stated that there would be a

significant difference in weight loss between subjects

receiving a choice of treatments and subjects assigned to

154

treatment types. Three measure of body weight were used in
the current research: pounds, body mass index (kg/cmz), and
percent over ideal body weight. Correlations of the three
measures, using the initial body weights obtained at the
first program meeting, suggest that the three measures are
highly interrelated. The correlation between pounds and
body mass index (BMI) is r = .84, r = .89 between pounds and
percent overweight, and r = .99 between BMI and percent
overweight. Results are presented in terms of all three
indices of body weight.

Table 12 presents the cell means and standard
deviations for the three body weight indices at the pre-test
measure, the post-test measure, and for the mean change
between pre- and post-test measures. Tables 13, 14, and 15
present summaries of the ANOVAs on body weight change for
pounds change, BMI change, and percent overweight change,
respectively. The overall results are highly similar for
each of the three body weight change indices. In each case,
the subjects who were randomly assigned to treatment types
lost more weight by the end of the program than did subjects
who had a choice of weight loss treatments. Figure 2
presents a graphic representation of weight loss trends (for
factor A) as measured in pounds. The probabilities
associated with the E-ratios differed slightly between the
three indices, with Q < .05 for pounds and percent

overweight and p = .054 for BMI. For subjects receiving a

155

TABLE 12
Cell Means of Body Height Measures at Pre-Test, Post-Test. and Changg

at the End of the Program

 

 

 

 

Condition
Choice Assigned
Measure NE 8M ET NE 8M ET
Pounds
Pre 196.1 196.8 195.6 201.2 192.0 205.3
(19.1)* (37.7) (44.9) (48.8) (40.7) (53.4)
Post 197.9 187.5 184.1 185.5 183.2 195.0
(20.6) (32.3) (42.0) (53.0) (40.1) (51.0)
Change + 0.14 - 5.8 - 5.1 - 7.5 - 8.4 - 5.9
( 5.2) ( 4.1) ( 4.0) ( 9.3) ( 5.2) ( 7.5)
Body Mass Index
Pre 33.1 33.4 32 2 32.1 31.8 33.2
( 4.3) ( 5 7) ( 6 0) ( 5.9) ( 6.7) ( 7.8)
Post 32.9 31.1 30.7 28.6 29.3 31.9
( 2 9) ( 5 3) ( 5.5) ( 3 8) ( 4.9) ( 7.3)
Change - 0.01 - 0.97 - 0.88 - 1.10 - 1.3 - 1.0
( .87) ( .72) ( .67) ( 1.41) ( .89) ( 1.3)
Percent Over Ideal Height
Pre 47.1 44.6 44.1 41.7 41.1 48.9
(17.4) (25.6) (28.2) (27.4) (28.3) (37.5)
Post 47.6 38.6 37.1 25.3 31.7 42.6
(15.5) (22.6) (25.6) (17.8) (24.0) (35.1)
Change - 0.07 - 4.3 - 3.9 - 5.2 - 6.0 - 4.4
( 3.9) ( 3.2) ( 3.1) ( 6.5) ( 3.6) ( 5.3)

 

* Standard Deviation in Parentheses

156
TABLE 13

Analysis of Variance for Body Weight Chagge in Pounds

 

 

 

 

 

 

Source df Mean Square F Sig. of F
Choice Factor (A) 1 165.0 5.16 .026
Treatment Factor (B) 2 56.7 1.77 .177

A x B Interaction 2 54.7 1.71 .188
Error 67 31.9

TABLE 14

Analysis of Variance on Body Mass Index Chaggg.

Source df Mean Square F Sig. of F
Choice Factor (A) 1 3.41 3.85 .054
Treatment Factor (B) 2 1.77 1.99 .144

A x 8 Interaction 2 1.15 1.30 .281

Error 67 0.89

 

157
TABLE 15

Analysis of Variance on Percent Over Ideal Weight Change

 

 

Source df Mean Square F Sig. of F
Choice Factor (A) 1 73.6 4.41 .039
Treatment Factor (B) 2 29.9 1.79 .175
A x B Interaction 2 25.6 1.53 .224

Error 67 16.7

 

158

330.5 cmcgmmo .m> 86:0 to”. E9605

lemon pco loco so mccsoa E 296; com: ”N 930E

 

me. é...
. . 02010 n olo mm;
pmcmfim< H Elm.
10m:
Tmmv
Tome
1mm?
IOON

 

[mom

spunod ug iqbgem

159

choice of treatments, the means for pounds change, BMI
change, and percent overweight change are -4.5, -0.77, and
-3.4, respectively. For subjects assigned to treatment
types, the means for pounds change, BMI change, and percent
overweight change are -7.4, -1.17, and -5.3, respectively.
There is, however, a tremendous amount of variation in the
data. As can be seen in Table 12, the standard deviations
of the body weight change measures are generally as big as,
and sometimes bigger than, the means.
The nutritional education groups lost fewer pounds (M =

-3.9, . = 8.4) than did the behavior managment groups

(M = -6.8, sggg = 4.7) or the exercise training groups (M =
-5.4, sggg = 5.5). Similarly, the nutritional education
groups had a smaller BMI change (M = -0.58) than did the
behavior management groups (M = -1.1) or the exercise
training groups (M = -0.93). The nutritional education
groups also showed a smaller change in percent overweight
(M = -2.8) than did the behavior management groups (M =
-5.0) or the exercise training groups (H = -4.1). As seen
in Tables 13, 14, and 15, however, the z-tests for
differences between treatment types (factor B) were not
significant for pounds change (9 = .177), BMI change (9 =
.144), or percent overweight change (2 = .175). The choice
factor by treatment type factor interactions were also not

significant for pounds change, BMI change, and percent

160

overweight change (9 = .188, p = .281, and p = .224,
respectively).

The weight change data were also analyzed within a
multiple regression format to control for differences in
initial body weight. Table 1 reveals that there were
differences between the treatment groups' means on the body
weight measures at the pre-test assessment, even though
these difference were not significant. Specifically, the
subjects receiving a choice of treatment types had a mean
pre-weight of 3.05 pounds less than the subjects who were
assigned to treatments. Conversely, however, the mean pre-
BMI and pre-percent overweight was slightly higher for the
subjects receiving a choice of treatments (the differences
were 0.16 and 1.25, respectively). This suggests that the
subjects receiving a choice of treatments were slightly more
obese than were the subjects who were randomly assigned to
treatment types. Differences in initial body weight also
existed between the three treatment types. Subjects in the
behavior management treatment had a mean pre-weight of 194.8
pounds, while the subjects in the nutritional education
treatments had a mean pre-weight of 199.03 pounds and the
exercise training subjects had a mean pre-weight of 199.5
pounds. The mean initial BMI and percent overweight scores,
respectively, for the three treatment types were:
nutritional education (32.5, 44.0), behavior management

(32.1, 43.2), and exercise training (32.6, 46.1).

161

The multiple regression analyses on body weight change,
conducted to control for initial body weight, used the
"forced" method of entering variables in a specified order.
Summaries of the results of the multiple regression analyses
controlling for the initial body weight measures are
presented in Tables 16, 17, and 18 for pounds, BMI, and
percent overweight, respectively. As can be seen, the
results indicate that the body weight pre-measures did
account for a significant amount of variance when pounds and
BMI were the measures (though not when percent overweight
was the measure), but that the basic pattern of results was
consistent with the ANOVA procedure.

Specifically, the pre-measure accounted for 6.5% of the
variance with pounds as the measure (p = .03), 5.8% of the
variance with BMI as the measure (p = .04), but only 3.7% of
the variance when percent overweight was the measure (9 =
.102). Adding treatment type to the model accounted for
only an additional 3.9% of the variance when pounds was the
measure (9 a .228), 4.2% of the variance when BMI was the
measure (3 = .21), and 3.6% of the variance when percent
overweight was the measure (9 = .269). Thus, treatment type
played a less significant role in the final weight change
than did initial weight, at least when pounds and BMI were
the measures. When the two levels of the choice factor were
entered into the equations, the regression models improved

significantly. When added to the pre-weight and treatment

162

 

 

4_o. no.4 an." a~o. Noe. 66—. cos. tongue masozo

«mo. ao.n ee.~ oNN. one. eo_. mam. tongue acoesmaue

one. 25.. «6.4 one. woo. woo. MAN. assumecsoe

m to .mem Le a sequo>o wwameo a «mango m x e «segues: assastm>
uo .wmm N N

 

 

mmdbmwum> hoboupoum mm acuumm acosumuuh pcm awOuumm mouoso .usmwozuoue

 

pcm onncuuc> couuouuuo no Aucssomv gamma: avom cu mwcmzo sow: swohasc< sensuouuwm oumwu~sz no musessm

c— mam<a

 

163'

 

 

mic. no.4 _~.n mmo. awe. one. can. touuae wusozo

nee. ao.m mn.~ o_~. Nee. oo_. o_n. sesame oeusummue

oeo. _~._ on.s oqo. mac. awe. ~¢~. «pa-22a

a co .mam on a asuta>o «mango a macago m a e «masses: meamssm>
Lo .wmm N N

 

muououpoum mm hobomh mouoso pcm mucuomm ucoeomouh .AZQIoLm

 

ccm oabmuum> cowuouquu mm

xoccm mum: moon cuhwwmmzo saw: mwoxumc< caummouuux oddqo~sz mo Necessm

: mama;

164

 

 

Nno. no.4 .N.N «No. coo. Nm_. an. sesame ousoeo

Nm_. ac.» Nm._ ooN. one. nNc. .NN. tosses scosumute

Nos. _N._ «N.N No_. Nno. Nnc. ns.. 8.;

“uswumzho>o ucmosom

a No .NNm cu m ”Nata>o mmcmeo N manage a N e «Nassau: 6.3ch=>
co .wmm N N

 

 

mo~pc~um> sououpmum mm acuomm ou—ozu pcc acuumm acetamouh .wmmuoauo>o ucwoummlocd

 

can ofibmuum> co«umu«uo mm ucwwozuo>o ucooumm ea omccso so“: mqmzumc< caummowwwm mamuo_:z uo amassem

a. uqn<a

165

type variables, the choice condition accounted for an
additional 6.2% of the variance when pounds was the measure
(3 = .028), 5.9% of the variance when BMI was the measure

(9 = .033), and 6.4% of the variance when percent overweight
was the measure (2 = .028).

Although there were not significant results for the
treatment type factor (factor B) in either the ANOVA or the
multiple regression analyses, Table 19 summarizes multiple
comparisons between the treatment types for each body weight
measure. This data is presented for further elucidation of
the data and to provide support for tests regarding
hypothesis twelve. Figure 3 presents the relationships
between the treatment types on pre-test and post-test body
weight with pounds as the unit of measurement. As can be
seen in Table 19, there are no significant differences
between treatment types on any of the weight loss outcome
measures. The nutritional education versus behavior
management contrast is very close to significance on all
three measures, providing further support for the evidence
suggesting that the nutritional education treatment was the
least effective in promoting weight loss.

In conclusion, the majority of the data suggests that
the level of factor A did significantly affect weight loss
in the program. Subjects randomly assigned to treatments
lost significantly more body weight than did subjects who

had a choice of treatments. Thus, hypothesis two,

166
TABLE 19

Multiple Comparisons for Treatment Types on Body Weight Change Measures

 

Measure and

 

Comparison Contrast S.E. T Sig. of T
Lbs. Change

NE - BM -6.86 3.56 -1.93 .058

BM - ET 3.24 3.05 1.06 .293
BMI Change

NE - BM -1.21 0.59 -2.04 .045

BM - ET 0.40 0.51 0.79 .433

N

Overweight Change

NE - BM -5.00 2.57 ' -1.94 .056
NE - ET —2.93 2.68 -1.09 .278
BM - ET 2.07 2.21 0.94 .351

 

df - 67, for all contrasts

167

mmab Eocboob cot 8589: 80.596

lemon pco Iota yo mccsoo E 3902, com—2 ”m 059.1

s...

r

2m
mz

111

mm—

1omF

rmmr

Tome

1mm—

1OON

 

rmON

spunod u! iqbgem

168

postulating that the level of factor A would affect weight
loss, was supported.
Bypethe§i§_lhree

Hypothesis three stated that subjects receiving a
choice of treatment types would have significantly different
attendance ratings for the program than subjects who were
randomly assigned to treatment types. The possible range of
attendance was from 2 to 10 sessions, and the overall
program mean for attendance was 7.61 sessions (standard
deviation = 2.77). Attendance data were analyzed only for
subjects completing the program to control for differential
attrition. Table 20 summarizes the results of the analysis
of variance on attendance and the cell means and standard
deviations are presented in Table 21. Attendance patterns
for the 2 X 3 design are graphically presented in Figure 4.

The mean attendance for subjects choosing their

treatment type was 8.98 sessions (s.d. = 1.07) and the mean

 

attendance for subjects assigned to their treatment type was
8.97 (sgg; = 1.22). This difference was not significant
(E = 0.01 [1,67], 3 = .929).

Differences between levels of factor B were not
significant in the ANOVA z-test (p = .436). The attendance
totals were lower for the nutritional education groups (M =
8.67, §A_d_é = 1.23) than for the behavior management groups
(M = 9.13, sgg; = 1.13) or the exercise training groups (M =

8.96, s.d = 1.08). The factor A by factor 8 interaction was

169
TABLE 20

Analysis of Variance on Attendance

 

 

 

 

Source df M.S. F Sig. of F
Choice Factor (A) 1 0.01 0.01 .929
Treatment Factor (B) 2 1.08 0.84 .436

A x B Interaction 2 1.84 1.43 .247
Error 67 1.29

TABLE 21

Cell Means and Standard Deviations on Attendance and Adherence

 

Means (S.D.)

 

Condition N Attendance Adherence
Sages
NE 7 8.6 (0.79) 10.9 (1.77)
BM 19 9.3 (1.00) 13.7 (3.07)
ET 16 8.8 (1.18) 13.3 (3.30)
Assigned
NE 8 8.8 (1.58) 9.0 (4.34)
BM 13 8.9 (1.28) 8.5 (3.69)

ET 10 9.3 (0.82) 11.4 (3.31)

 

170

82200 .8 mocopcoto cows. ”c 0.59.1

09$ 30:500.;

hm 2m uz

P D hr

60350 H Olo
cocEmm< H mlm

 

 

pepuano 80018393 10 JaqwnN

171

not significant (Q = .814), despite the appearance created
by the scale employed in Figure 4.

In conclusion, hypothesis three, predicting that there
would be a significant difference between levels of factor A
on attendance, was not supported. There was no significant
difference between levels of factor A on program attendance
in the current research.

Hypothesis flog;

Hypothesis four stated that subjects who received a
choice of weight loss treatments would have significantly
different adherence scores at the end of the program than
subjects who were randomly assigned to treatment type.
Adherence was assessed for each subject at each program
session. Subjects were rated for self—reported adherence on
prescribed behaviors with a three point scale: zero points
for no adherence, one point for partial adherence, and two
points for full adherence. Adherence data were analyzed
only for subjects completing the program to control for
differential attrition. The possible range of scores was
from zero to 18, and the mean adherence score across all

participants was 11.59 (s.d. = 3.83). The cell means and

 

standard deviations for adherence ratings can be found in
Table 21. Analysis of variance results for adherence are
summarized in Table 22. Adherence patterns in the 2 X 3

design are graphically presented in Figure 5.

172
TABLE 22

Analysis of Variance on Adherence

 

 

 

Source df M.S. F Sig. of F
Choice Factor (A) 1 193.86 17.41 .000
Treatment Factor (B) 2 23.20 2.08 .133
A x B Interaction 2 22.96 2.06 .135

Error 67 11.14

 

173

:oEncoo .3 mocsoc oocmeonco com: ”m 6505

web EoEbooch
gm mz

‘L‘J

606:0 H 010
nocEmm< 1 Elm.

 

aouaieqpv

174

The mean adherence ratings for subjects receiving a
choice of treatments (M = 13.07, sgg = 3.10) was higher than
the mean adherence rating for subjects assigned to treatment
types (M = 9.58, sggg = 3.85). This difference was highly
significant (F = 17.41 [1,89], 3 = .000).

The E-test for the main effect of the treatment type
factor was not significant (F = 2.08 [2,67]. 2 = .133), nor
was the ANOVA interaction effect between factor A and factor
B (E = 2.06 [2,67], 9 = .135). Nutritional education (M =
9.9, ség; = 3.4) produced lower adherence ratings than did

the behavior management treatment (M = 11.6, s.d. = 4.2) or

 

the exercise training method (M = 12.6, sggL = 3.4), but the
differences were not significant.

For purposes of discussion in Chapter V, the
correlation between adherence and body weight change in
pounds was r = -.12 (p = .161). The correlation between
adherence and BMI change was r = -.20 (p = .045). The
correlation between adherence and percent overweight change
was r = -.16 (p = .087).

In conclusion, hypothesis four, which postulated that
subjects who received a choice of treatment types would have
significantly different adherence ratings than subjects who
were randomly assigned to treatment type, was supported.

The greater level of adherence scores obtained by the

subjects who received a choice of weight loss methods, as

175

compared to those subjects assigned to their treatment type,
was larger than what would be expected by chance.
Mypothesis Five

Hypothesis five stated that subjects who received a
choice of treatment type would have significantly different
scores on a measure of perceptions of their treatment's
efficacy than would subjects who were randomly assigned to
treatment type. Scores from the perceptions of treatment
efficacy measure were expressed as an average of the scale's
items, and had a possible range of zero (low) to four
(high). The perceptions of treatment efficacy measure was
administered at the pre-test, the mid-program assessment,
and the post-test. Table 23 summarizes the repeated
measures analysis of variance on perceptions of treatment
efficacy over time. The cell means and standard deviations
of the perceptions of treatment efficacy measure at all
three measurement times are summarized in Table 24. Figure
6 graphically presents the perceptions of treatment efficacy
trends for the two levels of factor A; similar presentation
of trends in perceptions of treatment efficacy for levels of
factor B can be found in Figure 7.

