Title Page
 Table of Contents
 List of Tables
 List of Figures
 Key to symbols and abbreviatio...
 Introduction and conceptual...
 Theoretical background and review...
 Findings and discussion
 Summary and conclusions
 Biographical sketch

Group Title: impact of life events on psychiatric symptomatology
Title: The impact of life events on psychiatric symptomatology
Full Citation
Permanent Link: http://ufdc.ufl.edu/UF00097489/00001
 Material Information
Title: The impact of life events on psychiatric symptomatology
Physical Description: xvii, 153 leaves : ill. ; 28 cm.
Language: English
Creator: Holzer, Charles Elmer, 1943-
Copyright Date: 1977
Subject: Stress (Psychology)   ( lcsh )
Symptoms   ( lcsh )
Sociology thesis Ph. D   ( lcsh )
Dissertations, Academic -- Sociology -- UF   ( lcsh )
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
Statement of Responsibility: by Charles Elmer Holzer III.
Thesis: Thesis--University of Florida.
Bibliography: Bibliography: leaves 140-152.
Additional Physical Form: Also available on World Wide Web
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00097489
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000185166
oclc - 03329486
notis - AAV1748


This item has the following downloads:

impactoflifeeven00holzrich ( PDF )

Table of Contents
    Title Page
        Page i
        Page ii
        Page iii
        Page iv
        Page v
        Page vi
    Table of Contents
        Page vii
        Page viii
    List of Tables
        Page ix
        Page x
        Page xi
    List of Figures
        Page xii
    Key to symbols and abbreviations
        Page xiii
        Page xiv
        Page xv
        Page xvi
        Page xvii
    Introduction and conceptual framework
        Page 1
        Page 2
        Page 3
        Page 4
        Page 5
    Theoretical background and review of literature
        Page 6
        Page 7
        Page 8
        Page 9
        Page 10
        Page 11
        Page 12
        Page 13
        Page 14
        Page 15
        Page 16
        Page 17
        Page 18
        Page 19
        Page 20
        Page 21
        Page 22
        Page 23
        Page 24
        Page 25
        Page 26
        Page 27
        Page 28
        Page 29
        Page 30
        Page 31
        Page 32
        Page 33
        Page 34
        Page 35
        Page 36
        Page 37
        Page 38
        Page 39
        Page 40
        Page 41
        Page 42
        Page 43
        Page 44
        Page 45
        Page 46
        Page 47
        Page 48
        Page 49
        Page 50
        Page 51
        Page 52
        Page 53
        Page 54
        Page 55
        Page 56
        Page 57
        Page 58
        Page 59
        Page 60
        Page 61
        Page 62
        Page 63
        Page 64
        Page 65
        Page 66
        Page 67
        Page 68
        Page 69
        Page 70
        Page 71
        Page 72
    Findings and discussion
        Page 73
        Page 74
        Page 75
        Page 76
        Page 77
        Page 78
        Page 79
        Page 80
        Page 81
        Page 82
        Page 83
        Page 84
        Page 85
        Page 86
        Page 87
        Page 88
        Page 89
        Page 90
        Page 91
        Page 92
        Page 93
        Page 94
        Page 95
        Page 96
        Page 97
        Page 98
        Page 99
        Page 100
        Page 101
        Page 102
        Page 103
        Page 104
        Page 105
        Page 106
        Page 107
        Page 108
        Page 109
        Page 110
        Page 111
        Page 112
        Page 113
        Page 114
        Page 115
        Page 116
        Page 117
        Page 118
        Page 119
        Page 120
        Page 121
        Page 122
        Page 123
        Page 124
        Page 125
        Page 126
        Page 127
        Page 128
        Page 129
    Summary and conclusions
        Page 130
        Page 131
        Page 132
        Page 133
        Page 134
        Page 135
        Page 136
        Page 137
        Page 138
        Page 139
        Page 140
        Page 141
        Page 142
        Page 143
        Page 144
        Page 145
        Page 146
        Page 147
        Page 148
        Page 149
        Page 150
        Page 151
        Page 152
    Biographical sketch
        Page 153
        Page 154
        Page 155
Full Text







Copyright by
Charles Elmer Holzer III


Jean, Robin
Christopher Paul


I would like to thank Dr. George J. Warheit, chairman

of my doctoral committee, friend and colleague, for his

intellectual and emotional support throughout the creation

of this dissertation. His guidance has brought focus and

closure to an otherwise open ended process. I would also

like to thank the remainder of my committee, Dr Gerald

Leslie, Dr. Benjamin Gorman, Dr. Anthony LaGreca, and

Dr. Alan Agresti for counsel and support throughout

this endeavor.

This research is based on data collected under the

auspices of National Institute of Mental Health Grant

#15900 with John J. Schwab, M.D., as Principal Investigator.

This work has been continued while working on NIMH Grant

MH 24740 with Dr. George J. Warheit as Principal

Investigator. Without the direct and indirect support

of these grants, the present research would have been

impossible. Sue Legg, in particular, deserves mention

for her efforts in coordinating the follow-up interviewing,

but the staff members of these respective projects have

all made material contributions to this effort.

The analyses used in this presentation were performed

using the facilities of the Northeast Regional Data Center

and the Center for Instructional and Research Computing

Activities. Computing funds were provided by the Colleges

of Medicine and Arts and Sciences. Most of the computer

runs were made by Jean B. Holzer, an effort which facili-

tated my concentration on the writing process.

During the process of writing this report, and the

countless drafts which preceded it, Paul M. Cohen acted

as a sounding board for most of the ideas presented and

as an editorial consultant. Editorial assistance was

also provided by Lynn Robbins, Joye Barnes and Jean Holzer.

Throughout this drafting process the person who, more than

any other, kept the project alive was Linda Johnston. She

cheerfully survived the many waves of text which preceded

this last revision. The final typing of the manuscript

was performed by Sue Kirkpatrick and Nancy McDavid. They

labored well under excessive time pressure and should bear

no responsibility for the errors which I have managed to

slip past them.

I also want to acknowledge the support of a more

general nature, that of concern and enthusiastic encourage-

ment from many friends and colleagues. A few, Drs. John

Adams, John Kuldau, Meyer Maskin, Jim Fuson, and Ralph

Selfridge, stand out above the rest, but there are many more

than can be mentioned.

Finally, there is no one who suffered more through

the protracted effort of producing this dissertation

than my wife, Jean, and children, Robin Ann and Christopher

Paul. They quietly tolerated my absence from normal family

life and were still able to provide support when I needed




ACKNOWLEDGMENTS .. . . . .. .... .. . iv

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

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


ABSTRACT. . . . . . .. . . . . xiv


The Method of the Study .. . . 3
Plan of the Report . . . . . . 4

LITERATURE . . . . . . . . . 6
A Model of Stress. . . .. . . 6
The Definition and Quantification of Life
Events . . .. . . . . 15
Socioeconomic Status As Resources. . . 24
Psychiatric Symptomatology As Maladaptive
Response . . . . . . . 25
The Development of Working Hypotheses. .. 27

3 METHODS. . . .. . . . . . . 29
Research Design, an Overview . .. . 29
Sampling Procedures. . .. . . . 30
Construction of Life Event Indices . .. 36'
Measurement of Psychiatric Symptomatology. 46
Measurement of Socioeconomic Status. . 56
Plan of Analysis . . . . . ... 58
The Specific Analyses: Part 1 . . .. 59
The Specific Analyses: Part 2 . . .. 69

Distribution of Variables. ...... . 73
The Form and Strength of the Relationship. 76
Examination of Alternative Life Event
Measures. .. . . . . .. . 97
Discussion . . . .. . . . 118




Introduction . . . . . . .. .130
Summary of Findings. .. . . . . 130
Conclusions .. . . .. . . 133



. . . . . . . . 1 4 0

. . . . . . . 1 5 3



Table Page

2.1 A Summary of Selected Classifications of
Life Events Appearing in the Literature 17

3.1 Demographic Characteristics of the Original
and Follow-up Samples . . ..... .. 33

3.2 Paykel's Items; Their Weights and Frequency
of Occurrence Within The Follow-up Sample 38

3.3 Classification of Events as Entrances to and
Exits from the Social Field . . . .. 41

3.4 Classification of Events According to Judg-
ments of Their Dependence on Prior
Symptomatology . . . . . .. 43

3.5 Classification of Events by Area of Social
Activity . . . . . . . .. 47

4.1 Means and Standard Deviations for Major
Variables from 1970 and 1973 Surveys. . 74

4.2 Means and Standard Deviations for Life
Event Measures . . . . . . . 75

4.3 Transition of Respondents Between Case Cate-
gories by Paykel Scores, Controlling for
SES Level . . . . . . . . 77

4.4 Change in HOS Scores by Paykel Scores,
Controlling for 1970 HOS Scores . . .. 80

4.5 Regression Analysis of 1973 HOS Scores as
Change from 1970 HOS Scores, Using Paykel
Scores and SES Level as Predictors . 82

4.6 Regression Analysis of Paykel Scores Using
HOS-1970 Scores as a Predictor. . . .. 90

LIST OF TABLES (continued)

Table Page

4.7 Cross-sectional Regression Analysis of
HOS-1973 Scores, Using Weighted Paykel
Scores as a Predictor ... . . . 94

4.8 Regression Analysis of HOS-1973 Scores as
Change from HOS-1970 Scores, Using Unweighted
Paykel Scores as a Predictor. . . ... 98

4.9 Regression Analysis of Unweighted Paykel
Scores, Using HOS-1970 Scores as a Pre-
dictor. ... . . . . . . 100

4.10 Stepwise Regression Analysis of HOS-1973
Scores as Change from HOS-1970 Scores Using
Individual Events as Predictors . . .. 102

4.11 Regression Analysis of HOS-1973 Scores as
Change from HOS-1970 Scores Using Paykel
Entrances and Exits as Predictors ... 104

4.12 Summary of Regression Analyses Predicting
Life Event Scores for Entrances and Exits
from HOS-1970 Scores and Sociodemographic
Variables . . . . . . . ... 105

4.13 Regression Analysis of HOS-1973 Scores as
Change from HOS-1970 Scores Using Paykel
Symptom Relatedness Scores as Predictors. 107

4.14 Summary of Regression Analyses Predicting
Life Event Scores Classified by Rated
Symptom Dependence from HOS-1970 Scores
and Sociodemographic Variables .. .. .. 109

4.15 Regression Analysis of HOS-1973 Scores as
Change from HOS-1970 Scores Using Paykel
Scores from Life Events Classified by
Area of Social Activity as Predictors . ll

4.16 Summary of Regression Analyses Predicting
Life Event Scores for Areas of Social
Activity from HOS-1970 Scores and Socio-
demographic Variables . . .. . . 114

LIST OF TABLES (continued)

Table Page

4.17 Regression Analysis of HOS-1973 Scores as
Change from HOS-1970 Scores Using Paykel
Scores by Year of Occurrence as Predictors. 116

4.18 Summary of Regression Analyses Predicting
Life Event Scores by Year from HOS-1970
Scores and Sociodemographic Variables . . 117


Figure Page

2.1 Diagram of the Process of Adaptation. .. 13

3.1 Cumulative Percentage Distributions of HOS
Scores for Community and Patient Samples. 53

4.1 Estimated Change in HOS Scores Between 1970
and 1973 by Paykel Score and Level of 1973
SES . . . . . . . . . . 85

4.2 Partitions of Explanatory Variance for HOS-
1973 Scores as an Increment from HOS-1970
Scores . . . . . . . .. 87

4.3 Partition of Explanatory Variance for HOS-
1973 Scores from Cross-sectional Analysis 96



B = Regression coefficient

c = Error term in a regression equation

X2= Chi square statistic from test of independence

R2= Coefficient of determination

r = Pearson correlation coefficient


Abstract of Dissertation Presented to the Graduate Council
of the University of Florida in Partial Fulfillment
of the Requirements for the Degree of Doctor of Philosophy



Charles Elmer Holzer III

March, 1977

Chairman: Dr. George J. Warheit
Major Department: Sociology

This research analyzes the relationships between life

events, socioeconomic status and psychiatric symptomatology.

A review of the stress literature identifies a series of

three general stress relationships which, when applied to

the study of life events, produce three working hypotheses:

Hypothesis 1: For a specified level of socioeconomic
status, the level of psychiatric symptoma-
tology varies directly with the level of
life events.

Hypothesis 2: The strength of the association between
level of life events and level of psychia-
tric symptomatology varies inversely with
the level of socioeconomic status.

Hypothesis 3: For a specified level of socioeconomic
status, the level of psychiatric symptoma-
tology varies directly with the subsequent
level of life events.

These hypotheses are tested using a panel design based on

an original epidemiological survey (N=1645) conducted in

1970 and a follow-up survey (N=517) conducted in 1973.