There was not a large difference between the choice
(M = 3.1, sggg = 0.45) versus assigned (M = 3.05, ség; a
0.53) subjects on their perceptions of treatment efficacy at
the pre-test. By the mid-program measure, however, the

randomly assigned subjects showed a larger increase in their

TABLE 23

Repeated Measures Analysis of Variance on Perceptions of Treatment

Efficacy Over Time

 

176

 

 

Source df M.S. F Sig. of F
Choice (A) 1 1.42 3.59 .062
Treatment (B) 2 1.32 3.35 .041

A x B Interaction 2 1.20 3.04 .055
Error 66 0.39

Time (C) 2 1.69 7.94 .001

A x C Interaction 2 0.38 1.79 .171

B x C Interaction 4 0.44 2.08 .087

A x B x C Interaction 4 0.21 0.99 .416
Error 132 0.21

 

TABLEifli

177

Cell Means and Standard Deviations of Pre, Mid and Post Measures

of Perceptions of Treatment Efficacy

 

Means (S.D.)

 

Condition Pre N - 95 Mid N - 80 Post N - 73
Choice
NE 3.0 (0.42) 2.9 (1.2 ) 3.4 (0.52)
BM 3.1 (0.43) 3.3 (0.46) 3.5 (0.46)
ET 3.2 (0.50) 3.2 (0.44) 2.9 (0.79)
Assigned
NE 2.9 (0.57) 3.1 (0.61) 3.3 (0.38)
BM 3.1 (0.51) 3.5 (0.37) 3.5 (0.42)
ET 3.2 (0.57) 3.6 (0.55) 3.6 (0.35)

 

178

300.33 pocEmmo m> 820:0 toe 0E: .66 zooofm
acoEucob co 22369.3 co moeoom cows. no 659...

ecoEmmommo co oEF

 

E9695 Eoemoem Eocooen.
“nod 22 oun—

r 080:0 H.010 . . wN
oocmfimd. u mlm 1m.N
105,

e. _..m

1N6

Ind

14%

and

 

Aooogna iuawioaii ,Lo suondazuad

179

womb €083.00: toe 0E: .55 xooorco
€95.00: eo mcozaoocoa co mocoom coo—2 K 0.59...

acoEmmommo .6 08:.

 

Eocoocn. E8005 E6581
“mod 2.2 8m
E H Elm
2m H I
mz n «In .
0
n u

(4

L‘

m.N
rm.N
106
1rd
rNAm
Inn
lin

1nd

 

40001119 109011091) 10 suogidawad

180
TABLE 25

Multiple Comparisons for Treatment Types on Perceptions of Treatment

Efficacy at Pre-, Mid-, and Post- Program Measures

 

 

 

Comparison Contrast S.E. T Sig. of T
Pre-Program (df I 92)
NE - ET -0.25 .13 -1.97 .052
BM - ET -0.10 .12 -0.83 411
Mid-Program (df - 77)
NE - BM -O.36 .18 -2.00 .048
NE - ET -0.32 .18 -1.75 .084
BM - ET 0.03 .15 0.22 .828
Post-Program (df I 69)
NE - BM -0.13 .18 -0.71 .478
BM - ET 0.27 .15 1.77 .080

 

181

perceptions of treatment efficacy than did the subjects who
chose their treatments (M = 3.5 and M = 3.2, respectively).
This difference remained stable to the post-test measure.

As can be seen in Table 23, the effect of the choice
condition level approached, but did not reach, significance
(2 = 3.59 [1,66], p = .062). The effect of time was clearly
significant (p = .001), with all treatment groups increasing
in their perceptions of their treatment's efficacy over time
except those subjects who chose exercise training (who
showed a decrease in their opinion of the treatment from
mid-program to the post-test).

The factor A by factor B interaction also approached,
but did not reach, significance (p - .055). The differences
between the three treatment types did, however, reach
significance (E = 3.35 [2,66], p = .041). Table 25 presents
a summary of the comparisons between treatment types at the
three measurement periods. While all three treatment types
showed an increase in perceptions of treatment efficacy from
the pre-test to the mid-program assessment, the nutritional
education treatment subjects had consistently lower
treatment efficacy scores than the other two treatments.

The only comparison reaching significance, however, was at
the mid-program assessment between nutritional education
(M = 3.03, sggg = .93) and behavior management (M = 3.39,
s. . = .44). Continued increases in perceptions of

treatment efficacy from mid-program to the post-test occured

182

for the behavior management treatments (mean increase =
0.07) and the nutritional education treatments (mean
increase = 0.30). The exercise treatment subjects, however,
showed a mean decrease of 0.17 from mid-program to the post-
test. In fact, the subjects who were assigned to exercise
training did show a slight increase in treatment efficacy
perceptions (mean increase = 0.07), but the subjects who
chose exercise training decreased their treatment efficacy
perceptions from mid- to post-measure (mean decrease =
0.30).

For purposes of discussion in Chapter V, several
correlations with the perceptions of treatment efficacy
measure are presented. The perceptions of treatment
efficacy measure correlated with the self-efficacy measure
at pre-, mid-, and post-test as I = .069 (p a .253), r =
.298 (p = .004), and I = .398 (p = .000), respectively. The
mid-program and post-test perceptions of treatment efficacy
scores correlated with similar program satisfaction measures
at r = .455 (p = .000) and r = .518 (p = .000),
respectively. Similar mid-program and post-test
correlations between perceptions of treatment efficacy and
program reactance were 3 = -.119 (p = .147) and r = -.503
(p = .000).

In conclusion, hypothesis five, which postulated that
subjects who received a choice of treatment type would have

significantly different scores on a measure of treatment

183

efficacy perceptions than subjects who were randomly
assigned to treatment types, was not be supported.
es' 8'

Hypothesis Six stated that subjects who receive a
choice of treatment type would have significantly different
scores on a measure of program satisfaction than subjects
who are randomly assigned to type of treatment.

Satisfaction with the weight loss program was assessed at
the mid-program measure and at the post-test. Scores from
the program satisfaction measure are expressed as an average
of the scale's items. The possible range is from zero (low)
to four (high). A summary of the repeated measures analysis
of variance on program satisfaction over time is presented
in Table 26. The cell means and standard deviations for
program satisfaction are presented in Table 27. Figures 8
and 9 graphically present the program satisfaction trends
for levels of factor A and factor B, respectively.

The effect of time in the analysis of program
satisfaction scores was not significant (3 a 0.08 [1,67],

p = .781), nor were the interactions between time and factor
A (p = .469), time and factor B (p = .549), or time by
factor A by factor B (p a .162). There was no pattern of
significant change in program satisfaction scores from the
mid-program to the post-test measure.

There was not a significant difference between subjects

receiving a choice of treatment types and subjects assigned

184

to treatment types on program satisfaction (E = 1.98 [1,67],
2 = .164). The subjects who chose their treatments did,
however, have lower program satisfaction scores at mid-
program and post-test than subjects assigned to treatments,
even though these differences were not significant.
Subjects choosing their treatment had a mean mid-program
program satisfaction score of 3.35 (sgg = 0.44) while
subjects assigned to treatments had a mean program
satisfaction score of 3.53 (ségg = 0.46) at the same
measurement time. At the post-test assessment, subjects who
chose their treatment had a mean program satisfaction score
of 3.36 (sgg; = 0.50) while assigned subjects had a mean
program satisfaction score of 3.53 (sggg = 0.42).
Differences between the treatment types on satisfaction
with the program were significant (F = 4.02 [2,67], 9 =
.022). Table 28 summarizes the comparisons between the
treatment types at the mid-program and post-test
measurements. As can be seen, the nutritional education
treatment subjects had significantly lower levels of program
satisfaction at the mid-program measure than the behavior
management treatment (2 = .046) or the exercise treatment
(9 = .029). The significance, but not the pattern of, these
differences disappeared by the post-test measure. The mid-
program and post-test program satisfaction means for
nutritional education (3.20 and 3.26, respectively) were

lower than the means for the behavior management (3.47 and

185
TABLE 26

Repeated Measures Analysis of Variance on Program Satisfaction Over Time

 

 

Source df M.S. F Sig. of F
Choice (A) 1 0.56 1.98 .164
Treatment (B) 2 1.14 4.02 .022

A x B Interaction 2 0.72 2.54 .087
Error 67 0.28

Time (C) 1 0.01 0.08 .781

A x C Interaction 1 0.06 0.53 .469

B x C Interaction 2 0.06 0.60 .549

A x B x C Interaction 2 0.19 1.87 .162
Error 67 0.10

 

186
TABLE 27

Cell Means and Standard Deviations for Mid and Post Measure of

Satisfaction and Reactance

 

 

Means (S.D.)

 

Satisfaction Reactance

Condition Mid N I 80 Post N I 73 Mid N I 80 Post N I 73
Choice

NE 3.3 (0.35) 3.3 (0.33) 0.53 (0.5) 0.57 (0.6)

BM 3.4 (0.47) 3.5 (0.48) 0.78 (0.7) 0.46 (0.5)

ET 3.4 (0.46) 3.3 (0.56) 1.00 (0.7) 0.92 (0.8)
Assigned

.NE 3.1 (0.60) 3.3 (0.40) 1.22 (1.2) 0.47 (0.4)

BM 3.6 (0.31) 3.5 (0.40) 0.38 (0.5) 0.40 (0.4)

ET 3.7 (0.27) 3.7 (0.36) 0.48 (0.7) 0.28 (0.4)

 

187

8.00.330 c0c9mmo .m> 00_o:0 toe
0E5 c0>o cozooemzom Eccmoea :00: ”m 0.591..

E0Emm0mmo .8 08:.

 

 

E9521 Eoeuoem
anon 22
886 u I... .
c0cm_mm< n... Bin.
0 é
D m.

 

 

001100131103 woJBOJd

188

.0003 E0an0b co.— 0E5

c0>o cozooemzom E0395 :00: ”m 0.59.:

E0Emm0mmo to 0E?

Eocooem Eocmoem

to“. 22

 

2m
m2

1]

\
VAN

o.m

lam

 

U01lOD}S[lDS LUDJfion

TABLE 28

189

Multiple Comparisons Between Treatment Types on Mid- and Post-

Measures of Progpgm Satisfaction

 

 

 

 

Comparison Contrast S.E. T Sig. of T
Satisfaction: Mid-Program (df I 77)

NE - BM -0.27 .13 -2.03 .046

NE - ET -0031 014 -2023 0029

BM - ET -0.04 .11 -0.34 .735
Satisfaction: Post-Program (df I 70)

NE - BM -0.24 .15 -1.64 .107

NE - ET -0.18 .15 -1.18 .241

BM - ET 0.06 .12 0.49 .628

 

190

3.50, respectively) and exercise training (3.50 and 3.44,
respectively) treatments.

The choice condition by treatment type condition
interaction approached, but did not reach, significance (p =
.087). Close examination of the cell means listed in Table
27 reveals the nature of this pattern. At the mid-program
measure, subjects assigned to nutritional education had
lower program satisfaction scores than did subjects who
chose nutritional education, while the pattern was reveresed
for behavior management and exercise training subjects. At
the post-test assessment, the scores for assigned subjects
and those receiving a choice were identical in the
nutritional education and behavior management treatments but
quite discrepant between the two exercise training programs.

In conclusion, hypothesis six, postulating that
subjects who received a choice of treatment types would have
significantly different program satisfaction scores than
subjects who were assigned to treatment types, was not
supported.

Mypgthesis Seveg

The seventh hypothesis under study stated that subjects
who receive a choice of weight loss treatment types would
have significantly different program reactance scores than
subjects who were randomly assigned to treatment types. The
program reactance scores are expressed as the average of the

scaleis items. The possible range of scores is from zero

191

(low) to four (high). The program scale was administered at
the mid-program assessment and the post-test. The cell
means and standard deviations for the program reactance
measure can be found in Table 27. A summary of the repeated
measures analysis of variance on program reactance over time
is presented in Table 29. Figures 10 and 11 graphically
present the program reactance trends for levels of factor A
and levels of factor B, respectively.

The main effect for time was at the significance level
(E = 4.02 [1,67], 3 = .049), suggesting that overall levels
of reactance to the program decreased from the mid-program
to the post-test measures. The interaction between time and
factor A was not significant (p = .477), nor were the
interactions between time and factor B (p = .908) or between
time, factor A, and factor B (p = .151).

Subjects who received a choice of treatments did not
differ significantly from subjects who were assigned to
treatment tpyes on measures of program reactance (E = 2.67
[1,67], 9 = .107). Mean levels of reactance were, however,
higher for the subjects receiving a choice of treatments at

mid- and post-measures (M = 0.82, s.d. = 0.68, and M = 0.65,

 

sgg; = 0.66, respectively) than for subjects who were
assigned to treatment type (M = 0.64, egg; = 0.86, and M =
0.39, s‘g; = 0.40, mid and post respectively).

The difference between treatment types on program

reactance was not significant in the ANOVA (E = 0.99 [2,67],

TABLE 29 192

Repeated Measures Analysis of Variance on Program Reactance Over Time

 

 

Source df M.S. F Sig. of F
Choice (A) 1 1.49 2.67 .107
Treatment (B) 2 0.55 0.99 .375
A x 8 Interaction 2 1.33 2.38 .100
Error 67 0.56
Time (C) 1 0.73 4.02 .049
A x C Interaction 1 0.09 0.51 .477
B x C Interaction 2 0.02 0.10 .908
A x B x C Interaction 2 0.35 1.95 .151

67 .0.18

Error

 

193

300.330 00.5600 .m> 00_o:0 toe

0E: :06 00:380.. Eoeooa :00: ”OF 050E

E0Emm0mmo co 0Ec.

Eocmoi
ummm

Eocmoi
0.5

 

00_o:0
00:903.

@110
BIG

0 _..o
Tomd
Tend
1070
100.0
r000
#050
T00.0
100.0
700.—

 

900010091 tquBOJd

009.0. 808000.: to”. 08$
:0>o 00.03000: 80:09: :00: A: 0.59.:

808000000 00 08:.

8808 80.69.
ammon— d 0.5 n.

 

194

070
10N.0
10nd
1.00.0
700.0
100.0
105.0
100.0
0.00.0
T00;

Em
mz

111

 

900010091 woJBOJd

195

p = .375). Mean reactance scores for the behavior
management treatment at mid-program (M = 0.64, sggg = 0.65)
and post-test (M = 0.44, sgg£ = 0.47) were lower than the
mid- and post-measures for nutritional education (M = 0.90,
sgg; = 0.98, and M = 0.52, ségg = 0.50, respectively) and
exercise training (M = 0.79, sggg = 0.75, and M = 0.67, sgg;
= 0.71).

The test for an interaction between levels of factor A
and levels of factor B was not significant (3 = 2.38 [2,67],
9 = .100).

Mid-program and post-test correlations between the
program reactance measure and the program satisfaction
measure were ; = -.243 (p = .015) and r = -.338 (p = .002),
respectively.

In conclusion, hypothesis seven, postulating that
subjects who received a choice of treatments would have
significantly different scores on a measure of program
reactance than subjects randomly assigned to treatment
types, was not supported. Randomly assigned subjects had
lower program reactance scores, on average, than did the
subjects who chose their treatments. The difference was not
greater than what would be expected by chance at the .05
level of significance.

Wis-1.3.1311:
The eighth hypothesis was that subjects who received a

choice of weight loss treatment type would have

196

significantly different scores on a measure of self-efficacy
than subjects who were randomly assigned to a treatment
type. The self-efficacy measure was administered to program
participants at the pre-test, at mid-program, and at the
post-test. Scores from the self-efficacy measure are
presented as subjects' average rating on the scale's items.
The possible range of scores is from zero (low) to ten
(high). A summary of the repeated measures analysis of
variance is presented in Table 30. Cell means and standard
deviations for the self-efficacy data at the pre-test, mid-
program, and post-test assessments are presented in Table
31. Figures 12 and 13 graphically present the self-
efficacy trends for levels of factor A and levels of factor
B, respectively.

Overall, the self-efficacy scores increased over time
for all conditions (E = 16.35 [2,134], 9 = .000).
Interactions between time of assessment and the other
factors were not significant.

The pre-test self—efficacy mean for the randomly
assigned subjects (M = 4.99, gégE = 1.75) was not much
different than the pre-test mean for the subjects who chose
their treatment type (M = 4.86, gégé = 1.54). The
difference between levels of factor A increased over time,
but not significantly (E = 0.65 [2,134], g = .522). The
randomly assigned subjects had higher mean mid—program and

post-test self-efficacy scores (M = 6.06, s.d I 1.90, and M

197
TABLE 30

Repeated Measures Analysis of Variance on Self-Efficacy Over Time

 

 

Source df M.S. F Sig. of F
Choice (A) 1 23.14 3.44 .068
Treatment (B) 2 10.60 1.58 .214

A x B Interaction 2 6.13 0.91 .407
Error 67 6.72

Time (C) 2 13.15 16.35 .000

A x C Interaction 2 0.53 0.65 .522

B x C Interaction 4 1.51 1.88 .118

A x B x C Interaction 4 1.04 1.30 .274

Error 134 0.80

 

TABLE 31 198

Cell Means and Standard Deviations of Pre-, Mid-, and Post-Measures.

of Self—Efficagy

 

 

Means (S.D.)
Condition Pre N I 95 Mid N I 80 Post N I 73
Choice
NE 4.1 (1.2) 4.7 (0.6) 4.9 (0.5)
BM 4.9 (1.5) 5.5 (1.5) 6.2 (1.4)
ET 5.2 (1.6) 5.2 (1.9) 5.6 (2.1)
Assigned
NE 4.3 (1.9) 5.5 (1.9) 6.3 (1.6)
BM 4.9 (1.5) 5.8 (1.4) 6.3 (1.4)

ET 6.0 (1.4) 6.8 (2.3) 6.7 (1.9)

 

199

9.00.330 00.29000 .0> 00_o:0 to:

08: :06 00.600 4600501000 :00: ”NP 0.59:

808000000 00 08.1w

 

80.59.: 8959.: 89.09.:
000.0 2.5. 0...: .
030:0 H @110 O .M
00:900.... .11.. Elm
10.4
100
3.0
TON

 

10.0

0.1003 Aoooma—Ilas

200

000.3 808000.: .00 08:

.._®>O mGLOOm

00050.1:00 :00: ”me 0.59:”.

808000000 00 08:.

 

 

89.00:: 89.09.: 89.09.:
000.0 2.2 on: .
PM ..u.. BID 0 m.
2m .1. 010
m2 1!. 9L0 .
-o .4
50.0
T00
10.5
1.0.0

0.1005 [booms—“es

201

= 6.4, §Lg; = 1.58, respectively) than did the subjects who
received a choice of treatment types (M = 5.25, gég; = 1.58,
and n = 5.77, gégé = 1.65, repectively). Overall, the main
effect for the choice versus assigned condition did not
reach significance (3 = 3.44 [1,67], 9 = .068).