This design permits examination of change in symptom scores

over the three-year period as a result of intervening life

events, controlling for socioeconomic status.


The operational definition of psychiatric symptoma-

tology is the Health Opinion Survey (HOS) index, a

psychiatric community screening device developed by

Macmillan and modified by Leighton. It has been shown

to be an adequate measure for distinguishing psycho-

neurotic patients from community controls. The primary

measure of life events, based on the work of Paykel,

consists of a list of 61 items which might have occurred

during the period between interviews. Paykel had pre-

viously developed ratings showing the expected amount of

psychological upset typically produced by each event.

Scores on the Paykel measure consist of the sum of

weights for the events which occurred.

The third major variable is an index of socioeconomic

status (SES) based on methodology developed by the U.S.

Bureau of the Census. It is an average of the respondent's

rankings on education, occupation and income.

Additional life event indices are constructed using

categorizations based on entries versus exits, areas of

social activity, the dependence of the event on previous

symptomatology, and the year of occurrence. An unweighted

index of the full set of items, and the separate items

are also considered.

The data are analyzed primarily through the use of

multiple regression models, with equations for prediction

of 1973 HOS scores both cross-sectionally from 1973 data

alone, and as change over time by inclusion of 1970 HOS

scores as a control. A separate regression model is used

to predict the life event scores from the 1970 HOS scores.

The findings from these analyses are as follows:

(1) Only a small proportion (2.1%) of the change in

HOS scores can be attributed to life events.

(2) The association of life event scores with changes

in psychiatric symptomatology over a three-year period is

inversely proportional to the socioeconomic status of the

respondents involved. Life events are more strongly

predictive of changes in symptomatology for low SES

respondents than for high.

(3) Symptom scores for 1970 are predictive of life

events scores for the period following 1970, suggesting

a reciprocal relationship between symptoms and events

over time.

(4) Symptom levels are highly stable over the three-

year period. Symptom levels for 1970 predict more than

60% of the variance in 1973 HOS scores, suggesting that

the major factors determining symptomatology are chronic

rather than transient phenomena.

(5) Examination of the alternative indices of life

events reveals that some are much more predictive of

symptomatology than others, the most predictive being

exits, events with known or possible symptom dependence,

personal health and marital difficulties.


(6) The influence of life events on symptomatology

depends partially on the time elapsed between the occur-

rence of the event and the measurement of the symptoms.

Recent events are strong predictors of symptomatology.

Thus the above study has examined the influence of

life events on psychiatric symptomatology and has

demonstrated the presence of a small but significant

influence which is greater in strength for those with

lower than with higher socioeconomic status.




The primary objective of this research is to analyze

the relationships between life events, stress, and psychiat-

ric symptomatology. The past decade has witnessed a rapid

expansion in the number of studies dealing with this subject.

Despite the volume of these investigations, their theoreti-

cal and methodological limitations have hampered attempts to

clarify the strength and direction of causality between life

events and psychiatric symptomatology. As recently as the

final drafting of this report, a review article has been

severely critical of the current theoretical and methodologi-

cal state of the field (Rabkin and Struening, 1976).

At the theoretical level, the debate regarding the

medical versus social nature of mental illness remains unre-

solved. One consequence has been that both etiology and

definition have been poorly specified (B. P. Dohrenwend,

1975). At the methodological level, cross-sectional and

retrospective designs have not adequately controlled for

factors predisposing individuals either to the occurrence

of life events or to psychiatric symptomatology. Not only

have these theoretical and methodological limitations

seriously undermined attempts to specify the strength and

direction of causality for the relationship between life

events and psychiatric symptomatology, but they have also con-

tributed to the failure to specify the contextual factors

which shape it.

Three major issues must be addressed in order to spec-

ify clearly the relationship between life events and psy-

chiatric symptomatology. The first of these is the deter-

mination of the influence of life events, both general and

within specific subcategories, on the subsequent develop-

ment of psychiatric symptomatology.

The second issue is whether there exists reciprocal

causality between life events and psychiatric symptomatology

over time. Most studies have assumed that life events act

only as causes of psychiatric symptomatology. The longi-

tudinal design adopted here permits us to explore the ex-

tent to which life events themselves may be the result of

psychiatric symptomatology.

The final issue is the role of intervening factors,

such as resource availability, which influence the rela-

tionships between life events and psychiatric symptomatology.

Most often, socioeconomic statuses are seen as major

indicators of resources which an individual can utilize

in meeting the demands which life events place on him.

Implicitly and explicitly, research in the area assumes that

the impact of life events will be greatest for persons in

the lower socioeconomic statuses.

The Method of the Study

This research is designed to deal with each of the above

issues. A panel design has been adopted. This approach

builds on an epidemiological survey originally conducted

in 1970; in 1973 approximately one-third of the original

respondents were reinterviewed. The 517 respondents inter-

viewed at both times provide a panel which is representa-

tive of a broad segment of the community population.

For the purposes of this study, life events have been

operationalized in terms of the degree of upset typically

produced by an event (cf. Paykel et al., 1971); psychiatric

symptomatology has been operationalized as Health Opinion

Survey scores (cf. Macmillan, 1957), a measure of psycho-

neuroticism; and socioeconomic status has been operational-

ized as an index combining education, occupation, and

income (U.S. Bureau of the Census, 1960).

Analysis of change within this panel design provides

a means of observing the relationship between life events

and change in symptom levels over time. It also makes pos-

sible observation of the influence of symptomatology at a

particular point in time on the subsequent occurrence of

life events. Finally, the introduction of socioeconomic

status as a factor interacting with life events in its

influence on symptomatology helps to establish the impor-

tance of intervening factors in the influence of life


Only three prior studies have utilized panel designs

to study the impact of life events on psychiatric symptoma-

tology, and each of these has been seriously limited either

in the scope of the data on which analysis could be con-

ducted,or in the form of analysis which was applied to the

available data.

The present study has advantages over these previous

ones in terms of the size of the sample utilized and in the

application of multivariate statistical procedures which

facilitate a formal modeling of the interactions among the

life events, psychiatric symptoms, socioeconomic status, and

other control variables.

Plan of the Report

In Chapter 2 a review of several models of stress and

adaptation is presented. From this review three general

statements relating demand, resources, and maladaptive

response are identified; these provide guidelines for this

research. In the next section of Chapter 2, drawing from

the works of Paykel, life events are identified as sources

of demands, and an approach to quantify that demand is

presented. Then the definition of resources used in this

research is presented. Finally, a measure is identified

for psychiatric symptomatology, the central part of social-

psychological maladaptive response.

In Chapter 3 the methodology of the present study is

described. This description includes the design, sampling

procedures, interviewing, index construction, and finally

the plan of analysis.

Chapter 4 presents tests of hypotheses, which are

based primarily on regression analyses of the change in

levels of symptomatology over time. It also examines the

dependence of life events on prior levels of symptomatology.

The next section of Chapter 4 examines the cross-sectional

formsof the hypotheses and thus provides a point of com-

parison with most other life events studies. The remainder

of the chapter consists of reexamination of the hypotheses

based on alternate forms of the life event index and thus

addresses basic issues in the aggregate measurement of life


Chapter 5 provides a summary of the results and offers

conclusions regarding the relationships between life events,

socioeconomic status, and psychiatric symptomatology.

Finally, it discusses implications for future research.



Although the present study focuses on life events and

psychiatric symptomatology, the basis of the research is

found in stress theory. This chapter begins with a general

model of stress and identifies the three general findings

of stress research. It then traces efforts at quantifica-

tion of stress through the development of life event re-

search. After providing means for quantification of the

demand produced by life events, measures for socioeconomic

status, as resources, and for psychiatric symptomatology as

maladaptive response are provided. Finally a set of three

working hypotheses relating life events, psychiatric symp-

tomatology, and socioeconomic status is developed from the

general statement derived from review of the stress litera-


A Model of Stress

Origin of the Concept

The term stress traces its origins to engineering usage,

in which stress refers to a set of forces directed at a

physical material with the result that changes take place

in the material prior to its collapse. These changes are

called strain. The terms stress and strain have been adopted

as a metaphor describing the effects of an adverse environ-

ment upon a wide variety of phenomena.

Physiological Approaches to Stress

One of the milestones in stress research was Cannon's

(1939) original experimental study of the process of homeo-

stasis, the set of processes by which humans and other liv-

ing organisms maintain a dynamic balance in the face of

adverse environmental situations. This original work was

extended and codified by Selye (Selye and Fortier, 1950;

Selye, 1955, 1956, 1974) into the model of stress which has

been the point of departure for much of present stress re-

search. From experimental studies Selye discovered that the

process of adaptation to environmental disturbances involves

two and,sometimes,three phases. The first of these, an

alarm reaction, produces a consistent but general pattern of

physiological response termed the General Adaptation

Syndrome (GAS). The second phase is a progression to more

specific local adaptation syndromes in which specific

responses are directed at particular environmentally pro-

duced disturbances. The third phase is one in which spe-

cific responses are exhausted,once again leaving a

general response. In the case of an extreme and persistent

disturbance,this is maintained until the final collapse of

the organism. Essential to Selye's formulation is the

limited capability of the organism to maintain a response

without exhaustion or damage to the organism. This damage,

and the reaction promoting it, provide Selye's definition of

stress. Damage to the organism has been the basis for

widespread application of Selye's model to chronic health


The major criticisms of Selye's model for the present

purposes are its physiological focus, its failure to recog-

nize that, although the GAS is a patterned response, it may

vary among individuals (Appley and Trumbull, 1967), and the

confusion produced by Selye's application of the term stress

to that which engineers term strain.

The works of Cannon and Selye were preludes to wide-

spread exploration of environmental factors which produce

physiological disturbances. Much of this research has

remained physiological or focused on physical health, but

some has been extended to include social and psychological

aspects of stress.

Theories of Social-Psychological Stress

The generalization of the concept of stress into social

and psychological areas makes available a wide variety of

topics for study. These include social and psychological

situations which produce stress, cognitive and other sub-

jective processes involved in the individual's perception

of one situation as stressful but not another, psychological

defenses used to control the subjective reaction to a stres-

sor, cognitive processes used to solve problems, the use

of social structures as support systems,and finally, social

and psychological consequences of stress. Each of these

issues has been the object of considerable study,and com-

plete review is not possible. Instead, the focus here is on

those aspects of stress theory most directly applicable to

the present research.

McGrath: A Metatheory of Stress

McGrath (1970) presents a general paradigm for the

formulation of theoretical models of stress. His work is

used as a point of departure because it resembles a general

systems approach and thus provides a neutral terminology

for comparison of other approaches.

McGrath's formulation begins with an organism embedded

in an external physical and social environment. This

organism is exposed to objective demands which, through

evaluative processes, become subjective demands. The re-

sponses to subjective demands are based on the availability

of resources. These responses have consequences for the

organism and for its environment, thus influencing future

demands, as well as resources for meeting these demands.

Adaptation takes place over time, with feedback loops

reflecting the flow of events. Thus, adaptation, and the

stress accompanying it, constitute processes rather than

static conditions. McGrath points out that definitions of

stress focusing exclusively upon demand, whether subjective

or objective, or definitions focusing exclusively on re-

sponse to situations, fail to account for the active balance

between demands and resources. The balance between demands

and resources determines the formation of stress.

McGrath's formulation is eclectic and thus tends to

deal with generalities. For example, he fails to point out

that demands and resources must balance not only at the ag-

gregate levels but also within the major subsystems which

constitute the organism. Although limited, McGrath's model

provides two basic tenents of the present research: first,

that an excess of demands over the resources will produce

stress response, and, second, that response in one adapta-

tion cycle may affect subsequent levels of demands.

Dohrenwend: A Generalization of Selye's Model

B.P. Dohrenwend (1961), B.P. and B.S. Dohrenwend (1969),

and B.S. and B.P. Dohrenwend (1970) have extended Selye's

formulation into a model of social-psychological stress which

is more specific than McGrath's. They identify four elements

of stress: the antecedent stressor, conditioning or medi-

ating factors, the general adaptation syndrome, and the ul-

timate responses of the organism.

Stressors and mediating factors are generally equiva-

lent to demands and resources but receive fuller development.

B.P. and B.S. Dohrenwend (1969) developed an outline of life

events which have been identified as stressors. This four-

fold typology of stressors distinguishes achievement from

security and developmental from nondevelopmental stressors.