The E-test for differences between levels of the
treatment type factor on self-efficacy was not significant
(E = 1.58 [2,67], p = .214). Differences between the three
treatment types at the self-efficacy pre-test were
significant and have been discussed elsewhere. At the pre-
test, the nutritional education treatment subjects had lower
self-efficacy scores than the exercise training subjects
(p = .001). The comparisons were not significant between
nutritional education and behavior management or between
behavior management and exercise training (9 = .087 for
both). The mean self-efficacy scores of the nutritional
education subjects increased at mid-program (u = 5.13,

s. . = 1.47) and again at post-test (M a 5.66, ng& = 1.35)
as did the scores for the behavior modification subjects

(M = 5.60, §;Q; = 1.46, and g = 6.25, §;Q.= 1.39, mid and
post respectively). The same pattern was true for the
exercise training subjects but with smaller increments from
mid-program (M = 5.84, g‘gL a 2.20) to post-test (M a 5.99,
§&g; = 2.05). The behavior management subjects overtook the
exercise training subjects on self-efficacy from mid-program

to post-test. None of the treatment type comparisons were

202

significant at mid-program or post-test. The interaction
between levels of factor A and levels of factor B was also
not significant (E = 0.91 [2,67],p = .407).

In conclusion, hypothesis eight, postulating that
subjects who received a choice of weight loss treatment
types would have significantly different scores on a measure
of self-efficacy than subjects assigned to treatment type,
was not supported.

flypothesis nine

The ninth hypothesis stated that there would be an
interaction between levels of factor A (choice versus
assigned) and subjects' scores on the Desirability of
Control Scale (Burger and Cooper, 1979). Hypothesis nine
was constructed to determine whether desire for control
would be a useful predictor of outcomes in a choice of
treatments paradigm.

The desirability of control measure was administered at
the pre-test. Scores are presented as total scale scores.
The possible range of desirability of control scores is 20
(low desire for control) to 140 (high desire for control).
The cell means were presented in Table 1, as was the summary
of the analysis of variance on desirability of control.

The mean desirability of control score for subjects
receiving a choice of treatment was 98.75 (ggg; = 14.34).
The mean desirability of control score for subjects assigned

to treatment type was 103.00 (s.d2 = 12.90). The difference

203

was not significant at the .05 level. Multivariate analyses
of variance were conducted to directly test the possible
interaction between desirability of control and levels of
factor A with the family of all outcome variables as the
aggregate dependent measure. Table 32 summaries the results
of the MANOVA with factor A, factor B, and desirability of
control as independent variables and the seven post-test
measures as dependent variables (attendance, adherence,
self-efficacy, perceptions of treatment efficacy, program
satisfaction, program reactance, and weight change).
Similar results are summarized for MANOVA tests on the four
mid-program measures in Table 33 (program satisfaction,
program reactance, perceptions of treatment efficacy, and
self-efficacy) and for the two pre-test measures in Table 34
(self-efficacy, perceptions of treatment efficacy). The
desirability of control scores were split at the median and
entered into the MANOVA as either high or low desirability
of control. The distribution of desirability of control
scores was roughly normal, and did not possess any obvious
modal characteristics to guide an alternative categorical
division of the data.

The desirability of control by levels of factor A
interaction was not significant for the group of post-test

measures (EH = 0.08 [7,54], p = .999), for the mid-program

measures (EH = 0.20 [4,65], 9 .937), nor the pre-test

measures (EH = 0.19 [2,82], 2 .824).

204
TABLE 32

Multivariate Analysis of Variance for the Post-Program Measures

With Choice, Treatment Type, and Desirability of Control as Factors

 

 

Source df Hotellings F 513. of F
Choice Factor (A) 7,54 6.98 .000
Treatment Factor (8) 14,106 0.77 .696
Des. of Control (C)* 7.54 1.53 .179
Interactions

A x 8 14.106 1.06 .401

A x C 7.54 0.08 .999

B x C 14,106 0.54 .902

A x B x C 14,106 1.08 .384

 

* Entered using Median-Split

205
TABLE 33

Multivariate Analysis of Variance for the Mid-Program Measures

With Choice, Treatment Type, and Desirability of Control as Factors

 

 

Source df Hotellings F Sig. of F
Choice (A) 4.65 2.10 .091
Treatment (8) 8,128 0.83 .576
Des. of Control (C)* 4.65 1.13 .352
Interactions

A x 8 8,128 0.76 .635

A x C 4.65 0.20 .937

B x C 8.128 1.63 .123

A x B x C 8.128 0.33 .954

 

* Entered using Median-Split

206
TABLE 34

Multivariate Analysis of Variance for the Pre-Program Measures

With Choice. Treatment Type, and Desirability of Control as Factors

 

 

Source df Hotellings F Sig. of F
Choice (A) 2.82 0.48 .618
Treatment (8) 4,162 2.74 .030
Des. of Control (C)* 2.82 1.43 .246
Interactions

A x 8 4,162 0.08 .606

A x C 2.82 0.19 (.824

a x c 4.162 2.55 @041

A x B x C 4.162 0.97 .425

 

* Entered using Median-Split

207

The remaining E-tests in the MANOVA summaries in Tables
32, 33, and 34 parallel the results of the MANOVA summaries
in Tables 9, 10, and 11, or are otherwise irrelevant to
testing hypothesis nine, and will not be described further.
Multiple regression analyses were conducted to determine if
the desirability of control scores would yield more
predictive power if used as a continuous, rather than
discrete, variable. A simple regression model was
constructed to use desirability of control, levels of factor
A, and an interaction term as predictor variables. The
results were entirely redundant of the MANOVAs and, hence,
are not presented here.

In conclusion, hypothesis nine, postulating an
interaction between desirability of control and having a
choice of treatments or being assigned to treatment, was not
supported. There is no evidence to support the hypothesis
that desirability of control interacts with choice of
treatments to affect outcomes.

0 s' e

The tenth hypothesis stated that subjects who chose
exercise training as a treatment would be significantly more
internal on the Weight Locus of Control Scale (WLOC) than
would subjects who chose behavior management or nutritional
education. The WLOC was administered at the pre-test. WLOC
scores are presented in terms of the total score: the

possible range of scores is from 4 (internal) to 24

208
TABLE 35

Comparison of Subjects ChoosingExercise Vs. Subjects Choosing_8ehavior

Modification or Nutrition Education on Weight Locus of Control

 

 

 

Comparison Contrast S.E. T Sig. of T
ET - 8 BM - 3 NE -3.42 1.97 -1.73 .087

df - 89

TABLE 36

Correlations of Weight Locus of Control With Attendance and Weight

 

Change Measures

 

Weight Change
Attendance Lbs. BMI Z Overweight

 

WLOC .054 .122 .124 .128
p- .301 p- .153 p- .148 p- .141

 

209

(external). Cell means were presented in Table 1, but will
be discussed again here. Overall, the total sample was
highly internal in weight locus of control (mean = 7.31,
gggg = 3.34).

Subjects who chose exercise training had smaller mean
WLOC scores (M = 6.63, gggg = 2.75) than the subjects who
chose behavior management (M = 7.59, gggg = 3.05) or
nutritional education (H = 9.09, ggg; a 4.66). Table 35
summarizes the comparison between the subjects choosing
exercise training and an average of the subjects choosing
nutritional education and behavior management on weight
locus of control. A directional t-test was employed. While
the trend was for subjects choosing exercise to be more
internal in locus of control than were the subjects choosing
other treatments, the difference was not significant at the
.05 level (I = -1.73, g; = 89, p = .087).

In conclusion, hypothesis ten, postulating that
subjects choosing exercise training would be significantly
more internal on the Weight Locus of Control Scale than
subjects choosing behavior management or nutritional
education, was not supported. The difference was in the
predicted direction but did not reach significance.

o ev

Hypothesis eleven stated that internality on the weight

locus of control measure would be positively related to

measures of attendance and body weight change. Table 36

210

presents the correlations between weight locus of control
and all three body weight change measures as well as with
attendance. Attendance figures for all subjects were used
in this analysis (ie, drop-outs were not excluded). It
should be noted that body weight change indices were entered
as negative numbers, and thus positive correlations with the
weight change measures suggest negative relationships. In
other words, externality in locus of control was more
strongly related to weight loss than internality.

The correlations were small and not more probable than
chance at the .05 level of significance. Thus, hypothesis
eleven, predicting a relationship between weight locus of
control and measures of attendance and weight change, was
not supported.

WEE

Hypothesis twelve stated that there would be
significant differences between levels of factor B
(treatment types) on body weight change. The relationship
between treatment types on the three weight change measures
was discussed thoroughly in the presentation of results for
hypothesis two. Cell means and standard deviations for the
three weight change measures were presented in Table 12,
ANOVA summaries for the three weight change measures were
presented in Tables 13 - 15, and comparisons between the

three treatment types on each weight change measure were

211

presented in Table 19. The results are briefly summarized
here as a formality.

The only treatment type comparison that approached
significance was between nutritional education and behavior
management. The nutritional education groups had a smaller
degree of weight change than behavior management for pounds
change (9 = .058), BMI change (2 = .045), and percent
overweight change (9 = .056). Comparisons between
nutritional education and exercise training, and between
behavior management and exercise training, were not
significant for any of the three weight change measures.

In conclusion, the effects of the three levels of
factor B on the three body weight change measures were not
found to be significantly different on the analyses of
variance. However, post-hoc mutliple comparisons between
treatment types revealed a difference between nutritional
education and behavior management that was at the
significance level for BMI change and just above the
significance level for pounds change and percent overweight
change. On the basis of the more powerful ANOVA test,
hypothesis twelve was not supported. It is noted, however,
that the nutritional education treatments tended to have

less desirable results on all the outcomes.

212

W

Hypothesis thirteen stated that subjects' histories of
past attempts at losing weight would be related to scores on
program outcome measures. Table 37 presents the
correlations between both frequency of past weight loss
attempts and the variety of previous weight loss attempts
(ie, different types of methods tried) with the primary
outcome measures.

Four variables proved to have significant correlations
with the two indices of past weight loss efforts. The
frequency of past weight loss attempts was negatively
related to adherence (I = -.27, p < .01). The initial self-
efficacy measure was negatively related to both the
frequency (3 = -.24, p < .05) and the variety (I = -.23,

p < .05) of past weight loss attempts. In addition, variety
of past weight loss attempts was negatively related to the
mid-program self-efficacy measure (; = -.21, p < .05).
Finally, the frequency of past weight loss attempts was
postively related to the mid-program measure of program
reactance (r = .21, p < .05). No other outcome measures
were found to be significantly related to either of the
indices of previous weight loss attempts.

In conclusion, there was partial support for hypothesis
thirteen, which predicted a significant relationship between

subjects' histories of weight loss efforts and each of the

TABLE 37

Correlations of Frequency and Variety of Previous Weight Loss Attempts

With Outcome Measures

213

 

Previous Attempts

 

 

Outcome Variable Frequency Variety
Lbs. Change .03 -.01
8M1 Change .04 .02
Z Overweight Change .02 -.02
Attendance -.16 -.O4
Adherence -.27** -.16
Self-Efficacy

Pre -,24* -.23*

Mid -.14 -.21*

Post -.15 -.07
Perceptions of
Treatment Efficacy

Pre .00 .04

Mid .04 -.03

Post .11 .12
Program Satisfaction

Mid -.06 .09

Post .04 .00
Program Reactance

Mid 21* - 12

Post .12 - 11

* p<(.05

** p< .01

214

outcome measurements. In fact, weight loss history was
associated with four outcome measures: adherence, initial
self-efficacy, mid-program self-efficacy, and mid-program
reactance. The relationships between history of past weight
loss attempts and other outcome measurements were not

significant at the .05 level of significance.

CHAPTER V

DISCUSSION

The purpose of the present research was to examine the
role of offering a choice of treatment types to overweight
participants of a weight reduction program. Half the
subjects were offered a choice between three treatment
types, the other half of the subjects were randomly assigned
to type of treatment. Primary research hypotheses were
constructed to assess the effect of treatment choice on
relevant treatment outcome variables: attrition, weight
loss, attendance, adherence, perceptions of treatment
efficacy, program satisfaction, program reactance, and self-
efficacy attributions. Comparisons between the three
treatment types (nutritional education, behavior management,
and exercise training) were also examined. Secondary
exploratory hypotheses were constructed to evaluate the
potential predictive value of two constructs of personal
control (desirability of control and weight locus of
control) in weight loss research. A final exploratory
hypothesis examined the relationship between prior history

of weight loss efforts and the primary outcomes.

215

216

This chapter discusses the results of the data analyses
for the thirteen experimental hypotheses, and considers
their meaning for the central thesis of the research.
Following a review of the hypotheses and their results, the
discussion turns towards the limitations of the research
study, and suggestions for further research.

The first hypothesis was related to the role of
treatment choice in reducing program attrition. Attrition
data were analyzed separately for subjects dropping out at
the first program meeting, subjects dropping out during the
course of the program, and total program-related attrition.

In the present study, overall attrition rates were much
higher than anticipated. 38.1% of the initial 118 subjects
dropped out of the program. This figure is consistent with
early weight loss research efforts (Brownell, 1982; Wilson,
1985). The monetary incentive scheme, however, was included
in the present study to hold attrition rates down. In a
review of weight loss research studies, Wilson (1985) found
that monetary incentives reduced attrition rates to an
average of 9.5% in weight loss programs. In Mavis' (1987)
dissertation study, the attendance-contingent incentive
group (upon which the present scheme was based) had an
attrition rate of 20%. These figures are much lower than
what was obtained in the present study.

In the present study, monetary incentives clearly did

not help to reduce attrition and, at least for eight

217

subjects, may have caused attrition. 6.8% of the original
group of subjects dropped out at the first program meeting
listing the monetary incentive scheme as their reason for
not continuing participation. The reasons for the apparent
failure of incentives to reduce attrition is not clear. The
lottery method for distributing incentive funds may have
seemed too random to the subjects, undermining the
attendance contingency. No test of this hypothesis is
available from the obtained data, however. Another possible
interpretation may be that the ratio between program costs
and incentive costs worked to devalue the program.
Specifically, the program cost each subject five dollars,
but a forty dollar deposit was required to enhance
attendance. Subjects may have become suspicious of why so
much more money was required to facilitate attendance if the
program was really worthwhile. Again, this hypothesis
cannot be tested with available data.

Wilsons' (1985) review of weight loss research revealed
that 57% of the subjects dropping out of weight loss
programs did so because of the slow rate of weight loss they
were experiencing. In the present study, only one subject
indicated dropping out because the program was not helping
her to lose weight. Twenty subjects (17%) dropped out for
reasons of schedule conflicts. This situation is similar to
Hurray's (1976) attempt to implement a choice paradigm in a

weight reduction program: 56% of the original subject pool

218

was lost due to schedule conflicts. Murray (1976) was
forced to confound choice of program time with choice of
treatment type to continue his study.

Attrition due to schedule conflicts raises an important
issue regarding offering subjects treatment choices: perhaps
the choice of day and time of treatment is more important to
people than a choice of treatment type. Of the 13 subjects
dropping out at the first meeting because of schedule
conflicts, however, only 2 were from the group receiving a
choice of treatments. Of the seven subjects who dropped out
later in the program for reasons of schedule problems, again
only two were from the group receiving a treatment choice.
Thus, treatment choices may work, after all, to mitigate
attrition due to schedule conflicts. Receiving a choice of
treatments may leave subjects with a greater level of
commitment to their chosen treatment than is the case when
subjects are assigned to treatment type. Available data do
not allow an examination of whether schedule conflicts, as a
reason for dropping out, was a real concern or a convenient
rationalization.

At any rate, significant differences in attrition rates
were found between the subjects assigned to treatment types
and subjects receiving a choice of treatment types. Thus,
while the overall program attrition level was high,
receiving a choice of weight loss treatments was associated

with lower attrition.

219

While behavior modification had the lowest attrition
rate of the three treatment types, the 27.3% attrition
associated with this treatment in the present study is much
larger than what was reported in the literature upon which
the treatment was developed. Wilson (1985) and Brownell
(1982) claim that behavior modification treatments reduce
weight loss program attrition to 10% - 15% on average. Even
with the monetary incentive scheme, behavior modification
treatments in the present study had higher levels of
attrition than what is claimed as the norm in the behavioral
research literature. The exercise program's 36.6% attrition
rate, next lowest in attrition among the three treatments
offered, was somewhere near the norms of 25% - 30% described
in the literature. The 54.5% attrition obtained for the
nutritional education groups is among the worst attrition
rates reported in the weight loss research literature.

While exercise and behavior treatments did not
significantly differ from each other, nutritional education
fared significantly worse on attrition than the other
treatments. Interpretive statements about differences
discovered between the three treatments on attrition, or on
any other outcome variable, cannot be generalized beyond
this study, however. Comparisons between the three
treatment types are compromised by the fact that each
treatment was delivered by a different group leader. It is

impossible to know whether any of the outcome differences

220

between the three treatments were due to the nature of the
treatments themselves or to differences between the program
leaders. It was the experimenter's subjective observation
that the nutritional education treatment groups were rather
boring, that the behavior management groups were rather
lively, and that exercise training subjects developed an
esprit de corps. Again, however, these differences could
have been due to the treatment type, the treatment leaders,
or some other (unknown) factor.

A comparison of those subjects who dropped out with
subjects who remained in the program as participants
revealed that drop outs had a lower initial self-efficacy
score. Interestingly, the nutritional education groups also
had a lower intitial self-efficacy score than the other
treatment groups. These findings are consistent with self-
efficacy theory, which holds that self-efficacy expectations
work to determine the amount of effort and persistence that
an individual will put forth in the face of obstacles
(Bandura, 1977, 1982). Low self-efficacy is logically
associated with decreased persistence, and the present data
support that relationship --- for the group of drop outs as
a whole, and for the nutritional education subjects.

It is important to note that drop outs were not
differentiated from participants on their approval of
treatment assignment. That is, when asked whether they

thought that their specific treatment type was the method

221

most likely to help them lose weight, participants and drop
outs responded with similar proportions of yes (roughly 64%)
and no (roughly 35%). This finding lends credence to the
idea that it was some factor other than treatment
differences, such as self-efficacy, that was associated with
attrition.