While this typology encompasses situations which occur in

the lives of nearly everyone, these situations are considered

by the Dohrenwends particularly in relationship to social

class and ethnicity. A second point of elaboration in the

Dohrenwend model is the distinction between internal and ex-

ternal mediating factors. Internal mediating factors include

"inner drives, desires, and internalized rules or prescrip-

tions" (B.P. Dohrenwend, 1961:296), as well as other personal

attributes which facilitate dealing with a stressor. Ex-

ternal mediating factors, on the other hand, include general

and specific attributes of the individual's environment,

such as systems of social and economic support. Thus,

stressors and mediating factors correspond to demands and


In the Dohrenwend formulation, stressors altered by the

mediating factors produce an "adaptation syndrome, indicating

an intervening state of stress in the organism" (B.S. and

B.P. Dohrenwend, 1970:114). This, in turn, may lead to adap-

tive responses, but also may produce maladaptive responses

related to the physiological processes observed by Selye.

The Dohrenwends state that "the greater the intensity and

duration of stress, the greater its severity and the

likelihood of 'derailment' of the mechanisms of adaptation"

(B.S. and B.P. Dohrenwend, 1970:115).

B.P. Dohrenwend (1961) also elaborates on the form of

the adaptation syndrome and its maladaptive responses. The

adaptation syndrome involves affective, conative, and cog-

nititve facets which Dohrenwend illustrates in terms of com-

bat neurosis. Exact specification of maladaptive response,

however, is difficult due to the lack of an absolute standard

against which it can be measured. These social and psycho-

logical aspects of response to stress, as with evaluations

of mental health, are relative to a particular social context,

and thus Dohrenwend ultimately bases their evaluation on

socially determined standards of behavior (cf. B.P. Dohren-

wend, 1961:301). Nonetheless, the Dohrenwends' view of mal-

adaptive response is closely related to the occurrence of

what would conventionally be termed psychiatric symptomatol-

ogy, which will be the definition of maladaptive response

used throughout this study.

For the purposes of the present research we will adopt

the Dohrenwend model with only a few qualifications. These

center on the model's failure to emphasize the dynamic

nature of adaptation as a process taking place over time

and involving interrelated elements both internal and

external to the individual. The adaptive process is dia-

grammed in Figure 1.

decrease in demands

adaptive process






I increase in demands

Figure 2.1 Diagram of the Process of Adaptation

Although the above review has emphasized dynamic

aspects of the process of adaptation, the present study

must approach the practical issue of applying the findings

of the review of stress models to the research at hand.

Because a single study cannot assess all the different

aspects of a stress model, only three basic relationships

are identified. These will be used to guide the remainder

of the present research.

1. For a specified level of resources, the probability
of maladaptive response varies directly with the
level of demand.

2. The strength of the association between levels of
demand and maladaptive response varies inversely
with the level of resources for meeting that demand.

3. For a specified level of resources, the level of
maladaptive response varies directly with the sub-
sequent demand.

Each of these statements has been based on the re-

viewed stress models. Significantly, however, stress, the

intervening variable of stress models, is not included as

an explicit term, Although stress is a relevant concept,

it is omitted because it is not directly observable within

the present research context. In a similar fashion, the

statements do not specify explicitly whether demand and

resources refer to subjective or objective variables. It

is believed that objective and subjective appraisals will

be highly correlated, on the average, thus making the propo-

sitions applicable to either subjective or objective

definition. Because this study makes use of self-report in

measuring these concepts, a dimension of subjectivity is

introduced, even though objective measurement is attempted.

The Definition and Quantification of Life Events

Efforts to quantify the variables in life event research

have their origins in the life chart device of Adolph Meyer

and its further development in Harold G. Wolff's laboratory

at Cornell (Holmes and Masuda, 1974). Their research iden-

tified a series of life events which were usually stressful,

evoking psychophysiological reactions and often contributing

to the onset of disease. The direct successors to this work

have been Hinkle (1974), who has concentrated on longitudi-

nal studies, and Holmes and Rahe (1967), who have concen-

trated on quantitative methodologies for scaling life events.

These, in turn, have propagated a vast quantity of life

event studies by other researchers, some of which will be

discussed below.

Our immediate concern is the definition of life events

as a concept which can guide our theoretical considerations.

Unfortunately, however, the term life event has been left

ill-defined by virtually all who do life event research. The

reason for this difficulty has been the inherently diverse

nature of the phenomena which individuals experience as

salient enough to identify as an event. This is compounded

by the efforts of researchers and clinicians alike to

identify any potentially causative precursor to an observed

instance of disease or disability.

In dealing with the diversity of life events, two

nonexclusive approaches have been utilized. These are

classification and scaling. The former has focused largely

on the closely related dimensions of gain versus loss

(B. S. Dohrenwend, 1973a), social entrance versus social

exit (Paykel and Uhlenhuth, 1972), and desirable versus

undesirable (B. S. Dohrenwend, 1973a). Other dimensions

have also been explored, such as the amount of control the

respondent may have over the occurrence of the event, the

confounding of the event with previous mental or physical

illness (B. P. Dohrenwend, 1974), or the area of social

activity involved (Paykel et al., 1969). These are item-

ized in Table 2.1.

The efforts at scaling have particularly been associ-

ated with Holmes and Rahe (1967) and Paykel et al. (1971),

although B. S. Dohrenwend (1973a) has explored a major

modification to the Holmes and Rahe measure by incorporat-

ing desirability. Because the present research will pri-

marily utilize a scaling approach to life events (e.g.,

Paykel et al., 1971), we will explore these approaches in

greater detail.

Table 2.1. A Summary of Selected Classifications of Life Events
Appearing in the Literature

Dimension of Research



Factors Controlling
the Onset of the

Characteristics of
the Event and Its

Interval Measure-

Control vs. No Control

Confounding of Events
with Previous Mental
and Physical Health

Objective vs. Subjec-
tive Occurrence

Gain vs. Loss

Desirable vs. Unde-

Area of Activity
Involved in the

Entrances vs. Exits

Social Readjustment

Social Readjustment
Adjusted for Desir-
bility of Event

Amount of Upset
expected for the
Average Person

Brown et al., 1973b
B. S. Dohrenwend, 1973b

B. P. Dohrenwend, 1974
Cooper & Sylph, 1973

Thurlow, 1971

B. S. Dohrenwend, 1973a

Paykel et al., 1969
Myers et al., 1972
B. S. Dohrenwend, 1973a
Vinokur & Selzer, 1975

Paykel et al., 1969

Paykel et al., 1975

Holmes & Rahe, 1967

B. S. Dohrenwend, 1973a

Paykel et al., 1971

Social readjustment. Holmes and Rahe (1967) attempted

to estimate the magnitude of individual life events by

estimating the amount of change from the ongoing life pat-

tern of an individual required by each particular event.

They termed this change social readjustment. Holmes and

Rahe sought a measure which would make possible the cumula-

tion of change from individual events into a total measure

of social readjustment.

Their initial rating of events combined 43 events from

previous research into a Social Readjustment Rating Ques-

tionnaire (SRRQ) which was administered to a large number of

respondents. Instructions included a definition of social

readjustment and asked for ratings of the average amount of

readjustment for each event relative to an anchor event,

marriage, which was assigned an arbitrary score of 500. An

event perceived as requiring social adjustment half that of

marriage would be scored 250 while an event perceived to re-

quire twice the readjustment would score 1,000. In estimat-

ing the readjustment required by an event, the raters were

asked to consider direct and vicarious experiences. After

conducting a number of studies which examined variations

among individuals grouped sociodemographically, culturally,

and cross-nationally, Holmes and Rahe (1967) claimed

. a universal agreement between groups and among indi-

viduals about the significance of the life events under

study that transcends differences in age, sex, marital

status, education, social class, generation American,

religion and race" (p. 217).

On the basis of these findings, Holmes and Rahe (1967)

constructed an instrument for studying the impact of life

events upon individuals. This instrument, the schedule of

recent life events (SRE), has been used in studies examin-

ing the relationship between life events and a variety of

illness-related phenomena. Many of them are reviewed by

Holmes and Masuda (1974) and Rahe (1974).

Although the scale provided by Holmes and Rahe has

been used widely and successfully in the prediction of

illness and illness behavior, certain limitations of the mea-

sure have been recognized. Rahe (1974) reports continuing

efforts to improve the measure by changing its format,

modifying question content, and experimenting with cluster

scoring techniques.

Cochrane and Robertson (1973) provided a broader

criticism by noting three difficulties: (1) some items are

inappropriate, particularly those dealing with trivial

experiences, (2) some important topics are not included,

and (3) norms were not available, particularly within pa-

tient groups where much life event research was being con-

ducted. Other criticisms noted by Rahe (1974) have

focused on the assumptions about ratio scaling, which

comprise an essential feature of their approach. Because

of these limitations a number of researchers have reapplied

the original approach with modifications in both method and


A first alternative to Rahe and Holmes' Social Readjust-

ment is the use of an unweighted index based on the same

items. Rahe (1974) has stated that weighting of items may

provide no improvement in the prediction of health problems

in those situations having uniformly minor events. The

weighting becomes more important when some respondents have

experienced larger numbers of the extreme events.

A second modification to the social readjustment rating

scale was made by B. S. Dohrenwend (1973a). Myers et al.

(1971) had previously reported differences in the effect of

desirable and undesirable events; therefore, B. S. Dohrenwend

developed an index combining desirability and social read-

justment. This index was based on the original Rahe and

Holmes items, but added the weighted loss events to the

total score, and subtracted the weighted gain events from

the total score. Thus, gains serve to reduce the total

score obtained. This measure has been applied in re-

searches by Dohrenwend and also by Myers et al. (1974).

Paykel: Events As Upsetting. Paykel et al. (1971)

constructed a measure of life events based on the

degree to which an event is upsetting to the average person.

While recognizing the importance of life change in the

production of subjective stress, they pointed out that

Holmes and Rahe were concerned primarily with somatic

illness. Given Paykel's concern with psychiatric responses,

he gave primary consideration to the desirability of events,

a dimension intentionally excluded by Holmes and Rahe


We chose the concept of upset for several rea-
sons. The most important are concerned with its
meaning. We were particularly interested in
applying our scale to the precipitation of
depression and other psychiatric disturbance,
where stress appears closely related to per-:
ceived distress. Certainly the recognition of
life change as an element represents a distinct
advance over the frequent assumption that
desirable change entails negligible stress.
Nonetheless, in psychiatry questions of de-
sirability, value, and symbolic implication
have long been regarded as important in mediat-
ing the link between precipitant and reaction.
Threats rather than occurrences, and events
which involve little actual life change, such
as death of parent for an adult with separate
home or blows to self-esteem, often appear to
precipitate emotional disturbances. Promotion,
a desirable change in work responsibilities,
although not entirely blame-free, appears less
likely to induce disturbances than demotion,
its undesirable and ego-threatening equivalent.
(Paykel et al., 1971, 345-346)

Thus, although the concept of life change or social re-

adjustment is incorporated into Paykel's formulation, so

too are subjective notions of threat and psychological

process. The subjective mechanisms causing upset do not,

however, enter into either the definition of the resulting

state, being upset, nor are they part of the objective

event. Consequently, Paykel's conceptualization of sub-

jective mechanisms resulting in upset need not deter

empirical study of measurable relationships.

Paykel estimated the degree of upset associated with an

event by use of a panel of raters (N=373) who were asked to

rate 61 events on a scale from Oto 20 based on how upsetting

the event would be to the average person. They were given

the following directions:

Below is a list of events that often happen to
people. We would like you to think about each
event and decide how upsetting it is. Use your
own experience and what you know about other
people to make your decision. A particular
event might be more upsetting to some people
than to others. Try to think how upsetting
the event would be to the average person.
(Paykel et al., 1971:340)

Analysis of the ratings produced means and standard

deviations for each of the items. Comparisons were made

across sociodemographic groups and on methodological vari-

ables. In all instances the correlations exceeded r = .96.

A comparison of the upset ratings with an approximation of

social readjustment ratings provided a correlation of

r = .484, and a comparable correlation for 14 items having

identical wording with the original Holmes and Rahe analysis

yielded a correlation of r = .683. Finally, an examination

of the standard deviations suggested that, while application

of the ratings to individual cases is inappropriate, their

use for group comparisons is justified.

Evaluation of Life Event Measures

Assessments of life events made by Holmes and Rahe

(1967) and Paykel et al. (1971) relate to stress in different

ways. In both formulations change is an initial demand made

upon the individual. Holmes and Rahe ask how much change

an event produces while Paykel asks about the upset pro-


As a result, Paykel's approach can readily be inter-

preted as a measure of net demand, that is, the demand re-

maining after typical resources have been taken into account.