Since the attrition rate was higher among the assigned
subjects than the subjects receiving a choice, and since the
remaining assigned subjects ended up losing more weight than
the subjects receiving a choice (hypothesis two), attrition
may have created a selection bias during the course of the
study that favored the assigned subjects. For instance,
perhaps attrition tended to weed out those subjects who were
less committed to losing weight, but at differential rates
for the choice and assigned groups. Specifically, it may
have been that some of the subjects who received a choice of
treatments stayed in the program due more to a commitment to
their choice than to a commitment to lose weight.
Conversely, assigned subjects uncommitted to weight loss may
have had no special reason to be committed to their
treatment type, and thus dropped out at a higher rate,
leaving a group of remaining subjects more committed to
weight loss. If this were true, a smaller variance in
weight change measures would be expected for the assigned
subjects. The data reveal that this was not the case

however. In fact, the assigned subjects had slightly larger

222

variances on the weight change measures than did the
subjects receiving a choice of treatments.

It is still possible that the differential attrition
rates between assigned and choice subjects worked to create
a selection bias that mediated weight loss outcomes. Data
reveal that while the two groups were initially equivalent
in self-efficacy attributions, the assigned subjects had a
higher self-efficacy average at mid-program and post-test
measures than the subjects receiving a choice (hypothesis
eight). The same pattern was true for the perceptions of
treatment efficacy measures (hypothesis five). Thus, at
least for the attrition occuring during the course of the
program, the evidence suggests that the assigned groups were
left with subjects more confident in themselves and in their
treatments than were the groups where subjects had received
a choice. Data are not available to examine these patterns
for subjects who dropped out at the first program meeting.
Further, the data do not give any insight into potential
competing hypotheses that some factor(s) other than choice
may have mediated attrition.

The second hypothesis was related to the influence of
treatment choices on weight reduction. Three measures of
body weight were used to test this hypothesis: pounds, body
mass index, and percent over ideal weight. It was found
that the subjects assigned to type of treatment lost more

weight than the subjects choosing their treatments on all

223

three body weight measures. A discussion of differences
between the three treatment types on body weight change is
deferred to the discussion of hypothesis twelve.

The overall sample means on body mass index (32.4) and
percent overweight (44.4) suggest that the total sample did
represent an obese group of subjects. The norms specified
in the weight loss research literature for determining
obesity are a body mass index of 27.2 or greater for men and
29.2 or greater for women, and a percent overweight of 20% -
25% or over for men and 30% or over for women. By these
criteria, the total sample was still obese at the post-test
measure (mean BMI = 30.7: mean percent overweight = 36.8).

Literature reviews (Brownell, 1982; Blundell, 1984:
Wilson, 1985) indicate that the average weight loss in non-
medical programs is generally between 10 and 15 pounds over
an average 8 - 16 week period. Over ten week periods, as
was the case in the current research, a weight loss of 7.8
pounds to 8.25 pounds is expected (Wing and Jeffery, 1979).
The average weight loss in the current study, across all
experimental conditions, was 5.72 pounds. Thus, the overall
results in the current study do not compare favorably with
results reported in other studies. The comparison of total
program weight loss indices obtained in this study to
similar indices obtained in other studies may not be valid,
however. Most weight loss programs provide treatments that

include a combination of the three treatments that were

224

separated here in service of the experimental design.
Delivering nutritional, behavioral, and exercise treatments
separately may have compromised the effectiveness of each.

As Mavis (1987) pointed out, two factors should be
considered when viewing small weight change indices. First,
the use of a difference score to describe weight change
results in a measurement error larger than the individual
variances associated with the scores from which it is
derived. Thus, the significance of changes may be lost when
the magnitude of changes is small. Secondly, large within-
group variation also affects the analyses. A review of the
weight change indices obtained in the present study reveals
variances that are generally quite large relative to the
means. In an analysis of sample sizes and variances in
weight reduction studies, Wing and Jeffery (1984) suggest
that 45 subjects are needed in each experimental group to
find a five pound difference in weight change significant at
the .05 level. Groups of this size, however, become
unmanageable from group process and human resource
perspectives.

The 2.84 pound difference in mean weight loss between
the assigned subjects and the subjects receiving a treatment
choice was found to be significant in the ANOVA and multiple
regression analyses (despite Wing and Jeffery's findings).
The small differences favoring the assigned subjects were

also found to be significant for the BMI and percent

225

overweight measures on similar analyses. A rationale for
why the assigned subjects lost more weight (in absolute and
relative terms) than the subjects receiving a choice of
treatments is difficult to pin down.

In the discussion of attrition patterns, it was
hypothesized that a selection bias occured after subjects
dropped out of their respective groups such that the
remaining assigned subjects were perhaps more committed to
actually losing weight than the remaining subjects who
received a choice of treatments. The anticipated smaller
variance for assigned subjects on weight changes measures
was not found, however. Thus, if selection bias did occur
as a result of attrition, it must have been due to some
other intervening variable. Patterns of self-efficacy and
perceptions of treatment efficacy scores are consistent with
a hypothesis that attrition resulted in a selection bias
favoring the assigned groups, but competing hypotheses
cannot be ruled out on the basis of available data.

A possible alternative hypothesis is that the nature of
the program, and its advertisement, attracted subjects who
were likely to respond more favorably to being assigned a
treatment type than to receiving a choice. That is,
subjects who respond to a newspaper advertisement for a
weight loss program may represent a population of overweight
people who prefer that an "expert" tell them what to do.

The means of recruiting subjects may have pre-biased the

226

sample such that a choice of treatments may have been
inconsistent with the expectations subjects brought with
them to the program. In fact, at the pre-program meeting
where treatments and choices were described to the subjects
who received a choice, several subjects asked the
experimenter to make the decision for them (which, ofcourse,
was not done). It is possible, then, that offering subjects
a choice of treatments caused some cognitive and/or
affective responses that worked to diminish the
effectiveness of the treatments for subjects who received a
choice relative to subjects who were assigned a treatment
type. A comparison of the current sample with a sample of
the population of overweight individuals who do not respond
to advertisements for group weight loss programs would be
helpful here, but is not possible. Virtually nothing is
known about this comparison group. The discussion of a
potential cognitive/affective effect of choice is taken up
again following a discussion of the remaining hypotheses.
The third hypothesis was constructed to examine the
influence of offering subjects a choice of treatment types
on program attendance. The difference on attendance between
the assigned subjects and the subjects choosing treaments
was extremely small (0.01 sessions). Differences between
treatment types were also small: the greatest difference was
between nutritional education and behavior management, and

was only 0.46 sessions. Apparently the value of receiving a

227

choice of treatments was not powerful enough to influence
attendance. The possibility that potential differences were
prevented from occuring due to the attendance-contingent
monetary incentives is a plausable explanation for the
failure of the choice condition to generate differential
attendance scores. 1

The fourth hypothesis examined the influence of choice
of treatments on adherence. Analysis of the obtained
results suggested that having a choice of treatment types
had a powerful effect on adherence to prescribed behaviors
relative to the condition where subjects were assigned to
treatments. Subjects receiving a choice of treatments had
an average 72.6% adherence and subjects who were assigned to
treatments had an average 53.2% adherence. Unfortunately,
adherence was not strongly related to weight loss, as is
obvious in the results.

There are serious psychometric problems associated with
measuring adherence --- in general, and in this study
specifically. The adherence measure in the current study
was based upon the work of Stalonas et al. (1978) and Perri
et al. (1986). Each subject was asked at each program
meeting whether he/she had adhered to the previous week's
assignment; a score of 2, 1, or 0, was given based upon the
experimenter's subjective evaluation of full, partial, or no
adherence. The measure is weak relative to other measures

employed in this research study and contains no controls for

228

bias (intentional or unintentional) in the experimenter. No
evidence can be offered to rule out the competing hypothesis
that the results were biased by the experimenter, even
though the author believes that this was not the case.

Nonetheless, there is no explanation for why adherence
was unrelated to the weight change, other than to assume
that the adherence measure was in some way invalid. The
alternative explanation that adherence to weight control
behaviors is unrelated to weight control does not seem
acceptable. While very little research has actually been
done on adherence in weight control programs, the control of
body weight is clearly an adherence issue. The
physiological and behavioral research on body weight control
presented in Chapter I establish that body weight fluctuates
with changes in caloric intake (diet) and caloric output
(physical activity). Eating and physical activity are
behaviors, and adherence to these behaviors (in whatever
proportions) regulates body weight.

Thus, the results pertaining to adherence in the
current study, while consistent with hypothesis four, should
at best be taken cautiously.

The subjects in the nutritional education groups tended
to fare worse than the other two treatment types: they had
lower adherence ratings when compared to subjects in each of
the other treatments (though the differences were not

significant). The homework required of the nutritional

229

education subjects was the keeping of detailed food diaries.
The prescribed behavior did not vary much from week to week
as was the case in the behavior management groups where
there was a great deal of variety in homework assignments.
Thus, keeping the previously mentioned limitations of the
adherence measure in mind, nutritional education subjects
may have been less interested in the prescribed behaviors
than the behavior management subjects. While the exercise
training subjects also had the same homework assignments
from week to week (eg, walking), their failure to adhere to
prescriptions would be "sorely" felt at the next training
session, motivating them to step-up their between-session
adherence when it faltered. Without attempting to make any
broad generalizations, it is at least important to note that
the variety and salience of prescribed treatment behaviors
may have a large effect on a subject's willingness to
adhere.

The fifth hypothesis examined the influence of choice
of treatment types on perceptions of treatment efficacy.
The differences between the choice condition and the
assigned condition were not found to be significant at the
.05 level in the repeated measures analysis, though they
were close to significance (p = .062). It is important to
recall that the internal consistency measure of reliability
obtained with the current sample on the three item

perceptions of treatment efficacy scale was alpha = .61.

230

Thus, the scale did not possess impressive psychometric
strength.

Given the limitations of the assessment device, there
was no noteworthy difference in treatment efficacy
perceptions between assigned subjects and subjects receiving
a choice at the outset of the program, but a difference did
appear at the mid-program and post-test measures. While
both assignment conditions increased in treatment efficacy
perceptions over time, the assigned subjects showed a more
dramatic increase from pre-test to mid-program than did the
subjects who chose treatments. This finding may be due to
an attrition-related selection bias, as discussed
previously, but the mechanism is unclear. The pattern of
results for perceptions of treatment efficacy between
assigned subjects and subjects receiving a choice is nearly
identical to the pattern of changes on the self-efficacy
measure (hypothesis eight). While the two measures were
unrelated at the pre-test, they were significantly related
at mid-program (r = .298) and post-test (r - .398). Thus,
differences between assigned subjects and choice subjects on
perceptions of treatment efficacy may be due more to the
mediating effect of self-efficacy attributions than to a
direct effect of the manner of treatment assignment (though
the causal direction cannot actually be established with the

available data).

231

Mid-program and post-test perceptions of treatment
efficacy are correlated with program satisfaction measures
(3 = .455 and r = .518, respectively). Further, the post-
test treatment efficacy perceptions measure is correlated
with the program reactance post-test measure (r = -.503).
Additionally, satisfaction and reactance measures correlate
significantly with self-efficacy. That treatment efficacy
perceptions, program satisfaction, program reactance, and
self-efficacy are all related seems intuitively correct.
The direction of causal influences among these variables
seems open to speculation at the present time.

Differences between the three treatment types were
found to be significant in the ANOVA, but the pattern of
differences is confusing due to a nearly significant
interaction between the assignment method and the
treatments, and the significant effect of time. Initially,
the exercise training subjects had the highest perceptions
of treatment efficacy, followed by behavior management
subjects and nutritional education subjects. All treatment
groups showed an increase in treatment efficacy perceptions
over time except the exercise training choice group. This
group showed a decrease in treatment efficacy perceptions
over time.

The subjects who chose nutritional education had higher
initial treatment efficacy perceptions than their assigned

counterparts, but this situation reversed itself by the mid-

232

program measure, and reversed itself again by the post-test
measure. A similar, but less dramatic, pattern was seen for
the two behavior management groups.

After examining the relative scores of all six
experimental conditions (rather than simple treatment
comparisons) across measures with a known relationship to
perceptions of treatment efficacy (eg, self-efficacy,
program satisfaction, and program reactance) the author
concludes that there is no readily apparent pattern of
relationships that can account for the complex pattern of
scores among the six experimental conditions on perceptions
of treatment efficacy.

Hypotheses six and seven were concerned with the
influence of offering a choice of treatments on program
satisfaction and program reactance. Neither hypothesis was
supported. Satisfaction was not seen to increase over time,
while reactance decreased significantly over time.

While the meaning of program satisfaction is easy
enough to understand, reactance is a somewhat more difficult
concept. Reactance is not simply the converse of
satisfaction, but is rather an opposition or resistance,
aroused when individuals perceive a loss of freedom (Brehm
and Brehm, 1981; Perlmuter and Monty, 1979). Measures of
the two constructs should be negatively correlated, but not
perfectly so. An individual might, for instance, be

relatively unsatisfied with a treatment but not necessarily

233

have his/her reactance aroused. Correlations between mid-
program and post-test measures of satisfaction and reactance
were r = -.243 (p = .015) and r = -.338 (p = .002),
respectively. The program satisfaction measure had an
internal consistency reliability estimate of alpha = .83,
making it a reasonably solid measure. The program reactance
measure, however, had an internal consistency reliability
estimate of alpha = .57, making it somewhat less than
dependable. Thus, while there are differences in the
reliability of the two measures, they do seem to measure
distinct constructs (at least in a crude way).

The pattern of obtained results on program satisfaction
and program reactance is consistent with reactance theory
(Brehm and Brehm, 1981: Mann and Janis, 1982). The data
suggest, however, that it was the subjects who received a
choice of treatment types that felt a loss of freedom.

Brehm and Brehm (1981) argue that a perceived loss of
freedom increases reactance and may lead to a reduction in
satisfaction: precisely the result obtained in the present
study for the subjects who received a treatment choice
relative to assigned subjects.

While reactance theory predicts that this situation
should be maximally aroused when freedom is eliminated
altogether (ie, for the subjects assigned to treatments),
Brehm and Brehm (1981) acknowledge that other events might

alter this. They suggest that reactance may be higher for

234

subjects who choose treatments once they begin to experience
the consequences of their choice. In concert with this
possibility, it has been suggested that cognitive mechanisms
may interact with choice consequences (Brehm and Brehm,
1981; Mann and Janis, 1982). Thus, it may be possible that
subjects who received a choice of treatments felt a loss of
freedom which aroused their reactance and led to a decrease
in satisfaction. The choice subjects may even have
increased their perceptions of the value of the unchosen
treatments. Conversely, the assigned subjects may have
engaged in some "cognitive bolstering" of their treatments.
This explanation is also consistent with results obtained on
perceptions of treatment efficacy.

Brehm and Brehm (1981) also suggest that reactance in
choice situations may be mediated by differences between
individuals in their relative desire for control over
situations. Unfortunately, the Desirability of Control
Scale (Bruger and Cooper, 1979) used in this study did not
help to support this notion (hypothesis nine).

Nevertheless, reactance theory does adequately account for
the pattern of results obtained in the current study
regarding program satisfaction and program reactance.

Differences between the three treatment types on
program reactance were not significant, though subjects in
the behavior management groups tended to have less reactance

than subjects in the other groups. This is likely due to

235

the greater degree of variety and flexibility in the
behavior management treatment, which allowed for choices
within the treatment.

Differences between the three treatment types on
program satisfaction were significant, with subjects in the
nutritional education groups having lower levels of .
satisfaction than subjects in the other treatments. This
finding parallels other findings regarding the nutritional
education treatment. It should be noted that the lower
level of program satisfaction among nutritional education
subjects was associated with higher levels of reactance at
the mid-program measure, but not at the post-test measure.
Evidently the subjects who stayed in the nutritional
education groups to the end of the program had a more
favorable response to their treatment, even though they did
not necessarily lose weight. This is consistent with the
subjective reports of the nutritional education subjects as
related to the program leader.

Hypothesis eight was constructed to examine the
influence of treatment choice on self-efficacy attributions.
While self-efficacy scores increased significantly over time
across all conditions, the difference between assigned
subjects and subjects receiving treatment choices was not
found to be significant a the .05 level in the repeated
measures analysis (p = .068). The direction of differences,

however, was opposite from the predicted direction.

236

Since the self-efficacy scores increased over time
across all conditions, it is likely that the program worked
in some way to facilitate this increase. The question of
interest, however, regards why the assigned subjects tended
to show a greater increase in self-efficacy over time than
the subjects who received a choice of treaments. It has
been noted previously that the pattern of self-efficacy
scores between assigned subjects and choice subjects
paralleled other obtained results. Specifically, at mid-
program and post-test the assigned subjects had higher self-
efficacy, greater program satisfaction, less program
reactance, greater perceptions of treatment efficacy, and
more weight loss.

Marlatt and Gordon (1985) and Brehm and Brehm (1981)
have hinted at the possibility that reactance effects may
entail a two-stage process: reactance may be followed by a
sense of helplessness that serves to diminish outcomes.
Further, Bandura's (1977) conceptualization of self-efficacy
attributions as self-referent thought is consistent with the
idea that increases in feelings of helplessness might be
associated with lower self-efficacy. Thus, if receiving a
choice of treatment types was associated with a perceived
loss or restriction of freedom, which aroused reactance and
feelings of helplessness, then it is not surprising that
there were lower levels of self-efficacy among the choice

subjects relative to the assigned subjects. It is also not

237

surprising that these same subjects would not have as
favorable a view of treatment efficacy or program
satisfaction, and that they would not lose as much weight.

The "restriction of freedom" hypothesis, as an
explanation for the self-efficacy results and the overall
pattern of results, has three serious limitations. First,
the program measurement instruments were not all equally
reliable. Second, many of the obtained differences which
support the argument were not large enough to reach
statistical significance. Third, there is no empirical
basis for determining the causal order of relationships
among the relevant variables with the available data.

Generally, differences between the three treatment
types on self-efficacy attributions were not significant.
At the pre-test, subjects in the nutritional education
groups were significantly lower than subjects in the other
treatments. This difference became non-significant over
time, though the nutritional education treatment remained
associated with the lowest self-efficacy scores throughout
the program. Again, this is consistent with other results
obtained for the nutritional education treatment.