This is consistent with Paykel's definition of the measure

as the typical amount of upset resulting for the average

person. Adoption of Paykel's approach, therefore, calls for

the simultaneous adoption of a general measure of resources;

in this research socioeconomic status is used as such a

measure. Finally, in supporting Paykel's approach over that

of Holmes and Rahe, it must be emphasized that there is a

tendency for recurrent situations within society to become

socially defined. These social definitions lead to role

prescriptions and proscriptions. They further shape societal

reactions to the occurrence of the event and to the responses

of the individual in adapting to it. Thus, the socially

determined demands placed on the individual may go consider-

ably beyond the direct social readjustment experienced by

the individual.

A final issue in the quantification of life events is

the implication of forming an index of life events based

on the scale values obtained from the methods above. The

impact of an event may be modified by the occurrence of

multiple events. The ratings obtained above pertain to

single events occurring in isolation. But what happens if

a second event follows the first? Does the joint effect

equal the total of the two ratings or some other combination

thereof? Does the greater event govern or does the more

recent? These questions have not been fully answered. In

the formulation of the present research the effects of events

are assumed to be additive. The remaining alternatives must

await further research.

Socioeconomic Status As Resources

Prior to development of working hypotheses from the

stress propositions, the use of socioeconomic status as a

measure of the resources available for meeting the demands

posed by life events must be justified. In the three rela-

tionships summarized from the review of the stress literature,

resources were seen as diverse physiological, psychological,

and social elements which relate to specific types and

levels of demand. Nonetheless, two empirical factors come

together to justify socioeconomic status as an approximate

measure of aggregate resources. The first of these is the

possibility of conversion or substitution of one resource

for another. In physiological systems compensatory adap-

tations are possible. At the psychological level,

coping and defense are such mechanisms of substitu-

tion. Finally, social resources, particularly economic

ones, are often convertible to other social resources or

to support which augments psychological or physiological


The second factor is an empirical finding of associa-

tion between certain attributes and a variety of specific

resources. Socioeconomic status, composed of income, educa-

tion, and occupation,is one such point of convergence.

Income, education, and occupation represent resources in a

very direct sense, but each is also correlated with a number

of other specific resources. These include health, self-

confidence, and the availability of friends with resources,

among others.

As a point of caution it must be noted that the term

socioeconomic status is not equivalent to the standard

theoretical terms used to define formal aspects of strati-

fication, for example,prestige or class. As used here,

socioeconomic status refers to the general approach for

combining education, occupation, and income developed by

the U.S. Bureau of the Census (1960).

Psychiatric Symptomatology As Maladaptive Response

The dependent variable for the present research is

psychiatric symptomatology and, in particular, that opera-

tionalized by the Health Opinion Survey, a symptom index

for community screening of psychoneurotic disorders. The

rationale for this selection of dependent variable draws

from two sources: Selye's General Adaptation Syndrome (GAS)

and B. P. Dohrenwend's model of stress.

Selye (1956) reported that the GAS constituted a stable

pattern of physiological response to stress. He based his

observations primarily on physiological measurement, but

many of the elements of the GAS are also observed in the

reports of patients, particularly in those with psychoneu-

rotic anxiety. Similar symptoms are also reported by those

with psychoneurotic depressive disorders. Thus, there is

a partial concurrence between the GAS and the symptoms of

psychoneurotic disorder.

The second argument favoring an examination of psy-

chiatric symptomatology as stress response is found in B. P.

Dohrenwend's (1961) formulation of maladaptive response.

His qualifier "maladaptive" is not an absolute term but

instead rests heavily upon social definitions of acceptable

response. In the social and psychological areas, such defi-

nitions are strongly related to social definitions of mental

illness, particularly those institutionalized within the

medical profession. Thus it is appropriate to operation-

alize maladaptive response in terms of the HOS, a measure

which is designed to identify those in a community who

would be judged by a psychiatrist to be psychiatric cases.

The Development of Working Hypotheses

Because working definitions have been developed for

each of the terms, it is possible to recast the three

relationships drawn from the stress literature as working

hypotheses relating life events, psychiatric symptomatology,

and socioeconomic status. Paykel's formulation of life

events provides a specific instance of demand placed on an

individual. Dohrenwend's model of stress and Paykel's

criterion of upset provide sufficient justification for use

of psychiatric symptomatology as the specific social-

psychological manifestation of maladaptive response. The

final element needed to complete the development of the

working hypotheses from the stress proposition is the use

of socioeconomic status as a measure of the general level

of resources available for meeting the demands imposed by

life events.

These working hypotheses are assumed to be testable

with an implicit alternative hypothesis for each stating

that the relationship was not found to exist. Also implicit

is the traditional disclaimer, the assumption that other

factors are equal or are appropriately controlled. Finally,

the hypotheses are intended to bear interpretation both

as variation among individuals and as change within in-


The three working hypotheses are:

Hypothesis 1:

Hypothesis 2:

Hypothesis 3:

For a specified level of socioeconomic status,
the level of psychiatric symptomatology
varies directly with the level of life events.

The strength of the association between level
of life events and level .of psychiatric symp-
tomatology varies inversely with the level
of socioeconomic status.

For a specified level of socioeconomic status,
the level of psychiatric symptomatology varies
directly with the subsequent level of life

Having developed the theoretical framework and set of

hypotheses, the remainder of the study is devoted to provid-

ing operational definitions for the terms involved and then

testing these hypotheses using both general and'specific

measures of life events.



Research Design, an Overview

The data used in this research were drawn from two

community field surveys. The first of these was an epidemi-

ological field survey of health and mental health, and the

second consisted of reinterviews with 517 of the original

respondents three years after the initial survey. Because

the second survey reinterviewed respondents from the origi-

nal study, it is possible to examine change over time within

individuals. This is one of the defining characteristics

of panel designs, and a necessity for adequate test of

the working hypotheses.

The tests of the hypotheses consist of comparisons of

symptom levels at first and second interview with levels

of life events occurring between interviews and with socio-

economic status. These comparisons are based primarily on

regression models of the effects hypothesized.

The operational variables used in these tests of

hypotheses a re defined in this chapter. They consist

of (1) the Health Opinion Survey, developed originally by

Macmillan (1957) and modified by Leighton (1965); (2) an

index of life events based on ratings of psychological upset

due to the event (Paykel et al., 1971);(3) an index of socio-

economic status developed by the U.S. Bureau of the Census

(1960); and (4) relevant sociodemographic control variables.

In the last portion of the analysis, several alternative

indices of life events are used to examine the specificity

of the hypotheses to the primary index of life events.

Sampling Procedures

Procedures Used in 1970 Sampling

The original survey was conducted in 1970 as a general

epidemiological study of physical and mental health in a

county in the Southeastern United States. Because of its

epidemiologic purpose, it was designed to be maximally

representative of the adult, noninstitutionalized population

of the community. A systematic two-stage probability pro-

cedure was used.

In the first stage, a master sample of 2,315 households

was selected out of the total of 30,421. This was accom-

plished by drawing every thirteenth residential electrical

hookup. In some areas of the county it was necessary to

supplement this list with addresses drawn through area

sampling of county maps. A portion of this list was used

to draw a preliminary sample for development of instrumen-

tation. The remainder of the list was used to obtain 1,645


The second stage of the procedure, selection of one

respondent from each household, was based on a technique

reported by Kish (1965). Total non-response was 16.1% with

a refusal rate of 8.1%. Non-response also included 1.4% of

the total who were unable to complete the interview once

begun. These are not included in the refusal rate. The

remainder of non-response included not-at-home and invalid


The Kish technique selects a single individual from a

household regardless of size. This may cause a bias by

underrepresenting individuals from large households. Kish

recommends weighting the responses for individuals from large

households so as to reduce the bias. Because of the diffi-

culties of conducting analyses using sample weighting, it

has been suggested that such weighting can be eliminated

under conditions where relatively few households in the

sampling frame have exceedingly large household sizes (i.e.,

a typical household size), and where independent variables

are not strongly related to household size (Kish, 1965:400).

Both of these issues were examined. It was determined that

relatively few households (17.2%) had more than two adults

per household. Likewise, only 19.8% of the households con-

sisted of a lone adult. In these instances, the controls

normally provided in reporting obviate the need for weight-

ing because many of the findings are controlled for age and

other factors. Additionally, preliminary analyses were

using both weighted and unweighted respondent data. No

major differences were discovered between weighted and un-

weighted results. Finally, the association between major

dependent variables and household size was examined and

little association detected. Consequently, we followed the

procedure suggested by Kish, and reported results in unweighted

form. The demographic characteristics of the full 1970 sam-

ple are presented in Table 3.1 for comparison with the 1973

sample distributions.

Follow-up Sampling, 1973

A follow-up survey was conducted in 1973 for the pur-

pose of examining change in symptom levels over time. Sam-

pling was based on the 1,645 respondents included in the 1970

survey. Systematic stratified subsampling was used to ob-

tain the follow-up sample. To accomplish this, the 1,645

respondents were ordered on the basis of four variables:

race, sex, age (in five categories), and socioeconomic

status, i.e., SES (in five categories. This sorted list

first presented black males between the ages of 16 and 22

with low SES, followed by increments first in SES, then age,

sex, and finally race. Ten stratified subsamples were

obtained from this ordered list, through assignment on a

rotating basis of the first respondent to the first sub-

sample, the second respondent to the second subsample, etc.

The first, eleventh, twenty-first, and thirty-first, etc.,

Table 3.1

Demographic Characteristics of the Original and Follow-up

Full Sample Follow-up Sample Follow-up Sample
1970 Data 1970 Data 1973 Data
(N=1645) (N=517) (N=517)

Number Percent Number Percent Number Percent


60 +


(total family)
Under $3,000
15,000 & over

Common Law

Grade School
Some HS
HS Grad.*
Some College
College Grad.






























*Includes trade school graduates.







respondents in the ordered list were assigned to the first

subsample. The result was 10 stratified groups with race,

sex, age, and SES characteristics as similar to each other

and the original sample as possible.

A target size for the follow-up sample was set at 500,

based on economics and available manpower. Interviewers were

assigned follow-up respondents from the first sublist, and

then from successive sublists as earlier ones were exhausted.

The interviewers were instructed to locate and interview as

many as possible of the respondents within the first group

before proceeding to the second. Specific interviewer

assignments were made on a rotating basis. As a result, no

interviewer was able to select desirable respondents. This

required attempts to reinterview substantially more respond-

ents than anticipated. By the time the 517 interviews were

obtained, a total of 1,183 interview assignments had been

made. Of these the largest cause for non-interviewing was

inability to locate the respondent due to absence from the

county (513 or 43.36%). Of those contacted, the refusal

rate was 13.4%. Thus, the main reason for non-response was

high mobility among the original respondents between 1970

and 1973. Table 3.1 presents the sociodemographic charac-

teristics of the follow-up respondents as they appeared in

their original interviews in 1970, and also at the time of

follow-up in 1973. Since one o.f the main reasons for loss

of respondents is the high mobility of university students,

faculty, and staff, there were proportionately more blacks,

more females, and fewer college age or college educational

level respondents in the 1973 sample than in the 1970 sample.

Interviewing Procedures

Most interviewing in the 1970 and 1973 surveys was

conducted in respondents' homes, although respondents were

occasionally interviewed at their place of work. Inter-

viewers were typically white married females with college-

level educations and previous interviewing experience.

Interviewers received extensive training in the use of the

interview schedule and on methods of contacting and inter-

viewing respondents. Although interviewer turnover was low

during the course of each study, only a few interviewers

participated in both.

The 1970 interview schedule consisted of 317 items de-

signed to measure interrelationships among social and medi-

cal variables. Included in the instrument were (1) demo-

graphic data and a comprehensive social history; (2) informa-

tion on familial and other interpersonal relationships;

(3) items regarding life satisfactions, both interpersonal

and occupational; (4) indices concerning religion, racial

distance, anomie, perceptions of social change and social

aspirations; (5) a series of questions concerning attitudes

toward, and utilization of, health care services; (6) a

medical systems review and detailed physical symptom data;

and (7) a detailed inventory of mental symptomatology in-

cluding the Health Opinion Survey Instrument (Macmillan,

1957; Leighton et al., 1963; Leighton, 1965), and a series

of other scales developed to measure social-psychiatric

impairment. Typically, this interview took 90 minutes to

complete, although in several instances interviews were

either continued at a later time, or were as many as three

hours long (N=63). Rarely was a respondent unable to

complete the interview.

The 1973 interview retained much of the 1970 schedule

in order to preserve comparability. A number of items used

in 1970 were eliminated because of their apparent lack of

utility in analyses conducted during the interim period.

Few changes, however, were made in items being used in the

present study. One addition to the 1973 instrument which is

of particular import here was the inclusion of the inventory

of life events as developed by Paykel et al. (1971). Related

to this list of items was a series of probes which tapped

respondents' perceptions of each event and its general

influence on their mental health.

Construction of Life Event Indices

Measurement of Life Events

As stated in the discussion of life events in the

previous chapter, the definitions of the life event indices

are based on those of Paykel et al. (1971, 1972), ratings of

the degree of upset produced by events. That discussion

further pointed out an intention to assume that multiple

life events have additive effects. These decisions form the

basis of the main life event index.

Paykel's items are listed in Table 3.2, along with the

upset scores obtained for each. Because the interest here is

in all events occurring during the period between first and

second interview, the total event score is the sum of the

weights for those events. This is called the Paykel score.

Note that this particular form of scoring does not

account for multiple events of a single type, due to the

"yes" versus "no" response solicited for each event. This

is not a severe problem, however, because of the low occur-

rence of multiple events of the same type. Comparisons of

analyses using Paykel scores and scores based on multiple

occurrences of events, regardless of type, revealed no sub-

stantial differences. The correlation between the two forms

of index was r =.93. Also, the correlation between the

Paykel index and an unweighted equivalent, the count of

the different types of events, was r = .96.

Alternate Indices of Life Events

In addition to the Paykel index, alternative measures

of life events have been tested for use in the determina-

tion of the specificity of the results of the analyses.

Table 3.2 Paykel's Items; Their Weights and Frequency of Occurrence Within The Follow-up Sample (N=517)

Paykel's Number Percentage
Upset of Experiencing
Life Event Rating Responses the Event

1 Death of child 19.33 485 2.3
2 Death of spouse 18.76 484 2.9
3 Jail sentence 17.60 516 1.0
4 Death of family member 17.21 517 43.5
5 Extramarital affair (partner) 16.78 428 2.8
6 Major financial difficulties 16.57 517 9.5
7 Business failure 16.46 481 1.0
8 Fired 16.45 483 1.9
9 Miscarriage or stillbirth 16.34 368 1.9
10 Divorce 16.18 459 3.3
11 Marital separation 15.93 450 3.6
12 Court appearance 15.79 516 10.9
13 Unwanted pregnancy 15.57 372 1.6
14 Major illness of family member 15.30 516 29.5
15 Unemployed for one month 15.26 467 11.3
16 Death of close friend 15.18 517 26.5
17 Demotion 15.05 468 .6
18 Major personal illness 14.61 516 15.7
19 Begin extramarital affair 14.09 458 1.7
20 Loss of personal object 14.07 516 9.1
21 Law suit 13.78 517 2.1
22 Academic failure 13.52 469 1.1
23 Child married (not approved) 13.24 421 2.1
24 Break engagement 13.23 373 1.1
25 Increased arguments with spouse 13.02 431 8.4
26 Increased arguments with family member 12.83 504 5.2
27 Increased arguments with fiance 12.66 307 1.0
28 Take a loan 12.64 515 31.3
29 Son drafted 12.32 400 2.0

Table 3.2 (continued)

Paykel's Number Percentaqe
Upset of Experiencing
Life Event Rating Responses the Event

30 Troubles with boss or co-worker 12.21 454 6.8
31 Argument with non-resident family member 12.11 517 3.9
32 Move to another country 11.37 498 .6
33 Menopause 11.02 390 8.5
34 Moderate financial difficulties 10.96 516 15.7
35 Separation from significant person 10.68 517 10.1
36 Take important exam 10.44 481 7.5
37 Marital separation not due to argument 10.33 443 6.3
38 Change in work hours 9.96 473 14.4
39 New person in household 9.71 512 13.9
40 Retirement 9.33 472 4.7
41 Change in work conditions 9.23 481 10.8
42 Change in line of work 8.84 478 10.9
43 Cease steady dating 8.80 345 2.3
44 Move to another city 8.52 488 4.5
45 Change in schools 8.15 415 1.4
46 Cease education 7.65 424 5.0
47 Child leaves home 7.20 456 16.7
48 Marital reconciliation 6.95 437 3.4
49 Minor legal violation 6.05 515 8.7
50 Birth of live child 5.91 412 7.3
51 Wife becomes pregnant 5.67 272 7.0
52 Marriage 5.61 455 4.4
53 Promotion 5.39 467 14.8
54 Minor personal illness 5.20 517 23.4
55 Move in same city 5.14 505 15.8
56 Birth of child or adoption (father) 5.13 288 7.3
57 Begin education 5.09 472 7.0
58 Child becomes engaged 4.53 423 9.9
59 Become engaged 3.70 390 3.8
60 Wanted pregnancy 3.56 375 4.3
61 Child married (approved) 2.94 422 14.9

A major thrust in life event research has been analysis of

events by type rather than by means of a single index. The

variety of classifications used with life events was

identified in Chapter 2.

Although specific hypotheses were not stated for these

classifications, specification of results along these dimen-

sions is nonetheless considered to be important. The follow-

ing measures of life events have been developed for use in

later comparisons.

Exits and Entrances in the Social Field

Paykel et al. (1971) distinguish between exits and

entrances in the social field. The classification of life

events into exits and entrances is presented in Table 3.3.

The exit category includes those events which relate to

the loss of a social object or opportunity to perform one

or more roles. The entrance category includes events which

increase the respondent's role performances in the social

field. It is worth noting that Paykel reports more events

in the exit than in the entrance category, and that exits--

with few exceptions--have substantially higher scores on the

upset scale. These measures will be termed Paykel entrances

and Paykel exits respectively.

Independence and Symptom Relatedness of Life Events

B.P.Dohrenwend (974) has suggested that life events be


Table 3.3 Classification of Events as Entrances to and Exits from
the Social Field

Exits from the Social Field

Item Description Weight Pct.

Death of child 19.33 2.3
Death of spouse 18.76 2.9
Death of family member 17.21 43.5
Divorce 16.18 3.3
Marital separation 15.93 3.6
Death of close friend 15.18 26.5
Child married (not approved) 13.24 2.1
Break engagement 13.23 1.1
Separation from significant person 10.68 10.1
Marital separation not due to argument 10.33 6.3
Cease steady dating 8.80 2.3
Child leaves home 7.20 16.7
Child married (approved) 2.94 14.9

Entrances into the Social Field

Item Description Weight Pct.

New person in household 9.71 13.9
Marital reconciliation 6.95 3.4
Birth of live child 5.91 7.3
Marriage 5.61 4.4
Birth of child or adoption (father) 5.13 7.3
Become engaged 3.70 3.8

distinguished in terms of the likelihood that they may be

confounded with psychological or physical disturbances.

Paykel et al. (1975) recognizedJ a similar issue in distin-

guishing controlled from noncontrolled events. In accord-

ance with these concerns, events were classified in terms

of the probability that they might be precipitated by prior


The submeasures consist of weighted summations of events

within three categories. The first category consists of

events whose occurrence or scheduling is beyond the influence

of the respondent. Table 3.4 lists these events. They are

linked either to the life cycle (menopause, birth, or death

of others) or are externally scheduled or determined (take

an important exam, marital separation not due to an argu-

ment). These scores of symptom-independent life events will

hereinafter be referred to as Paykel-independent scores.

The second submeasure consists of events most likely

to have been precipitated by prior symptomatology. These

scores include items describing changes in interpersonal

relationships at home and at work. They also include items

related to changes in occupational or legal status having

a high potential for precipitation by symptom-related changes

in role performance. The items included are listed in Table

3.4. This submeasure of symptom-dependent life events will

hereinafter be referred to as Paykel-dependent scores.

Table 3.4 Classification of Events According to Judgments of Their
Dependence on Prior Symptomatology

Paykel Events Probably Independent of Prior Symptomatology

Item Description Weight Pct.

Death of child 19.33 2.3
Death of spouse 18.76 2.9
Death of family member 17.21 43.5
Miscarriage or stillbirth 16.34 1.9
Unwanted pregnancy 15.57 1.6
Major illness of family member 15.30 29.5
Death of close friend 15.18 26.5
Son drafted 12.32 2.0
Menopause 11.02 8.5
Take important exam 10.44 7.5
Marital separation not due to argument 10.33 6.3
New person in household 9.71 13.9
Birth of live child 5.91 7.3

Paykel Events Probably Related to Prior Symptomatology

Item Description Weight Pct.

Jail sentence 17.60 1.0
Extramarital affair (partner) 16.78 2.8
Fired 16.45 1.9
Divorce 16.18 3.3
Marital separation 15.93 3.6
Court appearance 15.79 10.9
Demotion 15.05 .6
Begin extramarital affair 14.09 1.7
Law suit 13.78 2.1
Academic failure 13.52 1.1
Break engagement 13.23 1.1
Increased arguments with spouse 13.02 8.4
Increased arguments with family member 12.83 5.2
Increased arguments with fiance 12.66 1.0
Troubles with boss or co-worker 12.21 6.8
Argument with non-resident family member 12.11 3.9
Cease steady dating 8.80 2.3
Minor legal violation 6.05 8.7
Minor personal illness 5.20 23.4

Table 3.4 (continued)

Paykel Events with Unknown Relationship to Prior Symptomatology

Item Description Weight Pct.

Major financial difficulties 16.57 9.5
Business failure 16.46 1.0
Unemployed for one month 15.26 11.3
Major personal illness 14.61 15.7
Loss of personal object 14.07 9.1
Child married (not approved) 13.24 2.1
Take a loan 12.64 31.3
Move to another country 11.37 .6
Moderate financial difficulties 10.96 15.7
Separation from significant person 10.68 10.1
Change in work hours 9.96 14.4
Retirement 9.33 4.7
Change in work conditions 9.23 10.8
Change in line of work 8.84 10.9
Move to another city 8.52 4.5
Change in schools 8.15 1.4
Cease education 7.65 5.0
Child leaves home 7.20 16.7
Marital reconciliation 6.95 3.4
Wife becomes pregnant 5.67 7.0
Marriage 5.61 4.4
Promotion 5.39 14.8
Move in same city 5.14 15.8
Birth of child or adoption (father) 5.13 7.3
Begin education 5.09 7.0
Child becomes engaged 4.53 9.9
Become engaged 3.70 3.8
Wanted pregnancy 3.56 4.3
Child married (approved) 2.94 14.9

The third submeasure is based on a residual category

of events: those not included in the symptom-independent

or symptom-dependent measures. The events falling into

this residual category are also listed in Table 3.4 and

appear to have somewhat less probability of having been

precipitated by prior symptomatology. These submeasures

will be referred to as Paykel-unknown scores.

It must be emphasized that the operational categoriza-

tion of events used in creating these submeasures is based

on judgments of the author and a colleague. As such, they

merely estimate the probability of symptom dependence.

However, this method was necessitated by the difficulties

in obtaining actual statistical estimates of symptom depen-

dence. This level of accuracy should be adequate for illus-

trating the symptom dependence of various classes of life


Area of Activity

The fourth categorization is based on the Paykel et al.

(1975) designation of areas of activity or roles affected

by the event. The five categories provided by Paykel are

work, health, family, marital, and legal. Because Paykel's

initial categorization was based on a sublist of only 32

events, it was necessary to modify the categories to include

all 61 items in the inventory. The categories used here

are education, work, legal and financial, personal health,

family, marital, and social, and will be named accordingly.

The family category excludes events involving only the

spouse, and the social category is largely residual. The

listing of these items by category is presented in Table 3.5.

Events by Year of Occurrence

The last set of life event measures stratifies the year

of event occurrence rather than the type of event occurring.

This set is designed to test the importance of how recently

an event occurred on the level of impact it may have on symp-

tomatology. Making use of the respondents' dating of events,

events were classified into approximate one-year intervals.

The periods are approximate due to faulty recall by respond-

ents and the unavailability of interview date within the

working files. This imprecision does not, however, cause

difficulty due to the one-year intervals used. The measures

are termed Paykel-1971, Paykel-1972, and Paykel-1973, based

on the year in which the time interval ended.

Measurement of Psychiatric Symptomatology

In the previous chapter psychiatric symptomatology has

been identified as one part of a syndrome of maladaptive

response to stress. The index of psychiatric symptomatology

used here, the Health Opinion Survey, or Macmillan Index,

was developed as a psychiatric screening device for use in

the community. As a screening device, it was intended to be

Table 3.5 Classification of Events by Area of Social Activity

Work Related

Item Description Weight Pct.

Fired 16.45 1.9
Unemployed for one month 15.26 11.3
Demotion 15.05 .6
Troubles with boss or co-worker 12.21 6.8
Change in work hours 9.96 14.4
Retirement 9.33 4.7
Change in work conditions 9.23 10.8
Change in line of work 8.84 10.9
Promotion 5.39 14.8

Legal and Financial

Item Description Weight Pct.

Jail sentence 17.60 1.0
Major financial difficulties 16.57 9.5
Business failure 16.46 1.0
Court appearance 15.79 10.9
Law suit 13.78 2.1
Take a loan 12.64 31.3
Moderate financial difficulties 10.96 15.7
Minor legal violation 6.05 8.7


Item Description Weight Pct.