While it makes some sense that the subjects who chose
exercise training had higher initial self-efficacy scores
than subjects choosing other treatments (ie, exercise might
have easily been perceived as requiring more persistence and

effort), it is curious that the assigned exercise subjects

238

had higher initial self-efficacy scores than all of the
other treatment conditions. It may well be that the
assigned exercise subjects raised their self-efficacy
attributions in anticipation of the perceived requirements
of their treatment. Unlike their choice counterparts, the
assigned exercisers had no advanced warning of their
treatment type. Unfortunately, measures of anticipated
expectations were not gathered in the current study.
Nevertheless, across all exercise training subjects, there
was a higher level of initial self-efficacy. The rate of
increase in self-efficacy attributions over time, however,
was lower for the exercise training subjects relative to
subjects in other treatments. The self-efficacy
attributions of exercisers leveled off at mid-program and
did not increase much past that point. This parallels
subjective reports among the exercisers that they reached a
plateau in their physical conditioning by mid-program,
finding it somewhat more difficult to push themselves to
higher levels of exercise duration.

The remaining hypotheses were exploratory in nature
and, hence, are secondary to the primary hypotheses one
through eight. Hypothesis nine was offered to explore the
possibility of an interaction between receiving a choice of
treatment types and the relative desirability of control.
Relevant literature suggested that there was a strong

likelihood that a high desire for control could enhance the

239

effects of choice on treatment outcomes while a low desire
for control could reverse the effects of choice on enhancing
outcomes (Krantz and Deckel, 1983; Brehm and Brehm, 1981).
It was hoped that statistical control of this potential
confounding variable would enhance primary analyses.
Unfortunately, interactions between the choice factor
and the desirability of control measure were not found to be
significant. A significant finding for the desirability of
control variable would have enlivened the discussion of
reactance effects. Since the Desirability of Control Scale
(Burger and Cooper, 1979) has acceptable psychometric
properties, it can only be concluded that either (1) the
desire for control simply does not interact with treatment
choice in any meaningful way, or (2) the Desirability of
Control Scale does not measure a form of desire for control
that is relevant to weight loss research outcomes. Based
upon available data, either alternative is tenable.
Hypotheses ten and eleven were offered to explore the
potential predictive value of another control variable:
weight locus of control. The Weight Locus of Control Scale
(Saltzer, 1982) was found to have an internal consistency
estimate of reliability slightly better than that reported
in the literature on the scale's development (alpha = .69 as
compared with alpha = .58), but was still not a highly

reliable measure.

240

The overall mean locus of control score for the entire
sample (M = 7.31) suggests that the sample was highly
internal in locus of control. This finding is not
consistent with previous research findings which suggest
that participants in group weight loss programs are
generally more external in locus of control (Wallston and
Wallston, 1978; Weiss, 1977; Goldney and Cameron, 1981).
This finding is also not consistent with the previously
offered suggestion that the program advertisement attracted
a sample of subjects that was more likely to desire being
told by an "expert" what to do to lose weight (ie,
externality).

A reading of the four questions contained in the WLOC
scale, however, sheds some insight into the obtained degree
of internality in the overall sample. It is the author's
opinion that the questions are too obvious in their intent.
Answering the items in the external direction would imply a
clear lack of readiness to engage in weight reduction
efforts and this would be in conflict with subjects'
intentions to lose weight as expressed by their enrollment
in the program. The internality of the sample likely
reflects a weakness in the WLOC scale and a cognitive
attributional process among subjects beginning a weight loss
program.

Hypothesis ten examined the idea that of individuals

who were offered a choice of treatment types, those choosing

241

exercise would be more internal in locus of control than
those choosing behavior management or nutritional education.
This hypothesis was originally offered by Goldney and
Cameron (1981) in their review of the locus of control
construct in the weight loss literature. It was suggested
that exercise involves more self-direction than other
treatments and that individuals external in locus of control
would be less willing to choose a treatment strategy that
required self-direction. The hypothesis had never been
directly tested.

The results of the present study indicate that subjects
who chose exercise training were indeed more internal in
their locus of control than subjects choosing other
treatments, but the obtained difference was not significant
(p a .087). However, since the analysis was based on a
small sample size and used a signficance level perhaps too
stringent for exploratory research, the results may be taken
to suggest that locus of control may yet possess some value
in predicting what types of treatment overweight individuals
will choose when given choices. Clearly, a better weight
loss relevant locus of control scale is in need of
development to pursue this possibility.

Hypothesis eleven was constructed to explore the
suggestion by Weiss (1977) that internal locus of control
was positively related to weight loss and program attendance

(which, in turn, is related to weight loss). Other authors

242

have also suggested this relationship (Weinberg et al.,
1984; Wallston and Wallston, 1978; Goldney and Cameron,
1981). The results obtained in the present study indicate
that the relationship between internal locus of control, as
measured by the WLOC scale, and outcomes of attendance and
weight change is small to non-existent. Correlations
between the weight change measures and locus of control were
non-significant. Thus, the locus of control construct may
have some limited promise in prediciting what types of
strategies overweight individuals will choose to reduce
their body weight, but has little promise in predicting
weight losss outcomes.

Hypothesis twelve pertained to differences between the
three treatment types on measures of weight change. In
general, the experiment did not concern differences between
the treatments, but rather the comparisons between assigned
subjects and subjects receiving a choice of treatments.
Hypothesis twelve (and any other discussion of differences
between the three treatments) was included as a formality.
Further, as has already been mentioned, obtained differences
between the three treatment types on outcome measures are
likely not generalizable to similar treatments outside this
study. In the present study, each treatment was delivered
by a different leader and this confounds the meaning of
comparisons. Each treatment was also offered in a "pure"

form (ie, not mixed with strategies from other methods), a

243

situation not entirely representative of the way these
treatments are offered in the commercial marketplace.

With these caveats in mind, it can be said that while
treatment differences on weight change were not found in the
omnibus tests for the three body weight change indices, the
multiple comparisons revealed that nutritional education was
the weaker of the three treatments in promoting weight loss.
The behavior management treatment was associated with
slightly more weight loss than the exercise treatment.

The overall weakness of the nutritional education
groups on most of the outcome measures in this study
suggests that, at least for the subjects and groups in the
present study, nutritional education was not equivalent to
the other treatments. This finding may be due to the nature
of the treatment, the nature of the program leader, or to
some a priori deficits in the nutritional education
subjects.

At the pre-test, the nutritional education subjects
tended to have lower self-efficacy, lower perceptions of
treatment efficacy, and were more external in locus of
control than other subjects. It is difficult to know
whether these findings are a cause or effect of the
treatment characteristics. It would seem that pre-test
measures should not have been affected by the treatment (as
the treatment had not occured yet), but this may not be

true. Subjects may have unsuccessfully tried the dieting

244

approach so many times in the past that they rated their
confidence (in themselves and in their treatment) as very
low.

It would seem more likely that the failure of the
nutritional education treatments was due to the nature of
the treatment or the program leader. Behavior modification
techniques, which can enhance both the face value and the
empirical effect of a dieting approach to weight loss, were
eliminated from the treatment so that it would appear
distinct from the behavior managment treatment. This may
have affected perceptions of the treatment. In addition, it
seems reasonable enough to assume that simply receiving
information about appropriate dietary plans for promoting
weight loss is not enough, by itself, to actually stimulate
a person to lose weight. In fact, it was the general
failure of nutritional approaches to promote weight loss
that provoked the application of behavioral techniques to
the weight reduction field in the first place.

In summary, it seems that there is enough accumulated
evidence in the current study to conclude that the
nutritional education treatment was inferior to the other
treatments (though not always significantly). This does not
affect the validity of having offered the three treatments
as roughly equivalent at the beginning of the study: what is
most important to the task of assessing the role of choice

is not so much the content of the separate treatments, but

245

that they are perceived as being equivalent. Nevertheless,
it does suggest that any future attempts at replicating this
research should attend to the ethical responsibility of
offering a "better" version of the nutritional education
treatment.

Hypothesis thirteen was constructed to assess the
relationship between subjects' history of prior efforts at
losing weight and the various program outcomes. This
hypothesis was offered purely for exploratory reasons that
were unrelated to the central thesis of the study. The
rationale for the hypothesis was that there is an unresolved
debate in the weight loss literature regarding whether
larger numbers of prior weight loss attempts work to
increase or reduce a person's chances of success in their
next attempt.

The correlational data presented in Table 37 suggest
that increases in prior weight loss efforts are unrelated to
most of the outcome variables. Significant relationships
did exist between prior efforts at weight reduction and
self-efficacy, adherence, and reactance. The relationships
observed, however, were in an unfavorable direction. Any
effect prior history had on self-efficacy and reactance
disappeared by the post-test assessment.

If self-efficacy theory (Bandura, 1977) is valid, it
can be assumed that the negative relationship between

initial self-efficacy and prior history of weight loss

246

efforts means that the subjects' previous efforts were
unsuccessful. A history of failure should reduce self-
efficacy attributions for future efforts. Lowered self-
efficacy may have played an important role in determining
the findings regarding adherence and program reactance, as
has been discussed previously.

The primary outcome measure in weight loss research is,
and should be, weight change. The data obtained in the
present study suggest that there is no relationship at all
between prior efforts at losing weight and weight lost in a
subsequent effort. It is likely that there is some truth in
both arguments regarding the effect of prior weight loss
efforts, and that the relationships are sufficiently complex
and individual-specific that global analyses cannot resolve
the issue.

In consideration of the volume of preceding analysis
and discussion, a few summary statements may facilitate
final theoretical discussion. A choice of treatment types
facilitated a smaller attrition rate and greater adherence.
A choice of treatment types was, however, also associated
with less weight loss. There were no obtained differences
on attendance between subjects choosing treatments and
subjects assigned to treatments. Though not significant,
the direction of obtained differences also seem to point

toward a negative (contrary to hypothesis) effect of choice

247

on perceptions of treatment efficacy, self-efficacy, program
satisfaction, and program reactance.

Discussion to this point has implied that it is the
author's belief that offering a choice of treatments to the
subjects caused some form of cognitive/affective response(s)
that worked to diminish the effectiveness of the treatments
relative to subjects assigned to treatment type. Self-
efficacy theory (Bandura, 1977, 1982), reactance theory
(Brehm and Brehm, 1981), and decision theory (Mann and
Janis, 1982) have been enlisted to explain the obtained
results.

It has been suggested that offering subjects a choice
resulted in perceptions of restricted or lost freedom and
that this aroused reactance. Data, while not terribly
strong, support the contention. It was further suggested
that increases in reactance were accompanied by increases in
feelings of helplessness and, thus, decreases in self-
efficacy. Perceptions of treatment efficacy and program
satisfaction were said to have decreased in concert with
these other changes. Consequently, the subjects who
received a choice lost less weight than the assigned
subjects. Again, while not overwhelming, the data were more
or less in support of the "theory".

Helson's (1959, 1964) theory of disconfirmed
expectations may offer some additional conceptual support

for the hypothesis that offering a choice of treatments had

248

a cognitive/affective impact on the subjects that worked to
diminish treatment outcomes. Helson postulated that two
varieties of "expectation" must be considered in this line
of research: expectations as preference choices and
expectations as anticipations. According to the theory, if
an expected preference elicits an emotional response, then
affective and motivational qualities of disconfirmed
expectations are determined by the direction (positive vs.
negative) and intensity of the discrepency. Thus actual
events, seen as more desirable than the anticipated
expectation, would result in positive affect and an approach
motivation. Conversely, actual events seen as less
desirable than anticipated would result in negative affect
and avoidance motivation. The result of disconfirmed
anticipated expectations is dependent upon preference
choices. In other words, it is possible that assigned
subjects who preferred a specific weight loss method had
preconceived ideas about the other methods, but after
actually experiencing the treatment, felt a reduction in the
intensity of their preference choice as a function of
finding their assigned method to be more desirable than
anticipated, hence working to enhance outcomes. This is
highly similar to the "cognitive bolstering" operation
described by Mann and Janis (1982). Conversely, subjects
who chose their treatment type may have found that the

treatment was less desirable than anticipated, and had to

249

grapple with the fact that they had chosen the treatment.
Hence, affect and motivation may have taken on negative
valences for subjects choosing treatments, and outcomes may
have been undermined.

A direct test of Helson's (1959, 1964) theory is not
possible with the data available from the current study.
Preference choices among the assigned subjects are not known
and anticipatory expectations were not measured for any of
the subjects. Nevertheless, differences between anticipated
expectations, preference expectations, and actual events,
may have had cognitive/affective effects that played a role
in the outcomes obtained in the present study.

Therefore, to the extent that the available data "fit"
into these theories, the conclusion of this research study
is that a choice of treatment types caused a cognitive
and/or affective reaction in the subjects such that
treatment outcomes were diminished for subjects who received
a choice relative to subjects who were assigned to
treatments.

The generalizability of the results obtained in this
research is compromised by several factors pertaining to
both internal and external validity. The most obvious
limitation should be that significance tests between the
three treatments were compromised by a lack of random
assignment to treatments for subjects in the choice

condition. Thus, while it has already been noted that

250

differences between the treatment types are not
generalizable due to the confound of group leaders, there is
also a statistical reason for using extreme caution in
handling results pertaining to the three weight loss
methods. Additional statistical threats to the validity of
this study include the relatively low reliability associated
with several of the measures and the small sample size
(particularly by the end of the program).

Another major area of limitation to the generalization
of these research findings regards the possibility that the
subject recruitment procedures attracted a non-
representative sample. It could be that subjects who
respond to advertisements for weight loss programs are more
likely to expect or anticipate being assigned to a method of
weight reduction and are more likely to be reactant to
receiving a choice of methods than would be the case for
other overweight individuals. It has been noted several
times, however, that there are suspected, but unknown,
differences between individuals who attend organized weight
loss programs and individuals who attempt to lose weight on
their own. It has also been stated that the results of this
study are not intended to describe the state of affairs for
"natural" weight reducers. Thus, as long as the population
of overweight individuals who respond to newspaper
advertisements for group weight loss programs is the

referent, the results can be generalized to that population.

251

The final threat to the study, the differential
attrition (ie, mortality) between assigned subjects and
subjects receiving a choice, has been discussed at length
previously. It is important to note that the attrition
patterns were results, or findings, in the present study.
Further, differences due to attrition were statistically
controlled in all relevant analyses subsequent to hypothesis
one.

There are several directions for potentially fruitful
future research that can be suggested based on the results
of this study. The obvious next step to exploring the role
of choice in weight loss would be to replicate this study
while measuring all subjects' anticipated expectations and
preference choice expectations. This would directly test
the validity of Helson's (1959, 1964) theory as an
explanation for the effect of choice in weight reduction
programs. It might also be helpful to explore where
subjects' expectations came from and what their determinants
are. Along these lines, future research should assess
subjects' reasons for wanting to lose weight. This is
likely an individual difference variable that interacts with
the effect of choice.

Research might also be directed to expanding the choice
of treatments paradigm to other health-related behaviors,
and in new ways. The resulting pyramid of knowledge would

allow researchers to learn more about the role of choice,

252

control, and behavioral health outcomes and specify where
treatment choices are helpful and where they might best be
avoided.

It might also be argued that since the effect of choice
was not entirely positive in this study, efforts might
better be directed toward developing a "super program" that
would maximize weight loss for all (or most) participants,
rather than expending more effort on differential treatment
prescriptions. When a more powerful weight loss method has
been developed, attention might return to the role of choice
in enhancing outcomes. Researchers might study the role of
mini-choices or micro-choices within a single, effective
type of treatment.

Clearly, this study was limited by the validity and
reliability of the measures employed. Empirical scale-
development research would be a most helpful in this line of
research. Particular attention should be devoted toward
developing valid and reliable, yet manageable, measures of
adherence behavior.

The results of this study suggest that the role and
impact of monetary incentives in weight reduction need
continued evaluation. Research should pursue the question
of which behavior is the best candidate for incentive
contingencies. Tying together the development of a better
adherence measure with an exploration of adherence-

contingent incentives would be a major contribution to the

253

weight loss literature. In addition, the "right" amount of
monetary incentive is far from being established at the
present time, and the current research suggests that the
amount of incentive money may interact with the amount of
money required for program costs to alter the effect of the
incentive.

Finally, individuals who lose weight on their own,
without the aid of an organized program, should be studied
in their own right. Much can be learned from these
"natural" weight reducers. When a sufficient data base has
been gathered about this population, comparisons between
program-reducers and "natural" reducers could revolutionize

weight reduction research and available treatments.

APPENDI CES

APPENDIX A

Physical Activity Readiness Questionaire

254

flurry! Return this form today!

 

 

 

 

Your name: Your age:
Your phone number: (day) (eve)
Eh1sical_nealth_fluestinnaire

Please complete the following questionaire. We cannot
include you in the weight loss program without this
information. Note: filling out this questionaire does not
mean you will be exercising, it is simply a precautionary
measure.

Please mark those items that apply to you:

Your doctor said you have heart trouble, a heart
murmur, or you have had a heart attack.

You frequently have pains or pressure --- in the left
or midchest area, left neck, shoulder, or arm ---
during or right after you exercise.

You often feel faint or have spells of severe dizziness

You experience extreme breathlessness after mild
exertion.

Your doctor said your blood pressure was too high and
is not under control. Or you don't know whether or not
your blood pressure is normal.

Your doctor said you have bone or Joint problems such
as arthritis.

You are male and over 45 or a female over 50 and not
accustomed to exercise.

You have a family history of premature coronary artery
disease.

You have a medical condition not mentioned here which
might need special attention in an exercise program
(for example, insulin-dependent diabetes).

None of the above items apply to me.

YES, I want to be included in the weight loss program.
Please contact me with details about the first program
meeting.

APPENDIX B

Description of Treatments
and
Instructions to Subjects

255

DESCRIPTION OF WEIGHT LOSS METHODS

A variety of weight loss methods are currently in
professional use in the United States today. The methods
used most often in professional weight loss programs are:

1) nutritional management or dieting,
2) behavior management training, and
3) exercise training.

These three methods have clearly been demonstrated to
be effective approaches for promoting weight loss. They
each represent differing approaches to losing weight, but
are roughly equivalent in their effectiveness.

Many professional weight reduction programs offer all
three of these weight reduction methods in abbreviated form,
covering only the highlights of each method, and without
instructors qualified as experts in each method. The
present weight loss program will offer a complete course in
each method, taught by instructors qualified in each area.
This will allow each participant an opportunity to receive a
thorough knowledge of an effective weight loss method.

Before briefly reviewing the three major weight
reduction methods offered in this program, you should keep

in mind a few things:

256
1. Weight loss is a relatively long-term process.
You should view the present program as a means for
you to begin losing weight.
2. In each group, you will be offered some
education regarding each of the three methods but
you will focus on one of the methods in depth.
3. The present program will last 10 weeks.
Following this, a follow-up program will allow
you to learn more about other methods of weight
loss. You will be gauranteed a position in the
follow-up program, though you will be free to
discontinue at any time.

Now, let's take a look at the three major weight loss
methods that will be offered:

H ! °!i J E: !'

... will consist of a program designed to help you
create sound and sane dietary plans to lose weight, based
upon a method developed by the National Dairy Council. You
will learn separate diet plans for losing weight and for
keeping the weight off once you reach your goal.

- nutritional approaches to weight management lead
directly to weight loss through a reduction in unnecessary
calories.

- a nutritional education program will help you locate
where in your diet the unnecessary calories really are. You

will learn to reduce caloric intake without having to

257
deprive yourself excessively --- as is the case in many
dieting plans.

- a rational nutritional manangement plan works
directly to decrease the serious health risks associated
with being overweight (eg, heart disease, diabetes, stroke).

- a rational nutritional management plan leads to
improvements in cardiovascular functioning.

- research has shown that a nutritional approach to
weight loss produces an average minimum 1 - 2 pounds per
week weight loss, initially. The rate of weight loss slows
down over time.

- the nutritional approach to weight loss is the most
widely used weight reduction method, and has the most
research support for its effectiveness.

2 . I . .

... will consist of a program of moderate physical
exercise and increased activity, designed to help you lose
weight and increase your level of physical conditioning
without excessive strain, pain, or risk of injury. Adapted
from the most successful exercise-based weight loss research
studies, this program can teach you the fundamentals of
stretching, warm-up, and low-impact aerobic exercise.
Walking will be the primary exercise used. Taught by an
expert in exercise physiology, the goal of this program is
not to train you to be an athelete, but to help you lose

weight.

258

- exercise works directly to promote weight reduction
by increasing the expenditure of calories and by decreasing
the size of fat cells.

- moderate exercise increases your metabolism so that
you burn more calories throughout the day, even when you are
at rest. This can also help counteract the slowing of your
metabolism that occurs with dieting.

- moderate exercise can reduce the serious health risks
associated with being overweight. Moderate exercise will
increase your cardiovascular fitness, and reduce the risk of
heart disease, diabetes, and strokes.

- moderate exercise will work to suppress your
appetite, thus indirectly helping your weight loss efforts.

- a moderate program of simple walking produces an
average minimum weight loss of 2/3 - 1 pound per week
initially. The rate of weight loss increases in a few
weeks. (The slightly slower rate of weight loss at first is
due to the fact that you are building muscle tissue at the
same time.)

- regular exercise is the best predictor of long-term
success in maintaining weight losses.

W

... will consist of a program designed to teach you
meaningful self-control methods for reducing your weight.
Based upon the most successful behavior management

approaches to weight loss, this program will: help you

259
modify how you eat, teach you to develop realistic weight
loss goals and plans, assist you in developing meaningful
self-rewards, help you deal with other people in your life
regarding your weight loss efforts, teach you how to stop
self-defeating behaviors, help you develop new ways to
manage stress (which can lead to weight gain).

- the behavior management approach to weight loss
actually consists of many helpful self-control methods, and
allows you more freedom than other approaches for a choice
of actual methods you will use.

- a behavioral approach to weight loss is one of the
strongest factors associated with success in maintaining
weight losses over long periods of time.

- a behavioral approach to weight loss has been found
to enhance the chances of success for people who have
trouble staying with a weight loss program.

- the behavior management training is also applicable
to other problem areas as well: smoking, alcohol
consumption, exercise, any habit problems, or any personal
goals. The strategies can be used to advantage in many
areas of your life.

- the behavioral weight loss method has been found to
produce an average minimum weight loss of one to one-and-one
half pounds per week, at a fairly steady rate.

- behavior management approaches to weight loss can

enhance your chances of success with dieting, exercise, or

260

any other method of weight loss. Motivation and self-

control are the focus of this method.

In summary, three widely used effective weight loss
methods will be offered in this program:

1. moderate exercise training,

2. behavior management, and

3. nutritional management or dieting.

All three methods promote an average one pound per week
rate of weight loss. Initially, nutritional management
produces a slightly faster rate of weight loss while
exercise produces a slightly slower rate of weight loss.
These differences equalize rather quickly over time. All
three methods help to reduce the health risks associated
with being overweight. Each method has its own distinct
strengths and benefits.

The three methods are roughly equivalent in promoting
weight loss, but they represent different approaches to the
task of losing weight.

Benembg;.... the present program is designed to help
you initiate your weight loss process. We will provide you
with ten weeks of in-depth training in one of these methods,
along with some education regarding the other methods. You
will be guaranteed a place in the follow-up program where
you will have the opportunity to learn more about all of the

weight loss methods.

261

INSTRUCTIONS TO SUBJECTS

Following the description of treatments, subjects in
group A.1 were given instructions regarding the choice of
treatment type and subjects in group A.2 were given
instructions regarding their assignment to treatment type,
as follows:

Subjects in Group A.2 -- You have been assigned to one
of the weight loss methods offered in this study for the
first ten-week phase of this program to help you begin the
weight loss process.

Subjects in Group A.1 -- We would like you to choose
gng weight loss method for the first (10 week) phase of this
program. Please choose the method which you feel would be
best suited to your needs. Each person is likely to respond
differently to each of these methods --- and for different
reasons. You may have some insight into which of these
three weight loss methods is best for you:

- perhaps you have had past experience with one of
these methods and have either positive or negative feelings
about it.

- you may already know about one of these methods and
want to know more about one of the others.

- you may already know about one of these methods and
want to know more about that same method. Or you may want a
review of what you already know. i

- you might prefer one of these methods as a strategy

262
to begin losing weight and desire another later on for
longer term weight loss efforts.
- you may feel your chances of succeeding are better
with one of the methods over the others.
Choose the gag weight loss method that you think is best for

you.

263

RESULTS OF THE PILOT STUDY:

EQUIVALENCE OF WEIGHT LOSS TREATMENT DESCRIPTIONS

EIQQEQQIQE

Nineteen undergraduate students, enrolled in a
communications training course, were used as subjects in
this pilot study to determine perceptions of equivalence in
the descriptions of three weight loss methods. Subjects
were 18 females and one male, ranging in age from 19 to 24
years. With the exception of one student, the sample
consisted of a normal weight group of undergraduate
students.

The subjects were given a description of a weight loss
program which involved three separate training conditions:
nutritional management, behavior management, and exercise
training. Subjects were asked to "pretend" that they had
signed up for the program and were hearing the description
of treatments to be offered. In addition to an oral
presentation of the treatment descriptions, subjects were
given written descriptions of the three treatments. After
the presentation of the treatment descriptions, subjects
were given a questionaire to assess their perceptions of the
three treatments described. After the questionaire was
completed, subjects returned the questionaire to the

researcher and were given an explanation of the study.

264

The questionaire used in this pilot study consisted of
five items. Item #1 asked the subject whether he/she
perceived the treatments, as described, to be roughly equal
in effectiveness for promoting weight loss. Item #2 asked
to subjects to list the treatment that was perceived as
being "least" gag "most" effective, if treatments were not
perceived as equal. Item #3 asked the subjects to rate the
degree to which they perceived each treatment as being
logical. Item #4 asked the subject to rate the degree to
which he/she was confident that each treatment type would be
successful in promoting weight loss. Item #5 asked the
subject to state his/her preference for one of the weight
loss methods.

33:31::

item #1: scale a 1 (strongly disagree) to 5 (strongly agree)
mean e 4.55
m; =- 0.62
n S 18

item #2: each treatment received gag vote as being the most
effective. exercise training received one vote as
being the least effective method; there were no
other votes for least effective method.

item #3: scale a 1 (not at all logical) to 5 (extremely

logical)
n = 19
nutrition -- mean a 4.68, s.g. = 0.48
behavior -- mean = 4.53, s.d. = 0.61
exercise -- mean = 4.53, §.g. = 0.61

item #4: scale a 1 (not at all confident) to 5 (extremely

confident)
n a 19
nutrition -- mean a 4.53, s.d. = 0.70
behavior -- mean a 4.00, s.d. = 0.82
exercise -- mean = 4.58, s.g. = 0.61

265

item #5: n = 19
nutrition 8 5
behavior a 4
exercise a 8
not scorable a 2 (chose all three)

A oneway ANOVA was conducted on the responses to items
#3 and #4, as follows:

item #3: SST-17.9, SSE-0.29, sswai7.62, g;=2,54
E a 0.438, n.s.

item #4: SST=31.3, SSB=3.93, SSW=27.37, g;=2,54
E = 3.85, n.s.

It should be noted that the grand sample size was
considered to be 57 (ie, "responses"). In fact, there were
19 subjects answering questions regarding 3 treatments.
Thus, the responses were not really independent.

In conclusion, the description of weight loss
treatments appears to contain equivalence of treatment

descriptions.

APPENDIX C

Consent Form

10.

Signed: Date:

266

PARTICIPANT INFORMED CONSENT

I have freely consented to participate in this weight
management study being conducted by Thomas C. Fuller.

I understand that this study is being conducted under
the auspices of Michigan State University and the School
of Health Education, Counseling Psychology. and Human
Performance.

I understand that the purpose of this study is to explore
the relationship between various personal characteristics
and the weight loss strategies offered. Full disclosure
of the complete research design will be made available to
me at the last program meeting.

I understand that I must contribute $45.00 to the
program. $5.00 will be used to help cover program

costs and will not be redeemable. $40.00 will be placed
in the program raffle deposit fund and I will have a
chance to win money for my participation in the study.
All monies are to be awarded and distributed in the last
program meeting. I must attend the last meeting in order
to receive money due to me (or make prior arrangements
with the program leader).

I understand that I am free to discontinue my partic-
ipation in the program at any time. However, if I
decide not to continue with the program, I understand
that none of my $45.00 deposit is refundable.

I agree to complete questionaires to be administered
during the course of the program.

I understand that the results of the program will be
strictly confidential. Only group results will be
reported; no individuals will be identified.

I understand that my participation in the program does
not guarantee any beneficial results to me.

I agree that at this time I AM NOT (a) diabetic,

(b) under medical supervision or treatment for high
blood pressure, or (0) pregnant. Should this change
during the course of the program, I will immediately
notify the program leader.

I understand that if I an injured as a result of my
participation in this research project, Michigan State
University will provide emergency medical care if nec-
essary, but these and any other medical expenses must be
paid from my own health insurance program.

 

APPENDIX D

Outlines of the Treatments

267

OUTLINE OF WEIGHT LOSS TREATMENTS

H ! .!. J E: !' I ! !

The nutritional management training method offered to

subjects in this study was based upon Lifestep§;_ng;

2erseaal_Elsn_fer_flsisbt_nanagsmeat (National Dairy
Council, 1985). Lijgfitepg was altered slightly to more

closely fit the research design. The program was
supplemented with additional materials regarding human

nutrition.
5 . H E . I .

1 Explanation of the program.
Discussion of self-monitoring.
Initial research assessment questionaire.

2 Discussion of basic food groups and
daily nutritional needs.
Establishing individualized calorie
reduction plans.
Discussion of serving sizes and portions.

3 Discusssion of menu planning, food
shopping, and substitute foods.

4 Discussion of low calorie foods and
"fad" diets.

5 Problem-solving discussion.
Discussion of food preparation and
recipe exchange.

Mid-program research assessment.

6 Discussion of binge eating and
secretive eating.

268

7 Discussion of dieting in "special"
situations: restaurants and holidays.
Calories and alcohol.

8 Discussion of between-meal eating.

9 Integration of nutritional information
and establishing short-term and long-
term diet plans.

10 Program summary and debriefing.
Final research assessment.
Raffle drawing for monetary prizes.
Wages:

The behavior modification program offered to subjects

in this research study was based upon behavior control

methods widely reported in the weight reduction literature.

Participants were given a weight reduction manual developed

by Brownell (1979) and revised by Mavis (1987). The manual

and lectures were supplemented with additional handouts on

behavior change methods, human nutrition, and exercise.

5 . H 3 . I .

1

Explanation of the program.

Introduction to behavioral aspects of
body weight management.

Initial research assessment questionaire.

Discussion of behavior specification,
goal specification, and self-monitoring.

Application of behavioral methods to
strategies of dieting and exercise.
Revision of individual behavior change
goals and plans.

Developing alternative behaviors:
response substitution.

269

5 Behavioral contracting.
Rewards, reinforcements and punishments.
Mid-program research assessment.

6 Developing control over eating cues:
stimulus control.
Behavioral management of special situations:
restaurants and holidays.

7 Discussion of "sabotage".
Relapse Prevention methods.

8 Cogntive aspects of behavior control:
internal dialogue.

9 Affective aspects of behavior control:
self-esteem, feelings, assertiveness.

10 Program summary and debriefing.

Final research assessment.
Raffle drawing for monetary prizes.

E . I . . I ! !

The exercise training program used in this research
study was based on a model established by the American Heart
Association (1984b). The basic program consisted of low-
risk moderate aerobic exercise, informational lectures and
demonstrations, and written handouts. Exercise participants
were given the Brownell (1979) manual as a supplement to the
program. Brisk walking was the primary form of exercise
promoted in this program and subjects were encouraged to

walk on their own 2 - 3 times per week between program

meetings.
5 . H 5 . I .
1 Explanation of the program.
Three-minute step-test of physical
fitness.

Initial research assessment questionaire.

10

270

Discussion of heart rate monitoring.
Discussion of exercise risks and risk-
management procedures.

Exercise training routine.

Discussion of calorie expenditure during
various types of exercise.
Exercise training routine.
Exercise training routine.

Exercise training routine.
Mid-program research assessment.

Exercise training routine.
Discussion of non-aerobic exercises.

Exercise training routine.
Exercise training routine.
Exercise training routine.
Program summary and debriefing.

Final research assessment.
Raffle drawing for monetary prizes.

APPENDIX E

Measurement Instruments

271

PERSONAL INFORMATION

 

 

Name: Date:

1. Age:

2. Sex: [ ] Female [ ] Male

3. Marital Status: [ 1 (1) Single, never married
I 1 (2) Married
[ 1 (3) Divorced or separated
I 1 (4) Widowed
[ 1 (5) Other:

 

4. Education: Number of years of school

 

5. Occupation:

 

[ ] Full time [ 1 Part time [ ] Not applicable

6. Ethnic Background: [ ](1) Black [ ](2) White
[ 1(3) Hispanic [ ](4) Other:

 

7. What is your height (without shoes)? feet

inches

 

 

8. Age-at-onset: Please indicate the age,at which you first
became overweight

 

9. How many serious attempts have you made at losing

weight?
1 1(1) a few (1-5)
C 1(2) several (6-10)
[ 1(3) numerous (ll-15)
E 1(4) tOO many to count (over 15)

10. Please describe the HEIGHT of the following people
in your life. (circle the appropriate response)

Father: underweight average overweight NA
Mother: underweight average overweight NA

Did/do you have one or more overweight siblings?
yes m__no

Spouse: underweight average overweight NA
Children: underweight average overweight NA

272

11. What are the attitudes of the following peopla about
your attempt(s) to lose weight? Are they:

NEGATIVE -- they disapprove or are resentful

INDIFFERENT -- they don't care or don't help
POSITIVE -- they encourage you

(circle the number representing your response)
(circle NA if the item is not applicable to you)

NEGATIVE INDIFFERENT POSITIVE

Spouse 1 2 3 NA
Children 1 2 3 NA
Mother 1 2 3 NA
Father 1 2 3 NA
Best Friend 1 2 3 NA

12. What different types of weight loss approaches have you
tried in the past? (check all that apply)

Drugs/Amphetamines

Surgical (Bypass or Stapling)
Jaw Wiring

Acupuncture
PsychotheraPY/Hypnosis
Self-help groups

Behavior Modification

Exercise programs/Fitness Clubs
Specific diet plans:

HHHHHHHHH
UHL—lh—ll—lt—lt—lt—AI—J
AAAAAAAAA
madman; GNU—-
vvvvvvvvv

JBeverly Hills Diet

JFat Counter Guide

JKemper Rice Diet/Duke Univ. Diet
JLast Chance Refeeding Diet
JPritikin Diet

JScarsdale Diet

JSlim Chance in a Eat World
JStillman Diet

JTake Off Pounds Sensibly (TOPS)
]Weight Watchers
JOther (specify)

HHHf—‘HHHHHHH

 

[ ](10) Some other method (specify)

 

273

Below is a list of questions designed to assess your view of
the weight loss method being taught in your group. Please
answer each question by marking the appropriate response.

1. The weight loss method taught in my group is logical.

strongly disagree

disagree

neither agree nor disagree
agree

strongly agree

2. The weight loss method taught in my group will help me
to lose weight.

strongly disagree

disagree

neither agree nor disagree
agree

strongly agree

3. The weight loss method taught in my group is likely to
help others lose weight.

4. I believe the

strongly disagree
disagree .
neither agree nor disagre
agree

strongly agree

most effective means for me to lose weight

is (circle one):

a. nutritional management
b. behavior management
c. exercise

274

W

Below is a list of situations in which people often have
trouble maintaining their weight loss efforts. Please read
each one carefully. Then circle the number which best
describes BOW CONFIDENT YOU ARE THAT YOU WOULD BE ABLE TO
MAINTAIN YOUR WEIGHT LOSS STRATEGIES IN THAT SITUATION. If
you are absolutely certain that you would continue to use
your weight loss methods. then circle 10. If you have no
confidence in your ability to continue using your weight
loss strategies in that situation, circle 0.

Most likely, your confidence will vary from situation to
situation. Circle the appropriate number to rate your
degree of self-confidence for each situation.
lot at all 5942:1141!
Confident ”Confident

 

1. When you feel really happy. 4 1 2 3 4 5 4
2. When you feel anxious. 4 1 z 3 4 5 s

3. When you want to sit back and enjoy 4 1 z 3 4 5 5
a cigarette.

4. When you are nervous. 4 1 2 J 4 s 6
5. When you feel annoyed. 4 1 z 3 4 5 4
6. When you want to relax. 4 l 2 1 4 s 4
7. When you are worried. I l 2 1 4 5 4
8. When you feel angry. I 1 2 3 4 5 4
9. When you feel tired. 4 1 2 J 4 5 4
10. When you feel embarassed. 4 1 2 3 4 5 4
11. When you feel bored. 4 1 2 1 4 4 4

12. When you feel you need more energy. 4 l 2 3 4 s 4

13. When you are drinking, or want to 4 1 2 4 4 5 4
drink alcohol.

14. When you see others eating. 4 1 2 1 4 4 4

15. When you want to reward yourself 4 1 2 J 4 5 4
for something you have done.

16. When someone offers you food. 0 l 2 J 4 5 6

Celletei:

._.______Cc:fi:':::

17.