Academic failure 13.52 1.1
Take important exam 10.44 7.5
Change in schools 8.15 1.4
Cease education 7.65 5.0
Begin education 5.09 7.0

Personal Health

Item Description Weight Pct.

Miscarriage or stillbirth 16.34 1.9
Unwanted pregnancy 15.57 1.6
Major personal illness 14.61 15.7
Loss of personal object 14.07 9.1
Menopause 11.02 8.5
Birth of live child 5.91 7.3
Minor personal illness 5.20 23.4
Wanted pregnancy 3.56 4.3


'Table 3.5 (continued)

Family and Parenting

Item Description Weight Pct.

Death of child 19.33 2.3
Death of family member 17.21 43.5
Major illness of family member 15.30 29.5
Child married (not approved) 13.24 2.1
Increased arguments with family member 12.83 5.2
Son drafted 12.32 2.0
Child leaves home 7.20 16.7
Birth of child or adoption (father) 5.13 7.3
Child becomes engaged 4.53 9.9
Child married (approved) 2.94 14.9

Marriage and Dating

Item Description Weight Pct.

Death of spouse 18.76 2.9
Extramarital affair (partner) 16.78 2.8
Divorce 16.18 3.3
Marital separation 15.93 3.6
Begin extramarital affair 14.09 1.7
Break engagement 13.23 1.1
Increased arguments with spouse 13.02 8.4
Marital separation not due to argument 10.33 6.3
Marital reconciliation 6.95 3.4
Wife becomes pregnant 5.67 7.0
Marriage 5.61 4.4
Become engaged 3.70 3.8

Other Social Events

Item Description Weight Pct.

Death of close friend 15.18 26.5
Increased arguments with fiance 12.66 1.0
Argument with non-resident family member 12.11 3.9
Move to another country 11.37 .6
Separation from significant person 10.68 10.1
New person in household 9.71 13.9
Cease steady dating 8.80 2.3
Move to another city 8.52 4.5
Move in same city 5.14 15.8

brief but, at the same time, able to identify those indi-

viduals most likely to be identified as psychiatric "cases"

if they were interviewed by a psychiatrist. This section

reports how the index was constructed, its current state of

review in the literature, and presents briefly some data

regarding its validity. Finally, some of the limitations

to its validity as a measure of psychiatric symptomatology

are discussed.

The Health Opinion Survey (HOS)

The measure of psychiatric symptomatology used in this

study is a 20-item index developed by Macmillan (1957) and

refined by Leighton (1965). Macmillan distinguished his

approach from other inventories by arguing that most

methods attempted "to study 'personality' as a whole" with-

out adequate theory. "Personality with all its ramifications

presented too great a range in complexity of phenomena

for a single test to deal with adequately" (Macmillan,

1957:326). Macmillan also cited efforts by the Armed Forces

during the Second World War to develop screening devices

more focused than those attempting to assess personality.

After reviewing a number of screening instruments, Macmillan

justified the development of a new one because "no test was

found which had been standardized on small-town and rural

adults--the major focus of the study in the Stirling County

research" (1957:327).

Macmillan's objective was "to detect those adults whose

responses to questions about their health approximated the

responses of psychiatric patients, and differed from the

responses of controls drawn at random from the community"

(1957:327). His comparison of hospital and general popula-

tion groups was based on a selection of items from the

Health Opinion Survey of the Stirling County study. These

had been drawn from the Army Neuropsychiatric Adjunct.

Macmillan found that 40 of the 75 items used in the survey

discriminated between hospital and general population groups

at the p < .01 level.

To reduce the length of the instrument, Macmillan

selected 20 items which distinguished neurotics from the

sample as a whole and from subgroups within it. The weights

obtained from a discriminant function analysis constituted

the initial scoring of the 20-item index. Using this set

of weights, Macmillan applied the measure to the community

sample and found that 25% of the 419 respondents had scores

in the case category. In contrast, nearly all (92%) of a

sample of neurotic patients fell into that category. In

addition, Macmillan had a psychiatrist visit 64 respondents

from the general population and found a high correspondence

between the psychiatrist's assessment and HOS scores. As

a result, he concluded that the index discriminated between

the general population and patients, particularly those who

were neurotic.

Validity of the Health Opinion Survey

Since the time Macmillan created the index, the HOS

has been used in a number of major epidemiological surveys.

These include the work of Leighton et al. (1963), Gurin et al.

(1960), Spiro et al. (1972), Edgerton et al. (1970), and

Warheit et al. (1975a). Concurrent with this widespread

application, several researchers have attempted to confirm

Macmillan's original claims for validity (Spiro et al., 1972;

Schwartz et al., 1973; Tousignant et al., 1974; and Kuldau

et al., 1976). Most of these studies provide support for

the validity of the HOS index as an epidemiological screen-

ing device, but not as a psychodiagnostic tool. Most of

the above authors identify theoretical limitations of the

HOS measure, but only one, Tousignant et al. (1974), appears

unwilling to accept the HOS as an epidemiological tool.

Discriminatory power. Macmillan's original work sug-

gested that the HOS is capable of discriminating neurotic

patients from a normal population. Yet, no index can be

properly validated against the same set of data used in its

construction. Validation against new sets of data--prefer-

ably drawn from different styles of psychiatric practice--

is required before firm conclusions can be reached regard-

ing the discriminatory power of the HOS measure.

Patients versus normals. In order to evaluate the dis-

criminatory power of the HOS, comparisons were made between

the scores of the sample population and a sample of patients

recently admitted to two general psychiatric inpatient

units (Kuldau et al., 1976). The HOS discriminated well,

both at the level of group means and for arbitrary cutting

points, as demonstrated by evaluation of the cumulative

percentile distributions presented in Figure 3.1.

Additional comparisons showed that respondents in the

general population possessing attributes indicating they were

at high risk for psychiatric disturbance had much higher

scores than individuals without these risk attributes. The

scores of these individuals were nearly as high as some cate-

gories of patients. Moreover, comparisons within patient

groups revealed that neurotics--the target group for the HOS

measure--obtained the highest scores, while other diagnos-

tic groups, including psychotics, scored much higher than

respondents in the general population. These comparisons

indicate that the HOS index does have the ability to dis-

criminate between patients and normals.

Item content and dimensionality. Seiler (1973) has

raised the issue of whether the discrimination of patients

from normals provides a sufficient criterion for validity.

Specifically, he expresses concern that such comparisons

deflect attention from questions regarding the item content

of indices such as the HOS. While these may be valid con-

cerns, they constitute to some extent an ideal standard of

validity which has not yet been attained.

90 -


70 A B C D E



03o0 A Community Sample: Low Risk
B : Medium Risk
SC : High Risk
20 D Patient Sample: Neurotic Anxiety
E : Neurotic Depressed

10 -
0 --- ^ --- i __ i __ i __ i __ i -- i __ -- -- i -- i -- I -- i -- -- --

20 22 24 26 28 30 32 34 36 38 40 42 44 46 48 50 52 54

Health Opinion Survey Scores

Figure 3.1

Cumulative Percentage Distributions of HOS Scores For Community
and Patient Samples

Nonetheless, it is useful to ask whether the HOS index

is composed of symptoms which are necessary but not suffi-

cient indicators of mental illness. If this were the case,

application of the index would overestimate the level of

mental illness in a sample population. At the individual

level, these cases would be false-positives. This aspect

of the validity issue can be constructively addressed by

studying individuals not hospitalized for mental illness but

identified as cases by their HOS scores.

Another issue related to the content of the index stems

from the fact that the HOS is designed to measure only a

single dimension of psychiatric symptomatology, caseness.

Nonetheless, many discussions consider the phenomenon to be

multidimensional (cf. Jahoda, 1958; Schwab et al., 1970).

While mental illness may well be a multidimensional phenome-

non and should probably be understood as such, parsimony

dictates that there be a point of departure for empirical

analysis. The HOS is not an end in itself but rather a

device for discriminating patients and non-patients which

will contribute to the subsequent discovery and analysis

of other dimensions of mental illness.

Stability and physical health. The two remaining issues

related to the validity of the HOS are its stability over

time and the strong physical health content of the component

items. Does the HOS therefore measure intermittent phe-

nomena--such as minor physical illness--which might be

assumed to vary substantially over time, or does it measure

more fundamental and stable processes?

The results of comparing HOS values over the three-

year period of the panel study, 1970 to 1973, reveal sub-

stantial stability (r=.73), suggesting that the HOS does

measure relatively stable processes. While such a high

degree of stability was unexpected and therefore may warrant

further study, it indicates that the HOS definitely does

not measure transient phenomena.

Transience would undermine the validity of the index

in several ways: by casting doubt on the seriousness of the

phenomena being measured and the consequent need for psy-

chiatric intervention; by suggesting that the index might

be tapping minor personal illnesses rather than mental

illnesses; by calling into question the appropriateness of

including a substantial proportion of physical health items

in the index; by undermining the utility of the HOS measure

in panel designs; and finally, by challenging the view that

there is neither a sound theoretical nor empirical basis for

dichotomizing health into physical and mental components

(Schwab et al., 1970; Eastwood and Trevelyan, 1972).

Fortunately, the stability of the HOS scores over a three-

year period militates against these potentially negative


Measurement of Socioeconomic Status

The SES measures constructed for the 1970 and 1973 data

are based on an adaptation of the methodology reported by

Nam and Powers (U.S. Bureau of the Census, 1960) but modified

to provide individual instead of family SES scores. The

methodology was applied using 1970 data on occupation (Nam

et al., 1975), education, and income ranks.

For respondents who were unmarried, no distinction was

made between family and personal SES. For 'married respond-

ents who had a spouse living with them, the procedure in

scoring SES was to determine the principal breadwinner. The

principal breadwinner is defined as the respondent or spouse

having the greater earned personal income.

National percentile ranks were obtained for each of the

components of the SES measure. These included earned per-

sonal income of the respondent, the last year of education

completed by the respondent, and an occupation score based

on the methodology of Nam et al. (1975). The occupation

scores were developed by applying the 1960 methodology to the

occupation categories of the 1970 census. This methodology

takes the average of income and education percentiles for

each individual falling nationally within an occupational

category. This average is then used to rank occupations,

providing a percentile score for each occupation. These

national ranks, based on aggregated values, are used as

scores for individuals. The process of aggregation

and disaggregation provides an occupational ranking independ-

ent of the individual's income and education. In their 1975

report, Nam et al. provide scores based on the 1970 census

for four groups: all employed males, all employed females

regardless of fulltime-parttime status, all fulltime

employed females,and finally, all employed individuals regard-

less of sex and fulltime-parttime status. It is this latter

figure which is used in computing occupation scores. Be-

cause Blau and Duncan (1967) and Nam and Powers (1968) have

reported great stability in the occupational structure,

1970 national norms were used for both 1970 and 1973 occupa-

tion scores. The ranking for education and income were

also based on data from the 1970 census. No correction is

made to the income figures to account for inflation during

the three-year period because of uncertainty about an appro-

priate figure and because such shifts can be controlled

directly in the analysis.

The individual SES score adopted for this analysis was

the average of the three measures of occupation, education,

and earned personal income rank scores for each respondent.

In the event that one or more of these items was missing or

unavailable, the SES measure was based on the average of

the remaining items. A major advantage of this individual

SES index is that it corresponds to the conceptual orienta-

tion advanced earlier by measuring resources available to

the individual in adapting to stressful life situations.

Plan of Analysis

Using the design and operational definitions developed

above, it is possible to test the working hypotheses relat-

ing life events, psychiatric symptomatology, and socio-

economic status. This section provides a detailed plan for

that test of hypotheses.

Before proceeding, a comment about hypothesis testing

is in order. Hypothesis testing should lend itself to the

verification and refinement of propositions developed as

part of a body of theory. To provide this verification,

hypotheses must be falsifiable. Nonetheless, the hypotheses

may relate to the parameters of a model, and thus the test

involves more than the statement that a given effect is pres-

ent or absent. Such an approach has been attempted in the

present research. Such an estimation approach requires

that the statistical models relating to the test of hypothe-

ses be made explicit. This is accomplished in the remainder

of the chapter. By convention, the p < .05 level of proba-

bility is used throughout the analyses as the minimum

standard for rejecting the null hypothesis.

The statistical analysis of the present research con-

sists of two major parts. The first part attempts to estab-

lish the general form of the relationship between life events,

psychiatric symptomatology, and socioeconomic status, and

thus focuses exclusively on the Paykel index as a measure

of life events. The second part has as its purpose the

further specification of the relationships among the central

variables by retesting the focal hypotheses with the alter-

native measures of life events. This retest serves to iden-

tify the differential influence of the various categories

of life events, and thus to provide specificity to the gen-

eral model developed in the first part.