18.
19.
20.

21.

23.
24.
25.
26.
27.

28.

29.
30.

When you are waiting for someone
or something.

When you feel uncomfortable.
When you want to cheer up.

When you want to avoid smoking
or drinking.
When you feel depressed.

When you want to take a break
from work or other activities.

When you are overly excited.
When you feel upset.

When you feel frustrated.
When

When you feel overwhelmed and
don't know what to do first.

When you are concerned about
money problems.

When a crisis occurs.

On special celebrations like
holidays or birthdays.

you are angry with yourself.

275

lot 44 411 lodetately 4034104411
enumn___tu114c-L_____Jm1m.'

I 1 2 J 4 5 4 1 I I H

276

W

Below you will find a series of statements. Please read
each statement carefully and respond to it by expressing THE
EXTENT TO WHICH YOU BELIEVE THE STATEMENT APPLIES TO YOU.
For all items a response from 1 to 7 is required. Use the
number that best reflects your belief when the scale is
defined as follows:

the statement doesn‘t apply to me at all.

the statement usually doesn't apply to me.

most often. the statement does not apply.

I am unsure about whether the statement applies
to me, or it applies to me about half the time.
the statement applies more often than not.

the statement usually applies to me.

the statement always applies to me

~20)“ fiUNH

It is important that you respond to all items.

44911:: 1/2 41:

 

doesn't tile. or 41:11:
apply allure 44:11::
1. I prefer a job where I have a lot 1 2 3 4 5 6 7
of control over what I do and when
I do it.
2. I enjoy political participation because 1 2 3 4 5 6 T
I want to have as much of a say in
running government as possible.
3. I try to avoid situations where 1 2 3 4 5 6 7
someone else tells me what to do.
4. I would prefer to be a leader rather 1 2 3 4 5 6 7
than a follower.
5. I enjoy being able to influence 1 2 3 4 5 6 7
the actions of others.
6. I am careful to check everything on 1 2 3 4 5 6 7
an automobile before I leave for a
long trip.
7. Others ususally know what is best 1 2 3 4 5 6 7
for me. '
8.'I enjoy making my own decisions. 1 2 3 4 5 6 7,
9. I enjoy having control over my own 1 2 3 4 5 5 7
destiny.
10. I would rather someone else took over 1 2 3 4 5 6 7

the leadership role when I'm involved
in a group project.

continued on next page

ll.
12.

l3.

14.

15.
16.

17.
18.
19.

20.

continued from previous page

277

I consider myself to be generally more

capable of handling situations than
others are.

I'd rather run my own business and
make my own mistakes than listen
to someone else‘s orders.

I like to get a good idea of what a
job is all about before I begin.

When I see a problem I prefer to
do something about it rather than
sit by and let it continue.

When it comes to orders. I would
rather give them than receive them.

I wish I could push many of life's
daily decisions off on someone else.

When driving. I try to avoid putting
myself in a situation where I could
be hurt by someone else's mistake.

I prefer to avoid situations where
someone else has to tell me what it
is I should be doing.

There are many situations in which
I would prefer only one choice
rather than having to make a
decision.

I like to wait and see if someone
else is going to solve a problem

so that I don’t have to be bothered
by it.

duurl
nu!

mm: 1]! Us

41041:
44411::

 

l

414:, 04

mm
4 5
4 5
4 5
4 5
4 5
4 5
4 5
4 5
4 5
4 S

7

278

BeliefthQutJieighLLms

Below are four questions designed to assess your beliefs
about losing weight. Please circle the appropriate number
to express the degree to which you agree with the statement.

4410441! 4410441!

 

41mm “n. um
1. Whether I gain, lose. or maintain my 1 2 3 ‘ 5 6
weight is entirely up to me.
2. Being the right weight is largely a 1 2 3 4 5 6
matter of good fortune.
3. No matter what I intend to do, if I
gain or lose weight. or stay the same 1 2 3 4 5 5
in the near future, it is just going

to happen.

4. If I eat properly, and get enough
exercise and rest. I can control my 1 2 3 4 5 -
weight in the way I desire. °

279

W

Please complete the following scales. indicating how you
feel about various aspects of the program by putting an X in
the appropriate space for each item.

The program leader is:

1. Pleasant ___:___:___:___:___ Unpleasant
2. Valuable ___t___:___:___:___ Worthless
3. Very Helpful __:_:___:_:__ Unhelpful
4. Very ___:___:___:___:___ Not Very
Motivating Motivating
5. Very Actively ___:___:___:___:___ Passively
Involved Involved

The program materials are:

6. Very Helpful ___ ___ ___:___:___ Unhelpful

7. Interesting ___:___:___:___:___ Boring
8. Very _:_:___:_:__ Not Very
Motivating Motivating

The weight loss methods taught in my group .re:

9. Very Helpful ___:___:___:___:___ Unhelpful
10. Interesting ___:___:___:___:___ Boring
11. Easy to do ___:___:___:___:___ Difficult to do

12. The weekly weight loss goals are much too difficult to
achieve.

strongly agree ___:___:___:___:___ strongly disagree

13. The weight loss methods suggested in my group are rigid
and very limiting.

strongly agree ___:___:___:___:___ strongly disagree

14. The program overall is too restricting --- there are
too many rules and regulations.

strongly agree ___:___:___:___: . strongly disagree
15. The monetary incentive scheme (or lottery) is unfair.

strongly agree ___:___:___:___:-__ strongly disagree

280

Was

PLEASE HELP ME complete my doctoral research by completing
the following questionaire and returning it in the enclosed
envelope. A few minutes of your time will mean a great deal

to me.
In August you joined a weight loss program at Michigan State
University. After a few weekly meetings you stopped coming.

Please help me with my research by indicating the reasons
why you stopped participating in the program.

Check any items that apply to you.

1. I did not like the 112: of weight loss program that
was being offered.

. I did not like the program leader.

___ 3. The program was not helping me to lose weight.

.___ 4. The day and time of the program meetings was not
convenient for me. (ie, schedule conflicts)

___ 5. Unanticipated events in my life interfered with my
ability to continue participating.

___ 6. I realized that I was not personally ready to lose
weight at that time.

7. Other (please specify):

 

 

 

Thankyou very much,

Thomas C. Fuller

BIBLIOGRAPHY

BIBLIOGRAPHY

Abraham. 5. (1983) WWW
Stetee, Vital Health Statistics, Series II, No.230, DHHS
Publication Number 83-1680, Public Health Service,
Washington, D.C.: U.S. Government Printing Office.

American Heart Association. (1984a) Exeteiee_egg_yett_heett.

Dallas, TX: American Heart Association.

American Heart Association. (1984b) Weltigg_tet_e_tee1thy
heett. Dallas, TX: American Heart Association.

American Heart Association. (1984c) Exeteiee_e;ezy. Dallas,
TX: American Heart Association.

American Heart Association. (1984d) figimming_fiet_e_heelthy
neezt. Dallas, TX: American Heart Association.

Annual Report of the National Institutes of Health program
in biomedical and behavioral nutrition research and
training, FY 1982. (1985) In J. Hirsch and T.B.
VanItallie (Eds.), e v ' obe 't esea ch:
IX. London: John Libbey and Co.

Appel, C. P. (1986) From contemplation to determination:
Contributions from cognitive psychology. In W.E. Miller

and N Heather (Ede-L W. New
York: Plenum Press.

Averill, J.R. (1973) Personal control over aversive stimuli
and its relationship to stress. 0 ' a et'
80, 286-303.

Balch, P. and Ross, A.W. (1975) Predicting success in weight
reduction as a function of locus of control: A
unidimensional and multidimensional approach. goutnal of

Q9nsultias_and_§linisal_£§22helegz. 43(1), 119-

Bandura, A. (1982) Self-efficacy mechanism in human agency.

W 37 122-147

Bandura, A. (1977) Self-efficacy: Toward a unifying theory
of behavior change. Esychotggieal Review, 84, 191-215.

281

282

Bellack, A.S. (1975a) Self-evaluation, self-reinforcement,

and locus of control. leuznal_2f_8e§earsh_ih
Eersenalitx. 9,-158-167.

Bellack, A. S. (1975b) Behavior therapy for weight reduction.
Addicti_e_neha_iera 1 73- 82-

Berchtold, P., Berger, M., Greiser, E., Dohse, M., Irmscher,
E., Gries, F.A., and Zimmerman, H. (1977) Cardiovascular
risk factors in gross obesity. ’ n o o

92:31:17 1: 219-229-

Bjorntorp, P. (1978) Physical training in the treatment of
obesity. IntsInatienal_leu:nal_ef_ghe§itx. 2. 149 156-

Bjorntorp, P., deJonge, H., Krotkiewski, M., Sullivan, L.,
Sjostrom, L., and Sternberg, J. (1973) Physical
training in human obesity. III. Effects of long-term
training on body composition. Meteteliem, 22, 1467-1475.

Bjorvell, H., Edman, G;, Rossner, S., and Schalling, D.
(1985) Personality traits in a group of severely obese
patients: A study of patients in two self-chosen weight

reducing programS- Internaticnal_leurnal_9f__besitx 9
257-266.

Blackburn, G.L. and Read, J.L. (1984) Benefits of reducing

-- revisited. R2stgraduate_nedisal_leurnal.
60(Supplement 3), 13-18.

Blundell, J.E. (1984) Behavior modification and exercise in

the treatment of obesity- E2stgraduats_usdisal_geurhal.
60 (Supplement 3), 37-49.

Borkovec, T. D. and Nau, S.D. (1972) Credibility of analog
therapy rationales. v'o

Excerimental_£sxshiatrx 3 257-260.

Bornstein, P.H. and Rychtarik, R.G. (1983) Consumer
satisfaction in adult behavior therapy: Procedures,

problems, and future perspectives. Beheyiet_1hetepy, 14,
191-208.

Bray, G.A. (1979) Obesity in America. t at' Jou nal
21.9besitx

, 3, 363-375.

Bray, G.A. (1978) Definition, measurement, and
classification of the syndromes of obesity.

Iaternatienal_ieurnal_ef_gbesitx. 2. 99-112.

283

Bray, J.H. and Maxwell, S.E. (1982) Analyzing and

interpreting significant MANOVAs. Bcvicw c: Educational
BEEQBIQD: 52(3): 340-357-

Brehm. 5.5. and Brehm. J-W. (1981) Ps22holosisal.reastanss_
. New York: Academic
Press .

Brownell, K.D. (1985) The contribution of psychological and
behavioral factors to a classification scheme for
obesity. In J. Hirsch and T. B. VanItallie (Eds.), Recent

adxansss_in_obe§itx_ressarsni_ly London: John Libbey and
co.

Brownell, K. D. (1982) Obesity: Understanding and treating a
serious, prevalent, and refractory disorder. Jccrncl_c;

Qonsgltins_and_glinisal_zsxsnologL 50(6), 820- 340.

Brownell. K-D. (1979) Behaxi2r_theranx_for_wsisht_oonfrol;_é
crcccmcn§_ncnccl. Unpublished manual.

Burger, J. M. and Cooper, H. M. (1979) The desirability of
COflttOl- Noti_atign_and_3motion. 3(4), 331-393-

Burton, R. T. Foster, W. R., md VanItallie, T. B., (1985)
Health implications of obesity: An NIH consensus
development conference. In;c:naticncl_1ccrncl_cfi
chsi;y,9,155-169.

Chambliss, C.A. and Murray, E.J. (1979) Efficacy
attributions, locus of control, and weight loss.

Qognitixs_1neranx_and_aesearsn. 3. 349-354.

Coates, T. J. (1977) Theory, research and practice in
treating obesity: Are they really all the same?
Addis§i_s_fien§_igzfi 2. 95-103-

Cohen, J. and Cohen, P. (1975) Applicc_mclciclc

-n --~ ., ., 5 . ., ... e -- -v . a

ccicnccg. Hillsdale, N. J.: Lawrence Erlbaum.

Colliver, J. A., Frank, S., and Frank, A. (1983) Similarity
in obesity indices in clinical studies of obese adults: A
factor analytic study. a u a o n 'ca

HBEIiSiQD 38: 640 547-

Colvin, R.H., and Crist, K. (1983) The futility of follow-
up in current behavioral studies of obesity. Behavior

Inerasist, 6(8), 152-153-

284

Consumers Union of the United States (1980) acclch
Qccckczy. New York: Holt, Rinehart, and Winston.

Costello, R. M. (1975) Alcoholism treatment and evaluation:

In search of methods. In1eInsf1onal_lournel_of_fhs
Addist1on§ 10(2). 251- -275.

Currey, H., Malcolm, R., Riddle, E., and Schacte, M. (1977)
Behavioral treatment of obesity: Limitations and results
with the chronically obese. ‘

Medisal_Assos1at1on 237(36). 2829-2831.

Dahlkoetter, J., Callahan, E.J., and Linton, J. (1979)
Obesity and the unbalanced energy equation: Exercise
versus eating habit change. 0 u t' and

Qlini£§l_E§¥QhfllQQ¥ 47(5), 898-905-

Demobrowski, T.M., MacDougall, J.M., and Musante, L. (1984)
Desirability of control versus locus of control: A
relationship to paralinguistics in the Type A interview.

fisal£h_£§¥§thQ§¥: 3(1): 15-25-

Dempsey, J.A. (1964) Anthropological observations on obese
and non-obese young men undergoing a program of vigorous
physical exercise. Bcccarcn_gg§;tcrly, 35, 275-287.

Devine, D.A. and Fernald, P. S. (1973) Outcome effects of
receiving a preferred, randomly assigned of non-preferred

therapy 12urnal_2f_Q2nsu111ns_snd_Qlinisal_2§xshologL
41(1), 104-107.

DiClemente, C.C. (1981) Self-efficacy and smoking cessation

maintenance: A preliminary report. cognicivc Ihcrapy and
Bfifiéslsh, 5(2): 175-137-

Dishman, R.K. (1982) Compliance/adherence in health-
related exercise. ficc1;n_£§ycnclccy, 1(3), 237-267.

Douglas, J.G., Ford, M.J., and Munro, J.F. (1981) Patient
motivation and predicting outcome in a hospital obesity

clinic. In1ernat12211_lournal_of_gbs§1tx. 5, 33-38.

Ducimetiere, P., Richard, J., and Cambien, F. (1986) The
pattern of subcutaneous fat distribution in middle-aged
men and the risk of coronary heart disease: The Paris
prospective study. u O e t
10(3), 229-240.

Duddleston, A. K. and Bennion, M. (1970) Effect of diet
and/or exercise on obese college women. Jou o the

Amsrisan.£1sts_12§_A§§921111.3 56 126- -129-

285

Dwyer, J. (1985) Classifying current popular and fad diets.
In J. Hirsch and T. B. VanIntallie (Eds.), gccccc advances

1n_obe§111_rs§earoh1_l_ London: John Libbey and Co

Eufemia, R.L. and Wesolowski, M.D. (1985) Attrition in
behavioral studies of obesity: A meta-analytic review.

Behaxior_Ther191§§. 8(6). 115-116-

Ewing, J.A. (1977) Matching therapy and patients: the
cafeteria plan. Br111§h_lournel_of_agdist1on, 72 13-18

Frankle, R.T. (1985) Dietotherapy in the treatment of
disease. In J. Hirsch and T. B. VanItallie (Eds.), Bccccc
adzan2es_1n_obs§112_re§eersh1_lxo London: John Libbey

and Co.

Franklin, B. A. (1984) Exercise program compliance. In J.
Storlie and H.A. Jordon (Eds. ), v'o e o
chccicy. Jamaica, NY: Spectrum.

Franklin, 8., Buskirk, E., Hodgson, J., Gahagan, H.,
Kollias, J., and Mendez, J. (1979) Effects of physical
conditioning on cardio-respiratory function, body
composition, and serum lipids in relatively normal-weight
and obese middle-aged women. Inccrngcicncl_1cczncl_cg
Qpccicy, 3, 97-109.

Garner, D.M., Garfinkel, P. E., Stancer, H. C., and Moldofsky,
H. (1976) Body image disturbances in anorexia nervosa

and obesity B:rshosoma112.fiedisine. 38(5), 327- -335.

Gatchel, R.J. and Baum, A. (1983) An incrcdcccicg cc health
REYEthQQY- New York: Random House.

Glass. G-V- and Hopkins, K-D. (1984) 51111511211_121hods_1u
sdusation.and_nsxsholosr- Engelwood Cliffs, N.J.:

Prentice-Hall Inc.

Goldney, R. D. and Cameron, E. (1981) Locus of control as a
predictor of attendance and success in the management of

obesity- Interna112nel.lournal_of_gbe§1tL 5. 39- 43

Gormally, J., Rardin, D., and Black, S. (1980) Correlates
of successful responSe to a behavioral weight control

clinic. lournal_of_§ounselins_£§xshologx. 27(2), 179-
191.

Gwinup, G. (1975) Effect of exercise alone on the weight of
obese women. v e 'c e, 135, 676-
680.

286

Hadjiolova, I., Mintcheva, L., Dunev, S., Daleva, M.,
Handjiev, S., and Balabanski, L. (1982) Physical
working capacity in obese women after an exercise
programme for body weight reduction. IBLQEDQLiQLQl

lonrnal.of_9besitL 6 405-410.

Hall, J.C., Veale, B., Horne, K., and Watts, J.M. (1984)
The nutritional knowledge scores of morbidly obese
patients selected for gastric bypass surgery.

In1erna1ional_lournal_of_gbesity. 8. 123-123.

Hand. Doc. and Taylor. Coc. (1987) Mulrixariate_analysis_21
zar1anse_and_repsate§_msasgre§- New York: Chapman and
Hall.

Harris, M. B. (1983) Eating habits, restraint, knowledge and

attitudes toward obesity International_lournal_of
g§c§1§y, 7, 272-186.

Hartigan, K. J., Baker-Strauch, D., and Morris, G. W. (1982)
Perceptions of the causes of obesity and responsiveness

to treatment- 1ournal_of_Q92nssl1ns_£§yshologL 29(5),
473-435.

Helson, H. (1959) Adaptation level theory. In S. Koch

(Ed-). 2sxsho1osx1_A_st1§y_o£_a_ssienss_IYol1_1l- New
York: McGraw-Hill.

Helson, H. (1964) Adaptationzleyel_1heory- New York: Harper

and Row.

Herman, P.C. and Polivy, J. (1975) Anxiety, restraint, and

eating behavior. 1onrna1.2:.Abnormal_£sysholosy. 84,
666-672.

Hirsch, J. (1985a) The present status of obesity research.
In J. Hirsch and T. B. VanItallie (Eds.), gcccgc_ccyccccc

1n_obes1§y_resssrsh11¥.- London: John Libbey and Co-

Hirsch, J. (1985b) Dietary treatment. In J. Hirsch and

T.B. VanItallie (Eds.), accent cdvcgccc 1g obcsity
:cccczcn1ly. London: John Libbey and Co.

Hoiberg, A., Berard, S., Watten, R. H., and Caine, C. (1984)
Correlates of weight loss in treatment and at follow-up.

In_erna112nal.£ournal_o£_9b_§11L 8 457- 465-

Holm, C., Jacobsson, B., Holm, J., Bjorntorp, P., and Smith,
U. (1977) Effects of submaximal physical exercise on
adipose tissue metabolism in man. nternat'ona Journal

of_Qbs§1ty. 1, 249-257-

287

Holmes, N., Ardito, E.A., Stevenson, D., and Lucas, C.P.
(1984) Maintenance of weight loss in a heavily
overweight population. In J. Storlie and H.A. Jordan
(Eds-). Beha21oral_manasemsn§_of_obesitr- Jamaica, NY:

Spectrum.

James, W.P.T. (1984) Dietary aspects of obesity.
’ ' , 60(Supplement 3), 50-55.

Janis, I.L. and Mann, L. (1977) Dcc1§1cn_mck1cc. New

York: Free Press.

Jeffery, R.W., Bjornson-Benson, W.M., Rosenthal, B.S.,
Lindquist, R.A., and Johnson, S.L. (1984) Correlates of
weight loss and its maintenance over two years of follow-
up among middle-aged men. Ersyentiys_ued1sins. 13(2),
155-168.

Jeffery, R.W., Snell, M.K., and Forster, J.L. (1985) Group
composition in the treatment of obesity: Does increasing
group homogeneity improve treatment results? fichcy1c;
Researsh_and_1herapx. 23(3). 371-373-

Jeffery, R.W., Wing, R.R., and Stunkard, A.J. (1978)
Behavioral treatment of obesity: The state of the art

1976. Behayior_1herapy. 9. 189-199-

Kaplan, R.M., Atkins, C.J., and Reinsch, S. (1984) Specific
efficacy expectations mediate exercise compliance in

patients with c0po. ngg1gn_2§ycng1ch, 3, 223-242.

Keesey, R.E. (1980) A set-point analysis of the regulation
of body weight. In A. Stunkard (Ed. ), ggcc1cy.
Philadelphia: Saunders Co.

Kirk, R.E. (1982) ' e ' : oc u e o the
behaxioral_§sieuse§- Monteray, CA: Brooks/Cole-

Kissin, B., Platz, A., and Su, W.H. (1971) Selective
factors in treatment choice and outcome in alcoholics.
In N. K. Mello and J. H. Mendelson (Eds. ), Reccnt advacces
1n_s12111s_2n_alsohol1sm- Washington, D C-: U.S.

Government Printing Office.

Krantz, D. S., Baum, A., and Wideman, M. V. (1980) Assessment
of preferences for self-treatment and information in
health care. l9BInél_Qf.£§£§2£§li£¥.éflé.§2§iél

Esysholosy 39(5)) 977‘990-

288

Krantz, D.S. and Deckel, A.W. (1983) Coping with coronary
heart disease and stroke. In T.G. Burish and L.A.
Bradley (Eds.), ' w o ' ' e ° esea c and

' ' . New York: Academic Press.

Langer, E. and Rodin, J. (1976) The effects of choice and
enhanced personal responsibility for the aged: A field
experiment in an institutional setting. Jccrnc1_cfi

2ersona1112.1nd_§osial_£§ysholosx. 34 191-198-

Larsson, B., Bjorntorp, P., and Tibblin, G. (1981) The
health consequences of obesity. ' u of

9233132: 5, 97'116-

Lauer, J.B., Wampler, R. S., Lantz, J. B., and Romine, C. J.
(1979) Psychosocial aspects of extremely obese women

joining a diet group Interna1ional_lournal_of_gbesity.
3, 153-161.

Lennon, D., Nagle, F., Stratman, F., Shrago, E., and Dennis,
S. (1985) Diet and exercise training effects on resting

metabolic rate- International_lournal_ofi_gbesitL 9 39-
47.

Leon, A.S., Conrad, J., Hunninghake, D.B., and Serfass, R.
(1979) Effects of a vigorous walking program on body
composition, and carbohydrate and lipid metabolsim in

obese young men. Amer1gan_lournal_of_slinisal_Nutr11122,
33, 1776-1787.

Lowe, M. R. and Fisher, E. B. (1983) Emotional reactivity,
emotional eating, and obesity: A naturalistic study.

lQEInél.91.fl§hé.i§£§l.fl§§i§ifl§p 5(2): 135 149-

MacKeen, P.C., Franklin, B.A., Nicholas, W.C., and Buskirk,
E.R. (1983) Body composition, physical work capacity,
and physical activity habits at 18-month follow-up of
middle-aged women participating in an exercise

intervention program- In1erna1ional_lournal_2§_gbs§111,
7, 51-71.

Mahoney, M.J. (1974) Self-reward and self-monitoring

techniques for weight control. achcvic; Thegapy, 5, 48-
57.

Mairken, G. (1981) How to pick the right therapist.
Qonsumer.s_n1se§t 20(4). 36-40.

Mann, L. and Janis, I. (1982) Conflict theory of decision
making and the expectancy-value approach. In N.T.
Feather (Ed. ), on e ta -va ue

289

m9g§1§_ig_p§yghglggy. Hillsdale, N.J.: Erlbaum.
Marlatt, G. A. and Gordon, J.R. (Eds ) (1985) e se

9 ‘V,! .'!' 92'! '12! 1 f, at‘- 1 1‘ e1 u‘n O.

aggig§12§_§gn§gig:§. New York: Guilford Press.

Marston, M. V. (1970) Compliance with medical regimens: A
review of the literature. Ngzsing_fig§§§;gn, 19, 312- 323.

Mavis, B. (1987) The use of monetary incentives in a social
context in the modification of weight loss. Unpublished
dissertation, Michigan State University.

Metropolitan Life Foundation (1983) Height and weight
tables. New York: Metropolitan Life Insurance Company.

Miller, J.W. and Schildkraut, M.L. (1983) How good are the

quick weight loss clinics? QQQQ_figg§ghggpigg, 196(4),
225-226.

Miller, P. M. and Sims, K. L. (1981) Evaluation and component
analysis of a comprehensive weight control program.

Interna112nal_lournal_of_gbesitx 5 57‘55-

Miller, W.R. and Hester, R. K. (1986) The effectiveness of
alcohol treatment: What research reveals. In W.E. Miller

and N- Heather (Eds.), Treating_Add1§tixe_Behaxiors. New

York: Plenum Press.

Murray, D.C. (1976) Preferred versus nonpreferred and self-
control training versus determination raising as
treatments of obesity: A pilot study. o 'ca

Bengzgs, 38, 191-198.

National Dairy Council (1985) L1fe_tep§__xour_nersonal_21_n
fig;_ggign;_m§nag§mgn;. Rosemont, IL: National Dairy

Council.

O'Connell, J.R., Price, J.B., Roberts, S.M., Jurs, S.G., and
McKinley, R. (1985) Utilizing the health belief model to
predict dieting and exercise behavior of obese and
nonobese adolescents. o a e , 12(4),
343- 351.

O'Leary, A. (1985) Self-efficacy and health. Behaylg;
B§§§3223_3nd_1n§2é211 23(4): 437‘451-

Orford, J. (1985) e 'v et't s. New York: John
Wiley and Sons.

290

Oscai, L. B. and Williams, B. T. (1968) Effects of exercise
on overweight middle-aged males. o th e 'can

Geriatries_fiosietx 16(7). 794-797

Parker, M.W., Winstead, D. K., and Willi, F. J. P. (1979a)
Patient autonomy in alcohol rehabilitation. I. Literature

review. Interna112na1_1surnal_2f_the_Addistion§ 14(7),
1015- 1022.

Parker, M.W., Winstead, D. K., Willi, F. J. P., and Fisher, P.
(1979b) Patient autonomy in alcohol rehabilitation. II
Program evaluation.

Addistions.14(8). 1177-1134-

Pekarik, G., Blodgett, C., Evans, R.G., and Wierzbicki, M.
(1984) Variables related to continuance in a behavioral

weight loss program. Addistixe.§ehaxior§, 9, 413-416.

Perlmuter, L.C. and Monty, R.A. (1979) ' 'ved
ggntrgl. Hillsdale, N.J.: Lawrence Erlbaum Associates.

Perri, M.G., McAdoo, W.G., MCAllister, D.A., Lauer, J.B.,
and Yancey, D.z. (1986) Enhancing the efficacy of
behavior therapy for obesity: effects of aerobic
exercise and a multicomponent maintenance program.

I2urnal_9f_Q9nsult1ns_end_§linigal_£§x§hol_gx 54(5);

670-675.

Pi-Sunyer, F. X. (1985) Effect of exercise on food intake.
In J. Hirsch and T. B. VanItallie (Eds.), Recent advances

in_obesitx_researsh1_1_ London: John Libbey and Co-

Piziak, V.K. (1983) Medical management of obesity.

Bosfgraduate_nedisine. 74(5). 158-163 and 166 and 171-
172.

Porcello, L. A. (1985) Characteristics of professional
weight control programs. In J. Hirsch and T. B.

VanItallie (Ede-). Besent_adxanses_1n_obesitx
reggaggnily. London: John Libbey and Co.

Rodin, J., Bray, G.A., Atkinson, R.L., Dahms, W.T.
Greenway, F. L., Hamilton, K., and Molitch, M. (1977)
Predictors of successful weight loss in an outpatient
obesity clinic. t J n 0 e 't l, 79-
87.

Rosenstock, I.M. (1966) Why people use health services.
' un u e , 44, 94-127.

291

Rotter, J. B. (1975) Some problems and misconceptions
related to the construct of internal versus external
control of reinforcement.

Qlini§§l_25¥§thQQ¥I 43(1)I 55-67-

Rotter, J.E. (1966) Generalized expectancies of internal
versus external control of reinforcement. Psychologigal

WI 80(1) I 1'28°

Sakoda, J. M., Cohen, B. H., and Beall, G. (1954) Tests of
significance for a series of statistical tests.

£§2§h912g1211_flulletinI 51(2)I 172-175.

Saltzer, E. B. (1982) The weight locus of control (WLOC)
scale: A specific measure for obesity research. Jo a

of_£er§ona1111_A§ses§mentI46(6)I 620-628-

Saltzer, E.B. (1978) Locus of control and the intention to
lose weight- Health_Edngation_Mon23112h§I 6(2). 118-128.

Sanchez-Craig, M. (1980) Random assignment to abstinence or
controlled drinking in a cognitive-behavioral program:
Short-term effects on drinking behavior. Addiggiyg
B§h§_iQI§I 5 35-39-

Sandifer, B. A. and Buchanan, W. L. (19 83) Relationship
between adherence and weight loss in a behavioral weight

reduction program. figh§_ig;_lngrapy, 14, 682- 288.

Sankey, K. (1984) The practical effect of obesity on the
insured life --- relative risks associated with disease
states --- "the impaired life". u e a

193;ng1,60(8upplement 3), 5- -12.
Sash, S. E. (1977) Why is the treatment of obesity a. failure

in modern society? on 0 t , 1,
247-248.

Schaller, W.E. and Carroll, C.R. (1976) Health, guackegy,
ang_;ng_ggn§gm§;. Philadelphia, PA: W.B. Saunders and
Co.

Schichting, P., Hoilund-Carlsen, P.F., and Quaade, F. (1981)
Comparison of self-reported height and weight with
controlled height and weight in women and men.

In2ernational_lournal_of_gbe§1_xI SI 67- 76.

Simopoulos, A.P. (1985) Research trends related to the
study and management of obesity. In J. Hirsch and T.B.
VanItallie (Eds.), e v ' ' e e h:
12. London: John Libbey and Co.

292

Sjoberg, L. (1985) Motivation and compliance: How assessed
and how important? In J. Hirsch and T. B. VanItallie
(Eds.), v ' e : V. London:
John Libbey and Co.

Snetselaar. L-G- (1983) nn1rition_ceunse11ne_skill§.
Rockville, MD: Aspen Systems Corp.

Stallones, R.A. (1985) Epidemiologic studies of obesity.
Annals_of_1nternal_ned121neI 103(6 pt- 2)I 1003- 1005

Stalonas, P.M., Johnson, W.G., and Christ, M. (1978)
Behavior modification for obesity: the evaluation of
exercise, contingency management, and program adherence.

I 45(3)I
463-469.

Sternberg, B. (1985) Relapse in weight control:
Definitions, processes, and prevention strategies. In

G.A. Marlatt and J.R. Gordon (Eds.), Eglgp§§_prgy§gg1gg.

New York: Guilford Press.

Storlie, J. (1984) Psychological factors related to eating
and activity behaviors. In J. Storlie and H. A. Jordan

(Eds.)I Exalua112n_and_treafmenf_ef_ebeeitx- New York:

Spectrum Publications.

Strain, G. W. and Strain, J. J. (1979) Psychological
impediments to weight 1053- International_ioerna1_ef

ngg1ty, 3, 167-170.

Stunkard, A.J. (1978) Behavioral treatment of obesity: the

current status. Internaf12nel.leernel_of_gbesitxI 2.
237-248.

Stunkard, R. B. (1975) From explanation to action in
psychosomatic medicine: The case for obesity.

Psychoegmatis_negicine 27 195- 235.

Stunkard, A.J. and Pencik, S.B. (1979) Behavior modification

in the treatment of obesity. Argh1yg§_g§_§gggra1
Reichieter 36. 801-806.

Thornton, C.C., Gottheil, E., Gellens, H.K., and Alterman,
A.I. (1977) Voluntary versus involuntary abstinence in

the treatment of alcoholics. lggrngl_gfi_§§gg1g§_gn
Algghgl, 38(9), 1740-1748

Tobias, L.L. and MacDonald, M.L. (1977) Internal locus of
control and weight loss: An insufficient condition.
C s ' ' 'ca 5 cho o , 45(4),

293

647-653.

Tremblay, A., Fontaine, E., Poehlman, E.T., Mitchell, D.,
Perron, L., and Bouchard, C. (1986) The effect of
exercise training on resting metabolic rate in lean and
moderately obese individuals. In;arna§1gna1_laarna1_af
Qpa§1§y, 10(6), 511-517.

Truswell, A. S. (1985) Obesity: Diagnosis and risks.
B1111§h_Medieal_£ournelI 291(7). 655-657.

VanItallie, T. B. (1978) Diets for weight reduction:
Mechanisms of action and physiologial effects.

IDE2ID3S12n3l_lQBID§l_Qfi_Q§§§i§II 2! 113-122-

VanItallie, T.B. and Abraham, S. (1985) Some hazards of
obesity and its treatment. In J. Hirsch and T.B.
VanItallie (Eds.), v ' 't e c
1_. London: John Libbey and Co.

VanKoten-Chappell, C. (1982) Self-efficacy as a predictor
of weight loss. Unpublished dissertation, University of
Oregon.

Vannicelli, M. (1979) Treatment contracts in an inpatient

alcohol treatment setting. lggzna1_gfi_§§ga1a§_afi
Algghgl, 40(5), 457-471.

Vertes, V. (1984) Very low calorie diets --- history,

safety, and recent developments. Egaggragaaga_nag1aa1
lggznal, 60 (Supplement 3), 56- -58.

Vickery, D.M. and Fries, J.F. (1976) Iaka gage g: yourse;§:
E , .3 ! 3° 1

. Reading, Mass.:
Addison-Wesley.

Wallston, 3.8. and Wallston, K.A. (1978) Locus of control
and health: A review of the literature. t u 'on

MQBQQIQRhfiI 6(2)I 107‘117-

Wallston, B.S., Wallston, H.A., Kaplan, G.D., and Maides,
S. A. (1976) Development and validation of the health
locus of control (HLC) scale. Jo onsu t'n and

911n1211_£§29h21231I44(4)I 580- -535.

Weinberg, R.S., Hughes, H.H., Critelli, J.W., England, R.,
and Jackson, A. (1984) Effects of preexisting and
manipulated self-efficacy on weight loss in a self-

control program- 12urnel_9f_Besearch_1h_2ersehalitxI 18,
352-258.

294

Weiss, A.R. (1977) Characteristrics of successful weight
reducers: A brief review of predictor variables.

Addigfize_flehaxier§. 2. 193-201-

Winer, B.J. (1971) ' ' ' ' ' e e ' e ta
gg§1gn. New York: McGraw-Hill.

Wing, R.R. and Jeffery, R.W. (1984) Sample size in clinical
outcome research: The case of behavioral weight control.

Behaxier_1heranx. 15. 550-556.

Wilner. D.N- (1968) A_nzasf1sal_guide_12_p§xsh21herenx- New
York: Harper and Row.

Wilson, G.T. (1985) Psychological prognostic factors in the
treatment of obesity. In J. Hirsch and T.B. VanItallie
(Eds.), v ' obes' c : V. London:
John Libbey and Co.

Wilson, G.T. (1980) An evaluation of behavior therapy in
obesity. InLernat12nal_lgurnel_ef_gbesifx 4 371- 376

Winefield, H. R. (1982) Reliability and validity of the

health locus of control scale. lggzngl_gfi_2g;§ggg11§y
Assessment. 45(5), 614-619.

Wing, R. R. and Jeffery, R. W. (1979) Outpatient treatments
of obesity: A comparison of methodology and clinical

reSU1t3- IDS3IDiEiQDQl_IQBID§l_Q§_QQ§§i§¥r 3: 251‘ 179

Zuti, W. B. and Goldberg, L. A. (1976) Comparing di et and
exercise as weight reduction tools. Th P 'a a

ansrts_ued121ne 4 49-53.