The Specific Analyses: Part 1

Transition Between Caseness Categories

Because the present design permits an examination of

change in symptom levels, the analysis begins with an exami-

nation of Hypothesis 1, which relates increases in HOS scores

over the period of study to the occurrence of life events as

measured by Paykel scores. Change can be defined in many

ways. This first analysis makes use of a simple definition

of change based on a dichotomization of HOS scores at the

value of 30. Scores of 30 or above have been shown in the

literature to correspond to elevated probabilities that the

individual would be considered a "case" of psychiatric dis-

order were he to be examined by a psychiatrist. Therefore,

if Hypothesis 1 is correct, then more frequent transition

from "non-case" to possible or probable "case" levels of

HOS score will be associated with high Paykel scores.

Hypothesis 2 is tested in a similar fashion through

presentation of transitions from non-case to case levels on

the HOS index for two levels of SES. SES scores have been

dichotomized (0-39:40-100) because the use of more categories

produces cells too small for appropriate statistical test.

Through this point only those with initially low HOS

scores have been considered. The above analysis is repeated

for those with high initial HOS scores by using retentions

within the case level of the HOS as evidence of stress.

Analysis of Change Scores

Although the previous analysis dichotomized HOS scores,

a more powerful analysis might deal with change in HOS scores

as an interval measure. Such an approach was used by Myers

et al. (1971) and Haberman (1965) and is suggested by

B.P. and B.S. Dohrenwend's (1969:128) examination of the

direction of change in symptom scores. Subtraction of HOS-

1970 scores from HOS-1973 scores results in a positive value

for increases in level of symptomatology and negative values

for decreases. Such an approach identifies change at any

point of the HOS score continuum and permits the use of more

powerful interval statistics. Unfortunately, however,

change scores introduce a bias which undermines their at-


Bias introduced in the use of change scores is termed

statistical regression toward the mean, a problem discussed

most recently by Coleman (1968), Bohrnstedt (1969), and

Cronbach and Furby (1970). This is a statistical effect

rather than a causal factor acting on individuals, and is

based on the probability that the retest scores of an indi-

vidual with extreme scores on any measure which has a com-

ponent of random error will, on the average, be closer to

the mean for that measure than the initial observation,

other factors being equal. In short, change scores for those

with very high initial symptom scores will tend to be nega-

tive and change scores for subjects with very low initial

symptom scores will tend to be positive. Thus, the use of

change scores introduces a bias whenever groups with selec-

tively high initial scores are examined.

In spite of the potential for bias, it is possible to

introduce a simple examination of change scores in which

control variables have not selected individuals with differ-

ing levels of initial score. Such a presentation has

strong intuitive appeal and sets the stage for illustration

of the bias described. Thus, this preliminary test of

Hypothesis 1 uses change in HOS scores as the criterion and

levels of Paykel score as the independent variable. The

potential bias is also illustrated.

Regression Analysis of Change in HOS Scores

A solution to the difficulties with change scores is

available by adopting a least squares regression approach

in which HOS-1973 scores are estimated from a model which

contains HOS-1970 scores as a predictor variable. This

approach has been advocated by Bohrnstedt (1969), Coleman

(1968), and Cronbach and Furby (1970). It is also incor-

porated into the change models of Duncan (1975b) Galtung

(1975), Kenney (1973), and Pelz and Andrews (1964). It

is the primary method used for the analysis of change in

this study.

The simplest regression model for analyzing change

within the present two-wave panel design contains two vari-

ables, the HOS-1970 and HOS-1973 scores. This model may

be stated as:

HOS-1973 = 8o + Bi HOS-1970 + e (3.1)

HOS-1973 is the dependent (time 2) variable; Bo is a con-

stant; HOS-1970 is the independent (time 1) variable; and

e is the term for random error. This simple model is easily

estimated by ordinary least squares methods. For explicit

discussion of the error term,see Bohrnstedt (1969).

In the model just described, Bi estimates the strength

of relationship between HOS-1970 and HOS-1973 scores. It

may also be seen as a measure of the strength of the regres-

sion effect occurring. The term Bo is a constant

which accounts for shift in the mean HOS score level from

1970 to 1973. The error term e is assumed to have a mean

of zero but otherwise accounts for individual lack of fit

in the model.

The model as stated includes no terms for estimating

other influences contributing to change between 1970 and

1973. This is provided by adding an additional term.

HOS-1973 = 0o + BI HOS-1970 + B2 Paykel+ c ( .2)

The additional term consists of a life event variable and

its regression coefficient. The coefficient estimates the

influence of Paykel scores on IIOS-1973 scores with a simul-

taneous control for HOS-1970 scores. It therefore provides

an estimate of the influence of life events on symptom change

between 1970 and 1973.

Only one term for estimation of the influence of life

events upon symptom change has been considered. In fact,

many additional terms might be included without affecting

the method of estimation. This procedure affects the inter-

pretation for each coefficient obtained only to the extent

that additional influences are being simultaneously controlled.

Additional explanatory terms are necessary, both for full

elaboration of the influences of life events, and for the

introduction of relevant sociodemographic controls.

Life Event and SES Interaction

Although a method has been developed for examining

change in HOS scores, the regression model of Hypotheses

1 and 2 is not yet complete. The hypothesized influence of

life events is positive but depends on SES. The conditional

nature of this influence is termed interaction. A method

must be provided for inclusion of this interaction effect

in the regression model.

The additive influences of Paykel scores and SES are

easily modeled as follows:

HOS-1973 = 0o+Bi HOS-1970 + B2 Paykel + 83 SES + E (3.3)

The coefficients 32 and 63 indicate the influence of Paykel

and SES scores on HOS-1973. In particular 82 indicates

that the influence of Paykel on HOS-1973 is incrementally

the same regardless of level of SES and therefore is not

equivalent to the interactive relationship of Hypothesis 2.

Hypothesis 2 suggests that the influence of Paykel is not

the constant suggested by 82 but rather a variable deter-

mined by the level of SES. If 82 is assumed to be a func-

tion of SES, then the term 82 can itself be modeled as:

62 = Ci + C2 SES (3.4)

Substituting this in the original equation provides:

HOS-1973 = Bo+38 HOS-1970 + (al + a2 SES) Paykel

+ 83 SES + E (3.5)

which expands to:

HOS-1973 = 8o+31 HOS-1970 + al Paykel + 02 SES x

Paykel + 83 SES + E (3.6)

Since the terminology of coefficients is arbitrary, the

model can be restated as:

HOS-1973 = Bo+31 HOS-1970 + B2 Paykel + 63 SES +

64 SES x Paykel + e (3.7)

Through the formation of a new variable, SES x Paykel, this

new model becomes estimable through ordinary least squares

multiple regression methods. The cross product term, SES x

Paykel, models the portion of the influence of Paykel on

HOS-1973 which is conditional on SES,while the simple

linear term, 62 Paykel, models the unconditional portion of

the relationship. Thus equation 3.7 accurately models the

influence considered in Hypothesis 2,although additional

sociodemographic controls may be added.

Multicollinearity. Although the inclusion of cross

product terms in regression models is a common practice, it

frequently creates a condition which distorts the coeffi-

cients for both the linear and interaction terms. This is

the problem of multicollinearity, a result of strong inter-

dependence among the predictor variables of the regression

model (cf. Blalock, 1972:457; Huang, 1970:154). Multi-

collinearity produces distortions both in the size of regres-

sion coefficients and in the testing of significance for

the terms affected.

The new variable used in the cross product term is

formed through multiplication of two or more original

variables and thus may be correlated with them (Blalock,

1972:463-364). Whether the cross product is correlated

with its components depends on the distribution of those

variables about the value zero. Cross products formed from

variables which have been centered, through subtraction of

their means, typically provide lower correlations with

their original components than cross products obtained from

uncentered data. This suggests that multicollinearity due

to cross products may be reduced through centering the

variables before forming the cross product. The use of

cross product terms formed in this way permits the estimated

linear effects of the variables to be unchanged by the intro-

duction of the cross product, while retaining an easily

understood interpretation of the interaction term as an

adjustment to the linear slope of the first variable based

on the respondent's deviation from the mean on the second


Overall, this approach is more direct and less con-

fusing than the attempt to interpret uncentered cross

products in the presence of multicollinearity, or the at-

tempt to interpret alternative methods for controlling that

multicollinearity (cf. Huang, 1970:149-150).

As a result of these considerations, the influence of

life events is tested in two parts, the unconditional linear

effect and the correction to the life event influences due

to deviation from the mean level on the SES variable.

HOS-1973 = Bo+B1 HOS-1970 + B2 Paykel + 83 SES +

84 (SES-SES) x (Paykel-Paykel) + E (3.8)

The regression modeling of the influence of Paykel

scores and SES on HOS scores both in this and subsequent

sections of the analysis is based on equation 3.8 with the

addition of terms for control variables. The control terms

in the equation include the age of the respondent in 1970,

age squared, dummy variables for race, sex, marital status,

and employment status as appropriate. Additionally, a

cross-product term for interaction between age and HOS-1970

is included; it is based on centered variables to reduce


The preceding discussions have outlined a regression

model for use in testing Hypotheses 1 and 2. The specific

test of Hypothesis 1 is a test of whether the regression

coefficient for the linear effect of Paykel scores, 82,

is significantly greater than zero. The test of Hypothesis

2 is a test of whether the regression coefficient for the

cross product of Paykel scores and SES is significantly

less than zero. This coefficient is negative due to the

inverse relationship of SES to HOS scores.

Regression Model Predicting Life Events

The fourth analysis provides a test of Hypothesis 3.

It examines the influence of initial levels of psychiatric

symptomatology, HOS-1970, on the events occurring during the

following three years. This analysis makes use of multiple

regression, in particular, the following model:

Paykel = N0oi HOS-1970 + control terms + E (3.9)

The control terms are selected as appropriate from

those identified in the above analysis, with the exception

that only 1970 values are used. The test of Hypothesis

3 in terms of this model is a test of whether the regres-

sion coefficient for the HOS-1970 term, 81, is significantly

greater than zero.

Cross-Sectional Models for HOS-1973

Although longitudinal studies of the influence of life

events are preferable to cross-sectional ones, few such

studies have been reported, and thus cross-sectional analy-

ses are a major point of comparison between the literature

and the present study. In addition, a comparison of cross-

sectional and panel analyses serves to highlight the differ-

ences between interpretation of our hypotheses as change

within individuals and differences among individuals. To

this end, Hypotheses 1 and 2 are retested using the 1973

cross section. An analysis for the 1970 cross section is

not possible due 'to the lack of event data for that inter-

view period.

The model for testing the hypotheses cross-sectionally

is different from the panel model in only two terms. These

are the terms for HOS-1970 and for HOS-1970 x age, a control

term. Thus, the regression equation is as follows:

HOS-1973 = 00+B2 Paykel + 83 SES + B4 (Paykel-Paykel) x

(SES-SES) + controls + E (3.10)

The cross-sectional test of Hypothesis 1 is a test that the

regression coefficient for the Paykel term, B,2 is greater

than zero. The verification of Hypothesis 2, that SES

determines the strength of influence of life events on

symptomatology, requires that the regression coefficient

for the cross-product term, ,, be less than zero. As in

the panel analysis, this coefficient must be negative due

to the inverse relationship between SES and HOS scores.

The Specific Analyses: Part 2

In Part 1 of the analyses, outlined above, the life

event variable used in each test of hypotheses was the

Paykel index. This index is based on a single formulation

of events, the typical amount of upset produced. As a

single index, the Paykel measure provides no means for

assessment of the differential impact of events of differ-

ing types.

The purpose of the second part of the analysis is to

provide greater specificity to our knowledge by retesting

the hypotheses with the alternative measures of life events

detailed above. Such a detailed examination, although

exploratory, can serve to refine the general model and to

suggest improvements in the measurement of life events.

Weighted Versus Unweighted Paykel Measures

The next analysis is a comparison of the unweighted

version of the Paykel index with the Paykel index used in

the analyses above. The hypotheses tested correspond to

panel versions of all three hypotheses. This test addresses

whether weighting of items by the typical level of upset

provides a model different from the one obtained above.

This difference is assessed in terms of the amount of

variance explained and the change in standardized partial

regression coefficients.

The second analysis involves the use of individual

events in test of Hypothesis 1. Because a large number of

items are considered, the interaction of each item with SES

is ignored and stepwise forward selection is used in ob-

taining a solution to the model. In this procedure, items

are entered in order of their contribution to the regression

equation at the end of the previous step. Thus, at each

step the item adding the most to the prediction achieved in

the previous step is entered. The procedure of adding

variables is terminated when no additional variable makes

a significant increase in the prediction. This procedure

serves to identify the events most strongly predicting

change in HOS scores but cannot provide an adequate esti-

mate of influence for each of the 61 events in Paykel's


Entrances and Exits

The next analysis is a retest of the hypotheses using

two event measures, one measuring entrances to the social

field and the other measuring exits. This test of Hypothe-

ses 1 and 2 identifies the measure with greater power to

predict change in HOS scores. The regression model and test

of hypotheses differ from the prior model of change in HOS

scores only in the duplication of life event terms considered.

Analyses equivalent to those on page 67 are used to test

Hypothesis 3. This is done separately for the two event


Symptom Relatedness of Life Events

This analysis is focused on the issue addressed by

Hypothesis 3, the dependence of events on prior levels of

symptomatology. Although Hypothesis 3 is addressed directly

in the section on regression model predicting life events,

this additional test helps to specify the particular kinds

of events involved. Based on theoretical and intuitive

judgments, items are classified by symptom relatedness.

If Hypothesis 3 is correct and the judgments about symptom

relatedness are correct, the three hypotheses should be

most strongly supported for symptom related events.

Life Events by Area of Activity

This analysis seeks to identify by area of social

activity the types of events for which the hypotheses hold

true. The literature reviewed above suggests personal

health is the area most related to the production of psy-

chiatric symptomatology (cf. Selye, 1957; B. P. Dohrenwend,

1974). Once again, the test of three hypotheses is parallel

to the above panel analyses.

Analysis by Year of Occurrence

The last analysis to be included in this report deals

with the time interval inwhich each event occurred. Through

this point in the analysis, all events are treated alike

regardless of when they occurred during the three-year

period of the panel design. Yet, as was discussed earlier,

the recency of events may be of importance if their major

effects are transient rather than long-term. The effects

examined are long-term ones; and the design was developed

accordingly. Long-term effects of this type should show an

impact from events from all three years of the study. In

contrast, a preponderance of transient effects would cause

the observed influence of events on symptoms in 1973 to be

limited to those events occurring most recently (i.e.,

Paykel, 1973). Thus a reexamination of Hypotheses 1 and 2

provides a test of one of the assumptions of the present




The purpose of this chapter is to test the working

hypotheses with the data from the present study. The first

part of this analysis establishes the form and extent of the

interrelationship between Paykel scores, HOS scores and SES;

the second part retests the hypotheses with life event

measures other than the overall Pavkel index. These

analyses identify the life event measures for which the

hypotheses hold true and thereby identify the sets of life

events with the greatest impact on HOS scores.

Distribution of Variables

The distributions of the variables used in the analyses

are presented in Tables 4.1 and 4.2 so that the reader will

be better able to interpret the analyses which follow.

One important comment is that many of the variables, par-

ticularly HOS and life event measures, are highly skewed

with many observations several standard deviations above

the mean. This skew is a natural phenomenon consistent

with assumptions about true distributions of the variables

involved. As a consequence, no transformation has been used.

Table 4.1 Means and Standard Deviations for Major Variables from

Mean S.D.

Mean S.D.

Between Years





Sex (female) '

Race (black)

Unemployed (yes)

Not married (yes)










*Age for 1970 is used throughout, in order to avoid introduction of spurious variance due to birth date.
#Sex and race are constant over the three-year period.
tThe means for the dummy variables are proportions of respondents in the specified category, and may be
converted to percentages through multiplication by 100.

--------------~--- -- _-T=-_~=~-_~-_~=T_~-

1970 and 1973 Surveys

Table 4.2 Means and Standard Deviations for Life Event Measures

Event Measure Event Item Count Weighted Score Upset
(Number of Items) Mean (N=517) S.D. Mean (N=517) S.D. Weight

Total Paykel (61) 4.73 3.57 53.10 39.56 11.23

Relation to Social Field

Exits (13) 1.27 1.09 16.91 14.19 13.31
Entrances (6) .33 .63 2.41 4.59 7.30

Relationship to Symptomatology

Independent (13) 1.47 1.16 21.19 16.45 14.41
Dependent (19) .83 1.22 9.06 14.97 10.91
Unknown (29) 2.42 2.35 22.86 22.84 9.45

Area of Living

Education (5) .20 .54 1.58 4.55 7.9
Work (9) .69 1.14 6.91 11.97 10.01
Legal & Financial (8) .80 .98 10.08 12.39 12.60
Personal Health (8) .66 .83 6.34 8.81 9.61
Family (10) 1.22 1.17 15.46 13.51 12.67
Marital (12) .39 .81 4.38 9.65 11.23
Other Social (9) .77 .89 8.33 9.66 10.82

Year of Occurrence

Paykel-1971 (61) .91 1.16 10.08 13.38 11.08
Paykel-1972 (61) .99 1.22 10.82 13.88 10.93
Paykel-1973 (61) 1.94 1.98 21.03 22.17 10.84

The main consequence of the use of such skewed data is

likely to be an increased estimate of error variance, and

thus any bias occurring is likely to be in the direction of

accepting the null hypothesis. Additional support for this

expected conservatism is seen in slight increases in the co-

efficient of determination when data from the full 1970 sur-

vey were redistributed in the form of a normal distribution.

The Form and Strength of the Relationship

Analysis of Transition Between
HOS Caseness Categories

An initial test of Hypotheses 1 and 2 is based on treat-

ment of HOS scores as a dichotomy between those with low

(non-case) levels of symptomatology (HOS=20-29) and those

with high (case)levels (HOS=30+).

The left-hand column of Table 4.3 reports only those

respondents with the HOS-1970 scores below the caseness

level. The figures presented for each level of Paykel

score show the percentage of respondents with low 1970

scores but whose HOS-1973 scores had increased to the case

level. Hypothesis 1 states that those with high Paykel

scores experience greater increases in symptomatology. This

is supported because the percentage of respondents whose

scores increased to caseness levels is nearly 20.7%, nearly

double the 10.4% and 11.6% found for the middle and lower

Table 4.3 Transition of Respondents Between Case Categories by Paykel Scores, Controlling for SES Level

Transition from Non-Case to Case Transition from Case to Non-Case
Number of Non-Cases, Percentage Becoming Number of Percentage Remaining
1970 a Case Cases, 1970 in the Case Level

All Respondents

Paykel Score

62.01 +


X2 = 6.21
p < .05


X2 = 11.43
p < .01

SES Levels 0-39

Paykel Score

62.01 +


X2 = 7.48
p < .05


X2 = 3.26

SES Levels 40-100

Paykel Score

62.01 +


X2 = 3.49


X2 =6.45
p < .05

Paykel levels. The initial test of Hypothesis 2 states that

the differences in transitions by Paykel level should be

greater for the lower SES levels than for higher SES levels.

This also is supported. For those with low SES scores and

high Paykel scores the percentage with HOS scores increasing

from non-case to case levels is 43.5%, more than three

times that for the low Paykel score group. In contrast,

the differences in transition between high and low Paykel

levels for the high SES group is slight, being only 15.1%

and 10.8%, respectively. Thus, life event scores are asso-

ciated with increased changes from non-case to case levels on

the HOS. The strength of that association is greater for

those with low SES levels.

The right-hand side of Table 4.3 presents the effect of

life events on those who had high initial HOS scores. Hy-

pothesis 1 is again supported by the greater retention at the

case level of those with high Paykel scores. The retest of

Hypothesis 2 fails in this particular context. Although re-

tention of respondents in the case category was associated

with Paykel scores at both levels of SES, the influence was

great or greater for those with high SES. This may reflect

a ceiling effect, because nearly all of the 30 respondents

in the high Paykel-low SES category (90%) were retained at

the case level.

Analysis of HOS Change Scores

A second analysis,to be considered only briefly, is the

analysis of HOS change scores. Table 4.4 presents the mean

of HOS change scores for three levels of Paykel scores. As

can be seen, those with highest Paykel scores show a

slightly greater increase in HOS levels than the other groups,

but not enough to reject the null Hypothesis 1 that no rela-

tionship exists between Paykel scores and change in HOS

scores. Thus, Hypothesis 1 is not supported by analysis

of uncontrolled change scores.

The remainder of the table confirms the methodological

assumption that change scores will, on the average, be posi-

tive for low initial HOS values and negative for higher

initial HOS values.

Regression Model of the Change in HOS Scores

In the methods chapter, a regression model for the

analysis of change in HOS scores has been formulated. This

regression equation predicts HOS-1973 scores from HOS-1970

scores, Paykel scores for the period from 1970 to 1973,

the respondent's SES in 1973, and additional control variables

for age, sex, race, and employment. Several of these vari-

ables have been handled in special ways. One term, included

to model the interaction between Paykel and SES scores, is

the product of the Paykel and SES scores, formed after

Table 4.4 Change in HOS Scores by Paykel Scores, Controlling for 1970
HOS Scores



Paykel Score

62.01 +






F = 1.43
df = 2,514

Non-Case (HOS 20-29)

Paykel Score

62.01 +

Paykel Score

62.01 +

Possible Case (HOS 30-34)

S -1.343 4.172
-1.269 4.441
3 1.848 4.969

Probable Case (HOS 35+)

Paykel Score

62.01 +






subtraction of their respective means. Among the control

variables, the square of age, after subtraction of its mean,

and the product of age and HOS-1970 after subtraction of their

means are included. The two remaining control variables,

race and employment, are treated as dummy variables: race

(black=l) and employment (not employed=l). The remaining

categories of the dummy variables are coded as zero.

Estimation of this regression model from the present

data (Table 4.5) reveals that the major determinant of HOS-

1973 scores is the respondent's HOS-1970 score. Although this

was anticipated, the strength of the prediction of HOS-

1973 scores by HOS-1970 scores is greater than expected for a

three-year period, explaining 53.4% of the HOS-1973 score

variance. This stability of HOS scores over time can be

considered as partial confirmation of the reliability of the

HOS measure, but consideration should be given to the other

factors which may govern this stability.

The two hypotheses to be examined concern the influence

of Paykel scores. The first, Hypothesis 1, examines the

linear effect of Paykel scores on the change in HOS scores.

The metric regression coefficient for the Paykel term is

B=.02209 indicating an increase in HOS scores of only two-

tenths of a point for 10 points on the Paykel measure, or

2.2 points for 100 points on the Paykel measure. Ten points

on the Paykel corresponds to a single event producing only a

Table 4.5 Regression Analysis of 1973 HOS Scores as Change
Level as Predictors

from 1970 HOS Scores, Using Paykel Scores and SES

Regression Coefficients
Metric Standardized Standard Signifi-
Variable Beta Beta Error Beta F chance

HOS-1970 0.64917 0.61304 0.03262 395.970 p < .001
SES-1973 -0.02799 -0.13330 0.00737 14.412 p < .001
Paykel 0.02209 0.14608 0.00470 22.093 p < .001
SES x Paykel* -0.00050 -0.08380 0.00017 8.628 p < .01
Age 0.01525 0.04194 0.01255 1.477 N.S.
Age Squared -0.00182 -0.08790 0.00061 8.899 p < .01
HOS x Age* 0.00737 0.11396 0.00187 15.591 p < .001
Race-Black 1.08978 0.08047 0.43551 6.262 p < .01
Not Employed-1973 1.32949 0.10992 0.39367 11.405 p < .001
(Constant) 1.19247

*Variables were centered before cross-product was formed.

Multiple R = 0.77940

R Squared= 0.60746

R Squared= 0.60049

Error = 3.78116

Analysis of Variance

Sum of Mean Signifi-
Source d.f. Squares Squares F chance

Regression 9 11217.27 1246.56 87.18 p < .001

Residual 507 7248.66 14.30

Total 516 18465.93

moderate amount of upset for the average person. One point

onthe HOS corresponds to increased occurrence of one symptom.

The interaction between Paykel and SES, modeled as the cross

product of Paykel and SES after subtraction of their means,

measures the change in the average influence of Paykel scores

for deviation from the mean SES score. Thus, the influence

for Paykel scores,given above,represents an average or

typical influence which varies with level of SES. The inter-

action term has a coefficient of B=-.00C5, indicating that a

decrease in SES produces an increase in the change in HOS

scores for every unit on the Paykel. This change in the

estimated influence of Paykel scores is nearly as great as

the strength of the Paykel coefficient itself. Over the 100

point variation of SES the change in the Paykel coefficient

is estimated to be .05,and thus the influence of life events

varies from zero through twice the average influence of life

events, depending on the level of SES. Thus both Hypothesis 1

and Hypothesis 2 are supported. These state that life events

have an influence on symptomatology, and that the strength

of that influence increases with decreasing socioeconomic


In the tests of hypotheses above, the regression model

provides control for a number of additional variables. This

control takes the form of an implicit linear adjustment of

the dependent variable for the effects of the controlled

University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs