The relationship between work-related life crisis events and mental health


Material Information

The relationship between work-related life crisis events and mental health
Physical Description:
xii, 125 leaves : ; 28 cm.
Ghyst, Dennis Earl, 1944-
Publication Date:


Subjects / Keywords:
Job stress   ( lcsh )
Work -- Psychological aspects   ( lcsh )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )


Thesis--University of Florida.
Includes bibliographical references (leaves 113-124).
Statement of Responsibility:
by Dennis Earl Ghyst.
General Note:
General Note:

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 000098170
notis - AAL3613
oclc - 06689714
System ID:

This item is only available as the following downloads:

Full Text







Copyright by
Dennis Earl Ghyst




I wish to thank Dr. George Warheit, my academic

chairman, for his guidance and support throughout my

graduate career. His constructive criticism and

encouragement have made it possible for me to bring

this dissertation to completion.

I would also like to express my deep gratitude to

Dr. Charles E. Holzer and to Mrs. Lynn Robbins for

their indispensable methodological feedback and

computer analysis assistance.

These remarks would be incomplete without giving

special thanks to my committee members: Drs. Anthony

LaGreca, Gerald Leslie, Charles Frazier and Robert

Ziller for their consistent support.

Finally, there is a special group of close friends

whose unstinting emotional and material support have

made it possible for me to complete this task. Among

these are Ann, Frieda, Bryan and T.J. And a heartfelt

thanks to Linda Johnston for her excellent job of typing

this dissertation.









Toward a Working Definition of
Mental Disorders . 6
Stress Literature . 9
Stressful Life Events 14
LCEs, Coping Variables and
Impairment . 20
Work and Mental Health 29

Background to Study . 48
Major Social-Psychiatric Variables 50
Measurement of Life Crisis Events 54
Demographic and Resource Control
Variables 57
Plan of Analysis: Foundations 57
Plan of Analysis: Formal
Hypotheses . 61
Other Measures of Work
Behavior . 65





Introduction .. .
Employed Sample Characteristics .
Description of WRLCE .
Regression Analysis of HOS and
PDS Change Scores ..

. 67
. 67
. 67
. 77

. 80

Introduction . .
Summary of Findings of Hypotheses
Implications for Future Research







Table Page

4.1 Demographic Characteristics of The
Original and Follow-up Worker Samples 69

4.2 Full Sample HOS Scale Scores by Sex, Age,
Race, Income . 70

4.3 Employed Sample HOS Scale Scores by Key
Social and Demographic Variables ... 72

4.4 Full Sample PDS Scale Scores by Sex, Age,
Race, Income . 75

4.5 Employed Sample PDS Scale Scores by Key
Social and Demographic Variables.. 76

4.6 Paykel's Work-Related Items, Their Weights
and Frequency of Occurrence . 73

4.7 Transition of Respondents Between Case
Categories Controlling for WRLCE ... 79

4.8 Regression Analysis of T2 HOS Scores as
Change from T! HOS Scores, Using Major
WRLCE and Sociodemographic Controls as
Predictors . 82

4.9 Regression Analysis of T2 PDS Scores as
Change from T1 PDS Scores, Using Major
WRLCE and Sociodemographic Controls as
Predictors . ... 85

4.10 Regression Analysis of HOS and
Sociodemographic Predictors of WRLCE
Scores . . 87

4.11 Regression Analsysi of PDS and
Sociodemographic Predictors of WRLCE
Scores . ... 88


LIST OF TABLES (continued)

Table Page

4.12 Regression Analysis of Time 2 HOS, Using
Income, Occupation and WRLCE as
Predictors . .. 91

4.13 Regression Analysis of Time 2 PDS, Using
Income, Occupation and WRLCE as
Predictors .. . 93

4.14 Regression Analysis of Time 2 HOS, Using
Social Support and WRLCE as Predictors 95

4.15 Regression Analysis of Time 2 PDS, Using
Social Support and WRLCE as Predictors 97

4.16 Regression Analysis of Time 2 HOS, Using
Combined Social Resource Variables and
WRLCE as Predictors . 98

4.17 Regression Analysis of Time 2 PDS, Using
Combined Social Resource Variables and
WRLCE as Predictors . .. 100

4.18 Cross-Sectional Regression Analysis of
Length of Time in Present Place of
Employment, Using PDS and Key Socio-
demographic and Social Resource Variables
as Predictors . ... 103

4.19 Cross-Sectional Regression Analysis of
Times Unemployed Using Key Sociodemographic
and Social Resource Variables as
Predictors . . 104



Figure Paae

2.1 Early Social Stress Model by Dohrenwend .. 11

2.2 Life Event: Sources, Adaptations and
Outcomes .. . 12

3.1 Different Possible Outcomes of HOS and
PDS Scores Over Time Controlling for
WRLCE . . ... 62

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



Dennis Earl Ghyst

March, 1980

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

This dissertation analyzes the influences of Time 1

psychiatric symptom and dysfunction scale scores on

intervening Work-Related Life Crisis Events (WRLCE) and

Time 2 symptom and dysfunction scores. The impact of

WRLCE on Time 2 symptoms and dysfunctions is also


Psychiatric symptomatology is assessed by the

Leighton Health Opinion Survey Scale (HOS) which was

revised from the original work of Macmillan, and the

Psychosocial Dysfunction Scale (PDS) developed by

Warheit. The major work behavior variable, WRLCE, is

derived from a weighted life crisis event measure

developed by Paykel which focuses on average degree of

"upset"expected for various life experiences. Standard

demographic and coping-resource controls are employed

throughout the analysis.

A review of literature concerned with social stress

and the relationship between mental health and work

behavior concludes with the following three general


Hypothesis 1: The level of prior work events (WRLCE)
is directly related to subsequent
psychiatric symptom (HOS) and dysfunction
(PDS) score levels.

Hypothesis 2: Those with high prior symptom and
dysfunction scores will have a larger
number of subsequent WRLCE than those
with low prior scores.

Hypothesis 3: The occurrence of WRLCE will produce
a greater increase in symptom and
dysfunction scores for those with few
resources than for those with many

These hypotheses are tested in a longitudinal design of

full and part-time workers (N=267) taken from an original

epidemiological survey (N=1645) in 1970 and its follow-up

in 1973. Multiple regression tests provide the primary

focus of data analysis.

The main findings of this study are as follows:

(1) Symptoms and dysfunctions maintain high

stability over time. By comparison, there is a low

association between antecedent WRLCE and Time 2

psychiatric scale scores, with less than 3% of the

variance explained. This provides only tentative

support for Hypothesis 1.

(2) Time 1 symptoms and dysfunctions have a strong

and highly significant impact on WRLCE. This strongly

supports Hypothesis 2.

(3) Time 1 dysfunctions better predict WRLCE than

do Time 1 symptoms, corroborating the efficacy of PDS

as a measure of work dysfunction.

(4) The social resource variables, occupational

level and income, are highly significant predictors of

time 2 symptoms and dysfunctions. This supports

Hypothesis 3.

(5) The weak association of social support with

Time 2 symptoms and dysfunctions, when controlling for

key variables, qualifies and limits Hypothesis 3.

Thus, this study confirms the influence of Time 1

psychiatric symptom and dysfunction scale scores on

intervening WRLCE and Time 2 symptom and dysfunction

scores. While the impact of WRLCE on Time 2 symptoms

and dysfunctions is given only limited support, the

mitigating effects of occupational level and income

variables on Time 2 scale scores are supported.

Finally, several ancillary measures of work-

related behavior are introduced at the end of this

dissertation. They are exploratory and useful

primarily in raising issues for future research.




A number of etiologic models have been developed to

explain the origins of mental illness. Among the primary

ones discussed in the literature are biogenetic (Slater

and Cowie, 1971; Myerson, 1976), biochemical (Perris, 1966;

Segal et al., 1976), psychosocial stress (Cannon, 1929;

Selye, 1955, 1956), social labeling (Scheff, 1966; Lemert,

1967), and learning models (Mowrer, 1960; Estes, 1970).

Falling within a social framework perspective are the

stress, labeling and learning models.

The present research focuses primarily on the stress

model. This model is most often employed to explain the

impact of events arising in the individual's environment

on physical and/or mental health. Consonant with this,

there is a large body of research supporting the view

that some mental disorders are the outcomes of events

which arise in the social environment (e.g., Holmes and

Rahe, 1967; Paykel, 1971, 1972). In contrast, a number

of studies (Murphy, 1976; Holzer, 1977; Warheit, 1979;

Myers et ai., 1972) report a great amount of stabil-

ity of symptomatology cross-culturally despite events.

This finding suggests that one's mental health influences

the occurrence of life events. It is therefore necessary

to examine the degree to which symptomatology may precipi-

tate life crisis events if one is to understand the rela-

tionship between life events, stress and mental disorders.

The stress literature suggests that life crisis

events (LCEs) produce increases in psychiatric symptom and

dysfunction levels. It is also consistent with this

literature that symptoms produce life events and, specifi-

cally, work problems. The research reported in this

dissertation is designed to examine the relationships be-

tween work-related life crisis events (WRLCEs) and

measures of psychiatric symptomatology and psychosocial

dysfunction. Related studies have shown that coping

resources such as income and family support networks help

to mitigate the deleterious relationship between events

and symptoms (Cassel, 1976; Cobb, 1976; Dean and Lin,

1977; Rabkin and Streuning, 1976).

Therefore, the present research proposes to answer

the following questions:

(1) To what extent do work-related events produce

changes in psychiatric symptoms and/or psychosocial


(2) To what extent do psychiatric symptoms and psycho-

social dysfunction produce work-related events?

(3) To what extent are work-related events inter-

related with prior psychiatric symptoms and psycho-

social dysfunction?

(4) And, to what extent do coping resources mitigate

the impact of work events?

The first two questions have often been treated as

alternatives in the literature. But they are not neces-

sarily exclusive; both may occur simultaneously.

Regarding the fourth question, on-going social and

emotional coping mechanisms have been demonstrated to

influence the relationship between stressors and mental

health. For example, an unskilled worker with relatively

few economic and familial resources to mediate the

effects of events such as illness or unemployment is more

vulnerable to their deleterious consequences. Because of

such considerations, this study will investigate the

impact of work crisis events on demographic groLps which

appear to have different levels of coping-adaptation


A longitudinal design was used for this research

based on an epidemiological survey and a follow-up study.

This has the advantage of providing a measure of psychiat-

ric symptomatology which is prior to and therefore

independentt of the current work situation. More specifi-

cally, this dissertation analyzes the influences of

Time 1 (T,) psychiatric symptom and dysfunction scores on

intervening life events and Time 2 (T2) symptom and dysfunc-

tion scores. Psychiatric symptomatology is assessed by the

Leighton Health Opinion Survey Scale (HOS) which was revised

from the original work of Macmillan (1957), and the Psycho-

social Dysfunction (PDS) measure developed by Warheit et al.

(1975a). The major work behavior variable, Work-Related

Life Crisis Events (WRLCE), is derived from a weighted life

crisis event measure developed by Paykel (1971) which

focuses on average degree of "upset" expected for various

life experiences. Standard demographic and coping-resource

controls are employed throughout the analysis. Hypotheses

testing the relationships between WRLCEs, resources, stress

and psychiatric symptoms and dysfunctions are introduced

at the end of Chapter 2 and discussed in the Chapter 3

plan of analysis.

Chapter 2 reviews the literature relating work to

psychiatric symptomatology, beginning with an overview of

the different theoretical approaches used to explain the

etiology of mental disorders. Then, the stress model is

presented and life crisis event literature reviewed. The

latter includes an exposition of the effect of resource

variables like social support networks and Socioeconomic

Status (SES), a composite measure of income, occupation

and education (See Nam and Powers, 1968). The core of the

chapter is comprised of a review of literature reporting

the relationship between work and mental symptomatology.

The chapter ends with a brief delineation of the general

hypotheses guiding this research.

Chapter 3 describes the methodology of this research.

First, the background of the Florida Health Study and its

three-year follow-up is outlined. The major variables

included in the present study are then described in detail.

These include the general psychoneuroticism scale (HOS),

psychosocial dysfunction scale (PDS), and work-related

life crisis event scale (WRLCE). This is followed by a

description of key demographic and resource controls. In

the plan of analysis, the hypotheses are formally intro-

duced and the application of appropriate statistical tests is

discussed. The chapter concludes with the presentation

of ancillary measures of work-related behavior and tech-

niques for their analysis.

Chapter 4 begins with a description of the distribu-

tion of main variables. A detailed analysis of each

hypothesis provides the core of this chapter. Alternative

analytic approaches for testing key questions of this

research are employed. Statistical tests appropriate for

crosstabular data presentation and multiple regression

analysis define the presentation of findings. Both

longitudinal and crosssectional data are examined.

In Chapter 5 a summary of the main findings is given

and examined for theoretical relevance. This dissertation

concludes with suggestions for future research.



The purposes of this chapter include the presentation

of an overview of the different theoretical approaches

used to explain the etiology of mental disorders. Then the

stress models are outlined and the life crisis event

literature is reviewed. The latter includes a discussion

of the effect of "resource" variables such as socioeconomic

status (SES) and social support networks. The remainder

of this chapter consists of a review of literature which

describes the relationships between work and mental


Toward a Working Definition of Mental Disorders

There are many divergent theoretical conceptions of

mental health and mental disorders. Researchers with a

biogenetic perspective (Slater and Cowie, 1971; Myerson,

1976), for example, maintain that mental health/illness

is a product of biological functions and dysfunctions

which are genetically transmitted. Other researchers,

however, claim that biochemistry, influenced by genes,

cognition, or environmental influences, is the proper

etiologic focus of mental illness (Perris, 1966; Segal

et al., 1976). The recent discovery that the body produces

endogeneous opiates (Hughes et al., 1975) has received

much attention in the popular literature and encouraged

hope that a new generation of highly effective, safe psycho-

tropic drugs is now on the horizon. Simultaneously, this

finding adds credence to the arguments posited by psycho-


Social models of etiology include labeling, learning,

and psychosocial stress models. Labeling theory (Scheff,

1966; Lemert, 1967) emphasizes the social interactional

processes whereby individuals are labeled mentally ill by

others. Such labeled individuals are described as commonly

internalizing and then acting out their socially defined

roles. Learning theory (Mowrer, 1960; Estes, 1970), by

contrast, examines the socioenvironmental processes which

are postulated to precipitate mental disorders. Psycho-

analytic and other developmental perspectives which

emphasize the social genesis of neurotic and psychotic

patterns of behavior exemplify this approach (Freud, 1949;

Adler, 1929; Jung, 1966). Finally, as was stated in the

Introduction, the psychosocial stress model (Cannon, 1929;

Selye, 1955, 1956) is most often employed to explain the

impact of events arising in the individual's environment

on physical and/or mental health. This model, which is

the framework for the present research, will be more fully

discussed in the following section.

The stress literature is for the most part character-

ized by the consideration of the negative dimensions of

mental health (although this is less the case for work-

related research). This is because the thrust of social

scientific research is generally towards amelioration of

"problem areas" of human existence rather than definition

of what is "normal" or "healthy." Nonetheless, there have

been some notable attempts to define the positive dimensions

of mental health. Jahoda (1958), for example, has develop-

ed a set of criteria for positive mental health based on

such things as positive self concept and self-actualization.

Similarly, Vaillant (1977) in Adaptation to Life examines

those features of positive adaptation which give some

individuals a decisive edge over others in dealing with

life's inevitable crises. For purposes of this research,

however, the negative dimensions of mental health provide

the primary focus. More specifically, mental health/ill-

ness is defined operationally by scores on two different

psychiatric scales: Leighton Health Opinion Survey (HOS)

and the Psychosocial Dysfunction Scale (PDS). This

approach represents a statistical normative definition of

mental health. While the definition of mental health/ill-

ness used in this research has definite theoretical

limitations, it is a useful one in that it avoids the

issues of diagnosis and etiology which cannot be resolved

at present given the current state of the social

and medical sciences. The statistical-normative approach,

as the name suggests, focuses on the distribution of

symptoms and dysfunctions among various social and demo-

graphic groups.

Stress Literature

The earliest pioneers in the field of stress research

(e.g., Cannon, 1929) researched the somatic aspects of

emotion. Selye (1955, 1956) both coined the word "stress"

as it relates to health and illness and did much to make

it a popular and respected topic of research in the

scientific community.

Selye'ssingle most important theoretical contribution

probably has been the General Adaptation Syndrome (GAS)

(Selye, 1956). This paradigm proposes that the initial

response to any kind of stressor, defined by Selye as any-

thing that produces stress, is alarm. The body then

quickly mobilizes its defenses (whether the threat is real

or imagined makes no difference) and if the threat recedes,

stability is quickly reestablished. If, however, the

pressure continues for an extended period or is of over-

whelming intensity at any given point in time, then the

defense mechanisms of the body are overcome and physical

deterioration results. It is of particular interest for

purposes of the present research that the GAS syndrome

concept can also be used to define the process whereby

stress precipitates behavioral disorders or psychiatric


Since the development of GAS, B.P. Dohrenwend (1961)

and others have further developed this model to include

coping and adaptation variables which determine the degree

of resistance one can muster against any stressor or

combination of stressors. Figure 2.1 (Dohrenwend, 1961) is

a simple representation of such a paradigm. But Warheit

(1979) has described this paradigm as a closed model with

inherent theoretical and analytical weaknesses. He

suggests the following model represented in Figure 2.2 as

an alternative. The advantages of the second model include

the fact that it does not assume unidirectional causality,

as does the first model, and, further, it allows for the

real life complexity of reciprocal interaction between

individual, cultural, and socioenvironmental variables.

Over time these variables are considered at the levels of

events, adaptive-nonadaptive screens (coping variables),

and stress outcomes. Furthermore, the author points out

that "resources" include such diverse things as individual

genetic makeup, social support networks, institutional

helping resources, and culturally based beliefs, values

and symbols.

Hence, an individual lacking adequate resources who

experiences intense life crisis events or debilitating

on-going stressors is likely to respond maladaptively.

Life Coping illness
Time 1 State --- Events --- Adaptation
Processes Success


Figure 2.1 Early Social Stress Model by Dohrenwend (1961)

C0 Z

LU 3< C

00- 2:2: -
= o-
UU -- j CLL.

(~ a mm jC) a-

- c /' z -, 2:- -
M 0) a: C
i- / -.O O
c -
C)) L) L. L


.- .J- L-e

LU -L L <1
U LU 1-- OC2O :


.. +cc -,

-C< C




00 C V) LU





30N3nbL3S 3WI1

Once a maladaptive response has been made, a stubborn

pattern of physical or mental symptomatology leading to

even greater deterioration will generally result if the

stressors are not removed or ameliorated. It should be

noted at this point that the present study is based on

longitudinal data and therefore examines such long-term

patterns of response to stress. In contrast, most stress-

related research is cross-sectional and tends to obscure

stress-related developmental factors.

McGrath (1970) viewed stress as resulting from either

undercapability or overcapability in the face of demands.

This is consonant with Selye's notion that some level of

stress is necessary for biological survival and that stress

can be either constructive or destructive depending upon

one's circumstances and corresponding abilities.

Mechanic (1967, 1970) takes a divergent view in

criticizing the "passive" stress model which he believes

characterizes most of the literature. He proposes that

greater attention be given to successful techniques for

mobilization of effort against various stressors. He

underscores the importance of "locating those aspects of

approaches and behavioral repertoires that lead to crises

and those that make (any) situation only an occasion for

further progress and mastery" (Mechanic, 1970:122). While

Mechanic's perspective has potential in sharpening or

augmenting clinical approaches which help individuals to

deal with stress, it is for the most part a digression

from the mainstream of research which deals with the

continuity of stress reaction patterns over time.

McQuade (1972) hypothesized that the inability to

cope with stress is the result of genetic and early socio-

environmental factors that diminish the capacity to adapt

to stressful events. In a similar manner, Hinkle and Uolff (1958)

note that "over time a small proportion of the subjects

have a majority of the illnesses, even though they have

not had a greater number of objectively determined (stress-

ful) experiences .. They have, however, perceived their

lives as more demanding, deprived, conflict ridden, and

threatening than their peers" (Hinkle-and Wolff, 1958).

Furthermore, Hinkle and 'olff question the utility of the stress

concept as applied to single event or "specificity"

explanations of physical and mental breakdown for two

reasons: the lack of uniform meaning and impact of crisis

events for different individuals, and the almost invariably

cumulative nature of stress events leading to illness

over time.

Similarly, no attempt is made in this research to

link any specific stressor with specific symptomatology.

Stressful Life Events

Life events research has its roots in Meyer's (1951)

life chart, a medical biographical device demonstrating

the relationship of biographical, psychological and socio-

logical phenomena to health in man. As early as 1949 the

life chart was employed in studying the nature and number

of life events of 5,000 patients in Harold Wolff's labor-

atory at Cornell University. Within the following decades,

Hinkle (1958, 1974) was pioneering in longitudinal life

crisis event (LCE) studies. Holmes and Rahe (1967), in

turn, were the first to quantify events and create LCE

scales. The degree of change in one's physical or mental

health was their sole criterion for determining the impact

of life crisis events and they made no distinction between

types of events. Most of the recent work in this area does

make such distinctions, however. Holzer (1977) in his

summary of selected classifications of life events appear-

ing in the literature notes the following dimensions of

research: desirable vs. undesirable events, objective vs.

subjective occurrence, gain vs. loss, entrances vs. exits,

and degree of upset.

According to Dohrenwend et al. (1978:207) the common

denominator of the many definitions of LCEs can be stated

as "objective occurrences of sufficient magnitude to bring

about change in the usual activities of those who experi-

ence them." For purposes of the present research, an .LCE

is defined as a life experience (such as being fired from

one's job) that is focused in time and that involves some

kind of change in role or status potentially disruptive
for tih individual (Paykel et al., 1971; Paykel and Uhlenhuth, 1972).

Most all LCEs examined in the literature have a direct

or indirect social origin. It is less clear, however, when

the impact of a life event is felt. In the case of being

laid-off from a job, does the event "occur" for the

individual at the time the layoff is announced, on the day

work ends, or at some point afterwards? (Cobb et al., 1966.)

In addition, by what mechanisms does the event affect

psychological symptoms? And why should some individuals

tend to experience an increase in psychiatric symptomatology

while others somatize stressful experiences? Another

salient issue involves the related topic of chronic stress

situations. For example, one might live with the threat

of losing his job for many years and never be fired. Or

an individual might live with all the pressure and dis-

advantage of being poor throughout life without ever

experiencing the crisis of bankruptcy. The more carefully

planned future research in this area should take such

issues into account.

In one of the few longitudinal studies of LCEs done

thus far Myers et al. (1972) collected interview data from

respondents on events and psychiatric symptoms at two

points, spaced two years apart. The findings showed that

an increase in LCEs over time was related to a worsening

of psychiatric symptoms. Still, it was not possible to

conclusively determine from the data whether a change in

events preceded a change in symptoms or whether changes

in symptoms more commonly preceded a change in events.

Another study suggesting that LCEs influence symptoms is

that of Rahe et al. (1970) which employed a prospective

design sampling of U.S. servicemen aboard ship who were

surveyed before they began a cruise. Events occurring in

a six-month period before the start of the cruise were

associated with a higher incidence of illness while aboard

ship, although the illnesses reported tended to be minor

physical disorders.

Until very recently, studies in the area of LCEs

seldom addressed the issue of whether high symptomatology

might precipitate events. In the past several years,

however, there has been an accumulation of literature

discounting or qualifying the impact of LCEs on mental

health. Dressler et al. (1976) make two primary points

in a study of the effects of LCEs on mental patients.

First, there is a wide variation in individual response

to LCEs which is obscured in most studies. Rabkin and

Streuning (1976) concur on this point in their study.

Also related to this finding are those of Paykel et al.

(1971) and Hurst et al. (1978) which indicate that ratings

(weights) assigned to LCEs differ between subjects who

have experienced given events and those who have not.

Second, Dressler et al. (1976) conclude that the incidence

of LCEs precipitated by psychological factors is also

commonly ignored or underestimated in the literature.

Their study revealed that:

(the patients) had accumulated a series of un-
resolved life stress events which occurred with
high frequency, prolonged duration, and high
emotional intensity resulting in an inner sense
of emotional fragility.. .the precipitating
event constituted a 'last straw' (which) assumed
personal importance because of its symbolic
meaning to the person, often reflecting an
important underlying life theme with roots in
early life experiences such as a sense of being
abandoned, etc. (Furthermore), it was dis-
covered that life events are often caused by
people rather than just happening to them...
particularly among those with a history of
psychological fragility or psychopathology.
(Dressler et al., 1976:557)

In their review of LCE literature, Rabkin and Streuning

(1976) find substantial support for the conclusions in

Dressler et al. (1976). In addition, they point to certain

methodological weaknesses commonly found in LCE studies

which considerably qualify the reported association between

events and symptomatology. For example, the large sample

size typically found in LCE research results in even very

small correlations passing the test of statistical signif-

icance. Also, correlation coefficients, when reported,

are typically below .30, indicating that life events may

account for only about 9% of the explained variance in


Cadoret et al. (1972) and Hudgens (1974) in their

studies of LCEs and psychiatric depression report .that:

while a clustering of life changes often precipitates

hospitalization, symptoms are relatively stable over time.

They speculate that it may be that life changes lead people

to seek help for long-standing psychiatric problems.

Tessler et al. (1976) in a study of out-patient medical

facilities also found that stress tends to increase

utilization of a medical facility rather than to increase


Hence, in attempting to understand the impact of LCEs

a number of questions should be considered. What sort of

crisis event has taken place and what symbolic meaning

does it hold for the individual involved? Do predisposing

psychological factors account for the precipitating of

certain events like accidents and acute illnesses rather

than the reverse? Also, what percentage of the variance

of psychiatric symptomatology do events actually account

for? And most importantly, do events alter long-range

symptomatology or is their effect just temporary?

The following section on coping-adaptation variables

makes inquiry regarding how factors such as social support

networks and financial resources are related to events and

their impact. Whether SES, racial, age, sex or marital

status variables amplify or reduce the impact of events

is involved in this question.

LCEs, Coping Variables and Impairment

The literature investigating the influence of coping

and adaptation variables on the relationship between LCEs

and psychiatric symptoms is relatively new. Nonetheless,

it has provided impressive evidence that such factors sig-

nificantly ameliorate the impact of LCEs. For example, a

study by Gersten et al. (1977) revealed that controlling

for life processes (resources), "events made no minimally

meaningful contribution to the prediction of any disturbed

behavior" in a sample of adolescent subjects for whom only

events outside the control of the subjects were examined.

The authors conclude ". .in other words, earlier disturb-

ance would be associated with later events to the same degree

as those events are associated with later disturbance"

(Gersten et al., 1977:229). Factors described by them as

contributing to the continuity of individual response over

time were social supports, differences in perception of

undesirability of events, differences in coping and

defensive styles, and differential emotional insulation

to events.

Social Support Networks

A number of recent articles have stated that social

support in many cases protects the individual against the

deleterious effects of stress (Cassel, 1976; Cobb, 1976;

Dean and Linn, 1977; Rabkin and Streuning, 1976; Kaplan,

Cassel and Gore, 1977). Furthermore, Zubin (1978) states

that 37 of the 43 events in the Holmes and Rahe (1967)

checklist involve a reduction of the social network relation-

ship of the affected individual. Consequently, the fact

that the majority of LCEs severely disrupt or alter existing

social support network relationships might well account for

the relationship between LCEs and symptomatology. Events,

then, involve in most cases a temporary or even permanent

diminution of social support. But it should be noted that

the same event (e.g., death of a parent) will have widely

different impacts on individuals, depending on the importance

of the lost or weakened relationship in one's social support


Eaton (1978) employs a panel regression technique in

his reanalysis of the Myers et al. (1975) data and finds

that following LCE, psychiatric deterioration is much more

likely for the unmarried than for the married. Kasl et al.

(1975) and Pearlin and Johnson (1977) report similar findings.

But in the Pearlin and Johnson study, multiple regression

analysis revealed that 69% of the original association

between marital status and non-psychotic depression is

accounted for by economic deprivation, isolation from social

networks, and frustrated or difficult parental responsibil-

ities. Economic deprivation was found to be the strongest

factor producing depression. Because the unmarried are

more often exposed to all these conditions, they are more

vulnerable to their effects. Social selection (downward

social mobility hypothesized to result from genetic and/or

social developmental disadvantage which manifests in high

levels of psychiatric symptomatology) is at most a marginal

factor in this relationship.

According to Mueller (1978), social support network

interpretations are generally consistent with the social

causation hypothesis (high levels of psychiatric symptoma-

tology described as resulting from social-structural re-

lated deprivation). However, with more serious mental

disorders like schizophrenia the literature suggests that

LCEs like those involving a weakening of social support

networks, merely trigger rather than form symptoms (Brown

and Birley, 1968; Brown et al., 1973). These latter studies,

then, are more consonant with the social selection

hypothesis in their emphasis on the importance of long-

range developmental factors in etiology.

Qualifying this relationship is the finding of

Brown et al. (1972) that there is a significant association

between dysfunctional social support networks and psychiatric

outcomes for schizophrenics. It was revealed that

psychiatric patients living with relatives expressing

high emotion at the time of admission were not as likely

to relapse within a nine month period following their re-

lease if they were able to avoid close contact with the

family and receive regular medication.

It is interesting to note that the social support net-

works of neurotics and psychotics are reported to differ

from those of the unimpaired in the following ways:

The primary networks of neurotics were relatively
small (about 10 to 12 persons), often including
persons who were dead or lived far away. The
density of the interconnectedness of neurotics'
networks tended to be low in comparison to the
normative sample. it is as if the neurotic
person is at the hub of a wheel, with individual
relationships like spokes that have no inter-
relationship. Also neurotics more often rated
interpersonal relationships with network members
negatively than did normal subjects. The
psychotic group, by comparison, had very small
primary networks (four or five persons) that
consisted mostly of family members.
(Pattison et al., 1975:1249)

Since social support is a multidimensional phenomenon,

such factors as source of support, kind of support given

(whether emotional or financial), and intensity of rela-

tionship (whether confiding or non-confiding) must be

taken into account. Brown et al. (1975), for example, found

that for women who had experienced "a severe event or

major difficulty" the consequent development of (non-

psychotic) depression was highly related to the absence

of an intimate, confiding relationship with husband or

boyfriend. For women experiencing stressful events, 38%

without such a confiding relationship developed depressive

symptomatology, while only 4% of the women with a close

male confidant did.

In a longitudinal study of life events, coping-adaptation

resources and depressive symptomatology, Warheit (1979)

discovered that Time 1 depression scores were the best

predictor of Time 2 scores taken three years later. The

data indicated that life event losses are cushioned by the

availability of interpersonal, familial and other resources.

It was also revealed that low SES individuals are especially

vulnerable to LCE because they have fewer resources and

experience more LCEs-which can be defined as losses.

Finally, there are a number of studies which demonstrate

a similar effect of coping-adaptation variables on various

physical illnesses and problems. In a frequently cited

study of pregnancy, Nuckolls et al. (1972) found that women

with high scores on a social support measure who experienced

high levels of change both before and during pregnancy had

one-third the rate of complications of women with low

social support scores. Social support, however, was not

related to complications for women with low life change

(low LCEs).

Indexes of familial and non-familial social support

networks will be used in this study. It is anticipated

that high levels of social support may lessen the impact

of work-related LCEs (WRLCE) on psychiatric symptoms and



According to B.P. and B.S. Dohrenwend (1969) the

highest overall rates of psychiatric disorder have been

found in the lowest social class in 28 out of 33 studies.

Furthermore, this relationship applies to schizophrenia

(in five out of seven studies) and personality disorders

(in eleven out of fourteen studies), although not to

neurosis or manic depressive psychosis. The research

done so far has provided no definitive answer to the

question of whether "social causation" or "social selection"

is the better hypothesis in explaining this relationship.

It is noteworthy, however, that Dunham, a pioneer in

epidemiological research on schizophrenia, has moved from

support of social causation to a position of qualified

support for social selection in his more recent work

(1965, 1966).

In a landmark study Dunham et al. (1966) reanalyzed

the Hollingshead and Redlich data (1958) employing a

different methodology. Their findings revealed that

schizophrenics are at a distinct disadvantage when entering

the job market compared with their fathers despite greater

educational achievement in their developmental years.

This results in downward mobility for a high percentage of

such individuals. While seeking employment, the

schizophrenic's "traits, attitudes, mannerisms and verbal

reactions become all too obvious and operate against

the securing of (or advancing in) a position. ." (Dunham

et al., 1966:225).

Langner (1963) found that the greater the stress, the

higher the mental health risk among lower social class

groups compared with other SES categories. Many other

researchers, including Warheit (1979), have had

similar findings. B.S. Dohrenwend (1973), for example,

discovered that lower class individuals experience more

unpleasant stressful life events which have a higher re-

adjustment or change impact than do persons higher in the

social system. In addition, it has been substantiated

that lower class individuals have more fragile social

support systems, disadvantaging them even further. More

specifically, the lower class is characterized by high

proportion of broken homes, ethnic minority status, unstable

employment, residential mobility and low participation in

community activities (Mueller et al., 1978).

Given all these factors, it is anticipated that work

events may have a stronger impact on symptomatology for

lower class workers than for middle and upper class workers.

Occupational level is by itself an important consideration

inasmuch as level of skill is directly associated with

social prestige, income, and one's related ability to

manipulate his environment and choose among alternatives.

Therefore, it is expected that stressful events will

probably be the most severe for unskilled blue collar

workers and low-level white collar workers when compared

with skilled blue collar workers and those in the profes-

sions and management.

It should be noted that SES will be broken down into

its component parts in the analysis of the present study

for reasons that will be explained in Chapter 3. Income

and an occupational-level index will constitute the primary

SES-related measures used in examining the relationships

described above.


Race is another factor that is closely tied into work

expectations and demands. Pettigrew (1964) discusses the

burdensome role of the Black in the U.S. which has included

many forms of invidious discrimination from employers.

The high unemployment statistics for Blacks in 1979

demonstrate how little things have changed in the past

fifteen years for the lower class Black worker.

Warheit et al. (1975b) found that Blacks have signif-

icantly higher scores than Whites on five scales of psycho-

pathology. But when their data are controlled for the

variables SES, sex, and age, the differences between Whites

and Blacks are reduced to non-significance (except on the

phobia scale). In every instance, low SES was the most

powerful predictor of high psychiatric symptomatology.

It is expected, then, that race, when controlled for other

major demographic variables, will have little impact on

the relationship between psychiatric symptomatology and

work in the present research.

Parenthetically, race (and ethnicity) can also be con-

ceived in terms of social support. Holmes (1975), for example,

found that controlling for social class, the highest rate

of TB in his Seattle survey occurred in those people who

lived in neighborhoods where they were of distinct, un-

accepted minority status. Similarly, it has been found

that Blacks living in predominantly White areas have

higher social deviance rates than Blacks living in mostly

Black areas with the converse also holding true (Faris

and Dunham, 1939; Klee et al., 1967).

Sex and Age

Although the situation has changed considerably in

recent years, occupation and occupational achievement are

still more important for American males than females in

terms of role expectations and demands. Because the sample

used in the present research includes a relatively small

number of female workers, a number of interesting

questions regarding such issues will have to be reserved

for future research.

A number of studies demonstrate that age is an

important variable in work satisfaction (Meltzer, 1963;

Spreitzer and Snyder, 1974). Older workers seem to be

more satisfied with their work. This is perhaps because

of lower aspirations, generational differences in terms of

exposure to different historical influences, etc. It is

not anticipated, however, that age will be significant

in articulating the relationship of psychiatric

symptomatology and work.

Work and Mental Health

"Among all those activities peculiar to humans, work

probably defines man with the greatest certainty" (O'Toole,

Work in America, 1973:1). But if the appelation "homo

faber" applies to man, what is the proper definition of

work? Paid employment is the most popular conception of

work. Because there is no more important status than

occupation, particularly in the context of American culture,

pay and social worth are commonly equated. Occupational

prestige, though less important than pay, must also be

included in this equation of social worth. Increasingly,

these factors are coming to apply to both males and females

as women augment their ranks in the professions.

There are a number of alternative definitions of work

that do not involve remuneration as an inevitable con-

comitant. For example, work has been defined as "any

activity that produces something of value for other people"

(O'Toole, 1973:3). Still more elegant is Gross' definition:

"Persons are said to be doing work when they have some

goal in mind that they are seeking to obtain in a disci-

plined manner" (Gross, 1970:62). These broader

definitions are not as useful for the purposes of this

research as is the concept of paid employment. Nonetheless,

work serves other social functions besides the economic.

One's place of work, for example, is a setting in which to

meet significant others, to achieve social acceptance or

marginality, and to affirm or reject previously held

personal values and norms. Furthermore, Gross contends

that our culture is still imbued with the Puritan Ethic

which "... .keeps people working by placing a negative

value on idleness through requiring that the enjoyment

of non-work be accompanied by an explanation"

Gross, 1970:65).

According to sociological theory, social roles have

a transforming effect on human nature, sometimes enhancing

an individual's sense of well-being and personal efficacy

while at other times undercutting one's perception of personal

worth and even one's basic health. Because of the unparal-

leled importance of the work role in American and other

modern industrial societies, a convincing theoretical case

can be made for work-related LCEs (WRLCE) having a

significant effect on psychiatric symptoms and functions.

The experience of unemployment, for example, may contribute

not only to a debilitating financial insecurity, but also

undermine one's basic sense of identity, self-esteem and

emotional health.

In order to put the above in proper perspective, it

should be noted that many people turn to other activities

(e.g., hobbies and sports) and institutions (e.g., the

family and religion) for their primary psychological

rewards. Also, there is some support in the popular

literature for the notion that this has been a "decade of

narcissism" in which the importance of the work role has

been undermined while individual physical and "spiritual"

development has been emphasized. A more important

qualification on the effects of work role is that there

seems to be significant and relatively constant individual

variations in resilience toward LCEs which perhaps apply.,

as well to WRLCE. Whether WRLCE more often precede

increased psychiatric symptoms and functions, or pre-

disposing psychiatric impairment more often precedes

WRLCE is difficult to determine. This crucial issue

will be carefully examined from both sides in the

remainder of this chapter.

Finally, it should be made clear that outside the

domain of the present study are the broad areas of work

stress, e.g., work role conflict, ambiguity and overload,

extent of responsibility for others, and repetitiveness

and pace of work. While the data do not permit a de-

tailed analysis of such topics, they do allow for a

diachronic study of the relationship of psychiatric

symptomatology to specific WRLCE.

A closer examination of research supporting the

view that work experience significantly affects mental

health will now be presented.

The Effects of Work on Mental Health

The literature on the mental health consequences of

work is relatively limited. The unemployment studies of

the late 1930's and early 1940's spawned by social and

political concern for the consequences of the Great

Depression, represent the earliest work of contemporary

interest in the area. These early studies are charac-

terized by a number of critical methodological shortcomings:

they employ subjective measures, they do not distinguish

short-term and long-term effects, and they do not include

control groups (Kasl, 1973). In addition, there is the

question as to whether conditions relevant to work and

employment forty years ago still apply today.

Kornhauser's Mental Health and the Industrial Worker

(1965) is perhaps the most quoted study articulating the

effect of work on mental health. His conclusions include

the finding that routine production workers have less

satisfactory mental health than workers in more skilled

and varied jobs. As with many other studies of this kind,

he emphasizes self-esteem as a factor linking work and

mental health:

Our interpretation is that job conditions impinge
on working people's wants and expectations to
produce satisfactions and frustrations which in
turn give rise to favorable or unfavorable
perceptions of self-worth, opportunities for
self-development, and prospective gratification
of needs. ..the evidence as a whole accords
with the hypothesis that gratification and

deprivation experienced in work and manifested in
expressions of job satisfaction and dissatisfaction
constitute an important determinant of a worker's
mental health.
(Kornhauser, 1965:89)

Regarding this specific point, Kahn and French (1970)

disagree. Based on findings of the University of Michigan

Survey Research Center research which examined the effects

of work environment on health, they assert, that self-

esteem "is a more or less stable characteristic of the

person that conditions the effects of social environment

on performance" (Kahn and French, 1970:245).

Kornhauser (1965) qualifies the effect of work on

mental health in his finding that job feelings such as

work satisfaction constitute intervening variables. For

example, he found that contrary to earlier findings,

mechanically paced, repetitive work does not really affect

mental health except perhaps in a symbolic way when workers

equate being on such a job with failure.

Kornhauser notes a more general limitation to his

work in the following:

Mental health affects job attitudes at the same
time as it is affected by them; and both job
feelings and mental health may have common
causes, for example, lifelong personality
attributes of cheerfulness or negativism, etc.
(Kornhauser, 1965:18)

Unfortunately, he does not control for such factors in

his design.

There are other difficulties in generalizing

Kornhauser's findings, according to Roman (1969). Among

the problems cited: Kornhauser's data are cross-sectional,

making it impossible to show the extent to which there was

self-selection for unskilled and repetitive jobs; and he

assumes homogeneity of workers at different job skill

levels. Furthermore, he overlooks social class differences

within different groups and confounds his dependent

variable by considering level of job satisfaction as an

index of positive mental health. Roman is generally

critical of other work literature which asserts or implies

an unqualified causal linkage between work and mental health:

A major contaminant in such studies is socio-
economic status, which typically varies with
occupational level, and which contains
explanatory variables other than work-based
experiences. A second major design problem
(in such research) lies in the general
absence of longitudinal data.
(Roman, 1969:564-65)

Kasl (1973) takes a more positive position regarding

literature which links work and mental health. He adds

the qualification that although there is much evidence

connecting the two variables, a precise causal interpre-

tation of such studies is difficult. Also, Kasl, unlike

Roman, is receptive to definitions of positive mental

health and the related use of indexes of positive psycho-

logical functioning. He notes that studies taking this

approach indicate that workers in low-level jobs seem to

limit their aspirations and expectations, to have low job

involvement, and consequently little, if any,self-

actualization. According to Kasl:

Studies of mental health in low-level, blue
collar jobs suggest a coping process which
essentially represents a trade-off: in order
to maintain some tolerable level of well-being,
self-acutalization is 'sacrificed' by severe
curtailment of aspiration and expected
satisfactions from work. (Additional trade-offs,
such as well-being in work role for well-being
in other social roles such as parent, spouse,
or not appear to be operative).
It is tempting to try to label the trade-off
between well-being and self-actualization as
'good' or 'bad'. .. This, of course is impossible
short of possessing some calculus for weighing
well-being and self-actualization. .
(Kasl, 1973:515)

Kohn and Schooler (1973) make a stronger statement

for the effect of work on psychological functioning in

their research findings. Using a two stage "least

squares" technique for estimating reciprocal synchronic

causal effects of fifty separate occupational conditions

with psychological functioning, they found the more

determining variable to be work in twelve out of fifty

different dimensions.

,the specific links between particular
occupational conditions and particular facets
of psychological functioning suggest that
men's ways of coping with the realities of
their jobs are generalized to non-occupational
realities. (For example), men whose jobs
require intellectual flexibility came not
only to exercise their intellectual prowess

on the job but also to engage in intellectually
demanding leisure-time activities. (Yet) no-
where in these data is there evidence that men
turn their occupational frustrations loose on
the nonoccupational world or try to find
compensation in nonoccupational realities for
occupational lacks or grievances.
(Kohn and Schooler, 1973:117)

Their findings are theoretically compatible with neo-

Marxian structural determinism inasmuch as they maintain

that job demands are the product of the economic and social

systems which shape workers' values and perceptions. But

it should be understood that their delineation of "psycho-

logical functioning," which emphasizes dimensions of

positive mental health, largely ignores dimensions of

symptomatology investigated in the present and other

research. Also, their findings are questionable inasmuch

as they are not based on longitudinal data.

In a later longitudinal study (1978) by the same

researchers, support is given their earlier arguments

regarding the positive relationship between "substantive

complexity" of occupation and intellectual flexibility

off the job. They add, however, that their research has

not yet demonstrated that substantive complexity directly

affects values, self-conception, or social orientation.

In other words, while their findings are provocative,

they do not demonstrate anything more than one linkage

in a complex nexus of interrelated work and mental health


The preceding studies in this section emphasize positive

definitions of mental health. Some observers maintain

that concern with positive mental health dimensions like

self-actualization reveals a middle class bias which has

no place in blue collar worker studies. Yet it is difficult

to discount the importance of the fact that lower class,

low-skilled workers demonstrate relatively low job and life

satisfaction, and poor self-concept and self-evaluation.

Kasl (1973) asserts that much more must be learned before

noxious elements in low-skilled jobs can be identified.

It should be added that more needs to be learned about non-

work-related noxious dimensions of lower class life as

well. Negative family and community influences, for

example, need more investigation. Also, the efficacy of

"social selection" explanations for the higher incidence

and prevalence of psychiatric symptomatology in the lower

class require further study. A better understanding of

such factors is needed before the effects of work can be

put in perspective.

Qualifying and helping to clarify the relationship

between work and mental health are studies which examine

the effects of resource-adaptation variables. Gore (1978),

for example, in a longitudinal investigation of the physical

and mental health consequences of unemployment due to two

plant shutdowns discovered that there were no differences

between the "supported" (workers who received emotional

support from their wives) and "unsupported" with respect to

length of unemployment or to economic deprivation. However,

during the period of unemployment, the unsupported workers

demonstrated greater physical and mental symptomatology

than the supported ones. The unsupported more commonly

expressed self-blame for being unemployed and had a greater

sense of economic deprivation than their counterparts.

Gore suggests that one reason for the difference is that

those receiving social support are not dependent on

accomplishments for self-esteem while the unsupported must

depend primarily on instrumental accomplishments for their

feelings of self-worth.

Brown (1975) affirms both the importance of the work

role and social supports for individual mental health in

a study of British working women. Four factors were found

to increase the chances of developing a psychiatric disorder

following a severe LCE: loss of mother in childhood,

having three children under the age of sixteen, lack of an

intimate or confiding relationship with husband or boy-

friend, and a lack of full-time or part-time employment.

Seventy-nine percent of the women who had the first three

factors working against them and who were also unemployed

"became disturbed" after experiencing a major LCE. Only

14% of the employed women who were in the same situation

did so, however. The authors caution that their findings

only hint at such a possible relationship because it is

based on a very small sample. Furthermore, a number of

difficulties were pointed out in interpreting the role of

employment: having a job might serve a protective function

by improving economic circumstances, alleviating boredom,

bringing greater variety in social contacts or an enhanced

sense of social worth. Although some combination of all of

these factors is probably operative, the authors continue


We were particularly interested in a few of these
women in our sample who took up employment a few
weeks after the occurrence of a severe event, none
of whom developed psychiatric disturbance. Their
comments suggest that a sense of achievement might
be crucial.
(Brown et al., 1975:244).

Coburn (1975) sustains the linkage between work and

mental health in a study which conforms to McGrath's (1970)

stress model discussed earlier. He found that a poor fit

between individual capacities and work demands may lead to

mental and physical health problems. This is particularly

the case for overly complex work. Overly simple work

has only moderate psychological and mild physical effects.

It was also discovered that perceived incongruence has a

much greater effect on mental health than does objective

incongruence. It would seem, then, that his study includes

too many subjective variables to be fully valid. But Coburn

maintains that he has controlled for the tendency to view

the world, job and health in a similarly positive or negative


Finally, Gross (1970) makes some general criticisms of

research describing work as a major precipitant of stress

and mental illness. He claims that such studies fail to

take into account the many sources of stress-producing

individual-group conflict which are outside the work

organization. Furthermore, he contends that it is the

small group as exemplified by the family and community that

is the most tyrannical in imposing discipline on the

individual rather than large work organizations. Also,

Gross questions those studies which implicitly assume that

sucn factors as job anxiety and aggression are in every case

negative factors for job adjustment and related mental

health. Such factors, he claims, are not only inevitable

but necessary to motivate workers to complete tasks and,

indirectly, to feel challenged or fulfilled by their work.

In summation, there are a number of shortcomings to

studies examining the effects of work on mental health.

First, much of this segment of the literature deals

primarily with positive dimensions of mental health which

has resulted in a frequent blurring of the distinction

between work attitudes and mental health. Furthermore,

the general applicability of a number of positive dimensions

of mental health like self-actualization is questionable.

Also, it is significant that most of these studies fail

to control for antecedent symptomatology and many of them

give only superficial consideration, if any, to SES

background factors. In addition, it should be noted that

in every study discussed, a large measure of the variance

of the relationship of work and mental health can be

explained in terms of subjective individual qualities which

precede the work situation. Finally, the literature is

said to overlook sources of stress outside the work

organization and to underestimate possible positive

functions of stressful work experience for the individual.

On the other hand, this segment of the literature

has provided evidence that negative work environment can

at least to some degree affect self-esteem, a dimension

of positive mental health that cannot be ignored. Also,

the literature establishes that aspects of the work

environment, like complexity of work, have effects which

carry over into nonoccupational areas of life. One study

claims that a poor fit between individual capacities and

work may lead to mental and physical health problems.

Significantly, another study suggests that employment may

act as a prophylactic to prevent psychiatric disturbance

in subjects otherwise lacking in coping-adaptation resources.

Inasmuch as this body of literature is relatively new

and incomplete, its inadequacies obscure the potential

for progress in future research. Giving direction to

future research efforts are a number of innovative topical

and methodological suggestions. Among them, Kasl (1973)

proposes that there should be longitudinal studies of

various naturally occurring events like job promotion,

demotion, and layoff. This dissertation endeavors to

augment research efforts in this specific area.

Now for another view of the relationship between work

and mental health, a review of that literature giving

direct support to the hypothesis that predisposing mental

health factors established long before the onset of career

are largely the basis for poor work adjustment and WRLCE.

The Effects of Mental Health on Work

There is much in the literature to suggest that mental

symptoms and functions affect work behavior. While poor

mental health is not a necessary cause of poor work adjust-

ment, a significant relationship has been demonstrated in

many studies. This is particularly the case with schizo-

phrenia and other serious mental illnesses. Tiffany (1970)

points out that the most serious mental problems will

affect the occupational sphere. This relationship is

further supported in a study by Cole et al. (1966) who

found that schizophrenics perform poorly on the job in

their ability to adjust to new situations, in their judg-

ment, quality of work, personal appearance, and cooperative-

ness. Also, Hamburger and Hess (1970) report that job

difficulties were experienced by 60% of schizophrenic

patients in the five-year period preceding admission to

a mental hospital.

McEwen (1974) discusses the significance of other kinds

of predisposing psychological factors affecting the

adaptability of workers in different types of jobs.

Retardates, for example, tend to be found in jobs where

noise levels and demands for exactness, speed, and team

work were relatively hign. Psychiatric cases generally

held jobs where all these factors were lower. McEwen

hypothesizes that the difference is the result of retardates

having a lower arousal level than psychiatric cases, thus

being able to withstand more stress before their performance


The literature supporting the effect of predisposing

mental health factors on work experience is at least

implicitly developmental in theoretical perspective. There

are a number of studies which directly examine related

issues. Robins (1966), for example, in her comprehensive

longitudinal study of sociopathically disturbed children

concluded that they are much more likely to remain

disturbed and/or deviant as adults than are control

subjects. Parenthetically, Robins reports that sociopathic

children in adulthood had consistently higher rates of

unemployment, longer periods of unemployment, lower occupa-

tional status, more frequent job changes and lower income

than other subjects. Similarly, Waldron (1976) found

that early childhood neurosis bears a significant relation

to psychosocial adjustment in adulthood. In comparing

the mental health of a group of young adults who had been

given psychiatric treatment for neurosis in childhood with

a control group, it was found that more than 75% of the

former patients were at least "mildly ill" at follow-up

compared with only 15% of the control group.

Apparently, similar definitions and responses to

mental illness occur cross-culturally and therefore have

something to do with a process which is at least in part

genetic. Murphy (1976), for example, reports that parallel

patterns of behavior are labeled abnormal cross-culturally

in five geographically scattered non-Western societies.

According to Murphy:

Explicit labels for insanity exist in (all the
cultures studied). (In each case) the labels
refer to beliefs, feelings, and actions that
are thought to emanate from the mind or inner
state of an individual (which) are essentially
beyond his control. ..The afflictions bear
strong resemblance to what we call schizophrenia.
.Almosteverywhere a pattern composed of
hallucinations, delusions, disorientations,
and behavioral aberrations appears to identify
the idea of 'losing one's mind,' even though
the content of these manifestations is colored
by cultural beliefs. (Furthermore) the
different cultures (all) react to people they
define as mentally ill with a complex of
responses. ...including an ambivalent-appearing
mixture of care giving and social control.
These reactions are not greatly dissimilar
from those that occur in Western society. Nor
does the amount of mental illness seem to vary
greatly within or across the divisions of
Western and non-Western areas.
(Murphy, 1976:1027)

In contrast with the other research discussed in

this section, Cumming (1963) and Neff (1968) contend that

the ability to work is part of a more or less autonomous

area of personality that has its own developmental process.

Turner (1977), for example, in a study of the work patterns

of schizophrenics discovered that a substantial majority

of those judged mildly to moderately impaired were

employed in stable full-time jobs. Surprisingly, even a

small majority of those judged severely impaired were

employed full-time. However, the unemployment rate of

those ever hospitalized was twice that for the never

hospitalized group. It should be noted that Turner at no

point in his study denies the strong association between

work failure and schizophrenia. His investigation simply

seeks to define the nature and limits of this relationship.

In a later related study, Turner and Gartrell (1978)

employ multiple regression in an analysis of factors

related to outcome of schizophrenia as measured by time

spent in hospital. Work performance, defined as the

ability to hold a job, was significantly related to

indexes of severity of pathology and social competence.

Severity of pathology was judged by a panel of psychiatrists

who had no knowledge of the patient's social competence.

This approach is questionable, though, since degree of

social functioning is often an important formal considera-

tion in both the admission and discharge of patients.

In conclusion, all of the studies discussed in this

section underscore the relationship between antecedent

psychiatric symptomatology, which is described as having

roots in socialization and genetic endowment, and later

work adjustment. Turner's findings perhaps qualify this

relationship but by no means account for it.


The relationship between work and mental health has

been extensively investigated in the literature with

divergent methodologies and ambiguous findings. This

research endeavors to add to our knowledge of people with

mental problems by investigating the reciprocal effects

of psychiatric symptomatology and its related psychosocial

dysfunction with work-associated life events. Based on

the literature and theory presented in this chapter,

three general hypotheses have been developed to guide

the present research. Key terms will be operationalized

in the next chapter.

First, there is a body of literature which claims

that stressful work experience results in a higher incidence

of mental problems for workers. This supports the credi-

bility of the following hypothesis:

Hypothesis 1: The level of prior work events (WRLCE) is
directly related to subsequent psychiatric
symptom and dysfunction score levels.

Second, recent research reporting the stability of

symptomatology cross-culturally and over time underscores

the importance of examining the social outcomes of mental

illness. In the context of the present study, this

suggests the desirability of measuring the effects of

prior symptoms and functions on subsequent work behavior

as articulated in the following:

Hypothesis 2: Those with high prior symptom and dysfunction
scores will have a larger number of sub-
sequent WRLCE than those with low prior

Third, there is increasing evidence in the literature

that resource-adaptation variables considerably ameliorate

the impact of LCE. In the present research, this is

translated into an interest in the degree to which

different kinds of resource variables like SES and social

support networks buffer the impact of WRLCE. Hence, the

third and final general hypothesis:

Hypothesis 3: The occurrence of WRLCE will produce a
greater increase in symptom and dysfunction
scores for those with few resources than
for those with many resources.



This chapter begins with a review of the Florida Health

Study and its follow-up. The major variables included in

the present study are then described in detail. These

include the general psychoneuroticism scale (HOS), psycho-

social dysfunction scale (PDS), and work-related life

crisis events scale (WRLCE). This is followed by a

description of key demographic and resource controls. In

the plan of analysis, the hypotheses are formally intro-

duced and the application of appropriate statistical tests

discussed. The chapter concludes with the presentation

of ancillary measures of work-related behavior and

techniques for their analysis.

Background to Study

This dissertation is based on data secured as part of

a psychiatric epidemiological study conducted in Alachua

County, Florida. Data on social and medical variables

were obtained by means of a 317-item interview schedule

administered initially in 1970-71. The instrument

included questions on demographic data and social history,

items concerning familial and interpersonal relations,

life satisfactions, and indices concerning religion,

racial distance, anomie, social aspirations and change.

In addition, detailed physical and mental symptomatology

inventories and items concerning attitudes toward and

utilization of health services were encompassed in the

survey (Warheit et al., 1975; Schwab et al., 1979).

The adult, non-institutionalized population of the

county constituted the target for the study. A statis-

tical probability sample of residential electrical hook-

ups, augmented by sampling from county maps in some areas,

resulted in a master sample of 2,315 of the 31,115 house-

holds in the community. Then the Kish (1965) technique

for randomizing respondents within households was

employed. It has been suggested that with this tech-

nique the weighting of samples is sometimes necessary to

avoid underrepresentation of large households. The

survey did not require such weighting, however, since

preliminary analyses using both weighted and unweighted

samples revealed no major differences.

A pretest of 322 respondents was conducted in order

to refine the survey instrument. Of the remainder of

the master sample list, total non-response was 17.5%

with a refusal rate of 8.8%. This resulted in a total

of 1,645 interviews obtained.

Then, in 1973, a systematic stratified subsample of

1970 respondents was re-interviewed in a follow-up which

replicated most of the material obtained in the 1970 study.

The non-response rate was 13.4%; 517 interviews were ob-


Major Social-Psychiatric Variables

To help measure social-psychiatric impairment several

scales were developed. These included the Health Opinion

Survey (HOS) a measure of general psychoneuroticism,

general psychopathology, cognitive impairment, psycho-

social dysfunction, anxiety, depression and phobia. These

scales, drawn from approximately 100 items relating to

psychiatric symptom and symptom-related dysfunctions,

were designed to measure clusters of symptomatology

considered to be components of social psychiatric impair-

ment. There was no attempt to diagnose individual dis-

orders since such identification remains beyond the

capability of field survey procedures. This general

approach is consistent with the epidemiological objective

of the research which was to identify groups that would

most likely constitute the case loads of Community Mental

Health Centers.

It should be noted that preliminary validation of

these scales has been established with a number of patient

groups. Furthermore:

Supporting evidence of their validity is provided
by the following: (1) many of the items included
were drawn from prior psychiatric research; (2) the
items were examined by a panel of experts and their
content was judged to be appropriate; (3) factor
analytic procedures empirically confirm their
grouping into scales; and (4) the scales have an
acceptable level of consistency as measured by
Cronbach's Alpha (1951).
(Warheit et al., 1975b:245)

Two of the above scales are utilized in this research;

tnese are the HOS and psychosocial dysfunction.

The HOS Scale

Until the development of Macmillan's Health Opinion

Survey (1957), most psychiatric screening tests had been

created for the selection of individuals in various

institutional situations, e.g., the military and business.

MacMillan devised his test as part of an epidemiological

project in order to detect those adults in the general

population whose responses approximated those of psychi-

atric patients and differed from those of controls drawn

at random from a number of rural communities. More

specifically, his aim was to identify community adults

who would likely be identified by psychiatrists as psycho-

neurotic cases.

The test was constructed from a number of sources:

Its core was 15 questions from the Army's Neuro-
psychiatric Screening Adjunct with additional
questions reported to be useful neurotic dis-
criminators by Eysenck, Rimoldi, and others.
(Furthermore), local general practitioners were
consulted by the author in order to establish
criteria as to the suitability of the various
(Macmillan, 1957:328)

Out of the many potential test questions, twenty items were

chosen which distinguish hospitalized neurotics from the

community samples as a whole and subgroups within it.

Employing a weighted score based on discriminant function

analysis, Macmillan found that 25% of the community sample

(N=419) had scores in the case category compared with 92%

of the neurotic hospital sample (N=78).

Macmillan's index and modifications of it have achieved

widespread use in the literature (e.g., Leighton, 1963;

Spiro et al., 1972; Warheit et al., 1975). In one version,

Leighton (1965) replaced several items of the index to

make it more applicable to general populations. Research

for this dissertation employs this modified version.

Although no claim has been made for the use of the

HOS as a diagnostic tool, the validity of the index as an

epidemiological screening device for psychoneuroticism is

supported by a number of studies (Kuldau et al., 1976;

Moses et al, 1971; Spiro et al., 1972; Schwartz et al., 1973).

Kuldau et al. (1976), for example, found significant differences

for the HOS between four out of five general population

risk groups and hospitalized psychoneurotics (p>.05).

Only those community respondents in the highest risk

category could not be distinguished from patients. This

is not surprising since the highest risk group, by virtue

of the items defining risk, included individuals who had

prior treatment for psychiatric problems and/or who had

severe physical health problems. Analysis of the data used

in the present study has revealed that the HOS, in con-

formance with its intended use, taps psychoneuroticism

better than any other diagnostic category. The data also

demonstrate that the HOS clearly distinguishes between

neurotic and psychotic patients (p<.01), with neurotics

receiving significantly higher scores. Analysis of these

data and other studies also give evidence that the HOS

successfully identifies psychoneuroticism. Kuldau et al.

(1976) defend this quality of the index in the following:

(The HOS index) has the advantage of showing high
levels of symptomatology which are not character-
istic of a limited number of specific diagnoses
but cut across the entire range of patients com-
monly treated by community mental health centers.
(Kuldau et al., 1976:1)

Another important issue is the stability of the index

over time. A comparison of HOS values over a three year

period resulted in a high degree of stability (r.=.73),

discounting suggestions that the index measures only

transient phenomena like physical illness or temporary

stress resulting from unusual LCE.

Tousignant et al. (1974) have rejected the HOS as an

epidemiological tool. Their main criticism of the instru-

ment is what they consider to be the biasing effect of

items tapping physical illness. Yet, according to Schwab

et al. (1970) and other researchers, many patients' ill-

nesses cannot be dichotomized solely as psychiatric or

medical. In fact, it seems that individuals with a high

record of physical symptomatology commonly manifest high

psychiatric symptomatology as well. Neither factor has

been identified as antecedent to the other in the litera-


The Psychosocial Dysfunction Scale

The Psychosocial Dysfunction Scale (PDS) was created

as part of the Florida Health Survey in order to measure

psychiatric symptom-related inabilities to perform social

roles. The scale contains items gauging the degree to which

worry, nervousness or fear of having a nervous breakdown

interfere with respondents' functioning. These items tap

functions which are common to work, family and social life.

The scale was analyzed for internal consistency and

found to have an overall Cronbach's Alpha of .88 and item-

whole correlations ranging from 0.48 to 0.67. This

indicates high reliability.

Measurement of Life Crisis Events

For purposes of the present research,an LCE has been

defined as a life experience, such as being fired from

one's job, which is focused in time and involves some kind

of change in role or status that is potentially upsetting

for the individual. In the measurement of these events,

Paykel et al. (1971) employed the average degree of

"upset" created by 61 different events rated on a scale of

from 0 to 20. They gave primary consideration to the un-

desirability of events since they believed this to be

highly related to psychiatric symptoms, which was their

primary concern. The more frequently cited Holmes and

Rahe (1967) schedule of recent life events, by contrast,

estimates the degree of readjustment required by events

and ignores the dimension of desirability. It should be

noted that B.S. Dohrenwend (1973) and Myers et al. (1971)

have reported significant differences in the effects of

desirable and undesirable events. Hence, this disserta-

tion employs work-related events included in the Paykel

index since the objectives of this study are basically

consistent with those of the Paykel index.

Paykel et al. (1971, 1972) found that the majority

of crisis events fell into exit (loss) categories such

as social separations. These events were also more

heavily weighted and more strongly related to psychiatric

symptomatology. Also, the great majority of events (48)

were judged by Paykel et al. (1972) either as "probably

symptom dependent" or as having an unknown relationship

with prior symptomatology. Only 13 of the events were

defined as having no probable relationship with prior

symptomatology. These were events that were either

externally determined like the major illnesses of a

family member, or somehow connected with the life cycle

such as births and deaths. Most of the events included,

then, are seen as potentially symptom-related.

While the Paykel index does not measure multiple

occurrences of the same event, a comparison with a similar

measure which did, revealed virtually no difference (r.=.96)

(Holzer, 1977).

Measures of Work-Related Life Crisis Events (WRLCE)

WRLCE are measured in an unweighted scale of total job

events. Previous research (Holzer, 1977) demonstrates no

significant difference between analyses employing weighted

and unweighted versions of LCE. This measure of total

job events is complemented by two measures which differen-

tiate between those events which are probably symptom-re-

lated and those which are not. The first of these

measures, "Major Event,"is a composite of five events

having a likely association with prior symptomatology:

business failure, demotion, fired, troubles with boss or

co-workers and unemployed for one month. The second

measure, "Minor Event," is a composite of three events

having an unlikely association with prior symptomatology:

change in line of work, change in work conditions, and

change in work hours. While there are cases in which

crises such as being fired or having troubles with one's

co-workers are the consequences of healthy self-assertion

in an unhealthy, repressive work environment, there is

evidence that such situations are not typical (Tiffany,

1970; Robins, 1966).

Demographic and Resource Control Variables

Standard demographic controls are employed throughout

the presentation of findings. Among these are race, sex,

age, and marital status. Income, occupation,

and interpersonal resources function as coping-resource


Dunham et a!. (1966) provide evidence that an aggregate

SES measure is an imprecise tool for psychiatric epidemiol-

ogy since they found educational attainment to be a less

reliable predictor of level of functioning than level of

income or occupation. They also found that individuals

who were high in education but low in income and average

in occupational level were more likely to experience higher

dysfunction than those who had high income but were low

in the other two areas. Hence, because this study is

concerned with work and work-related functioning, it is

better to differentiate between the three components of

SES so as not to confound their influence.

Plan of Analysis: Foundations

Employing the above operational definitions based

on the scales utilized, it is possible to formalize the

three major hypotheses introduced in Chapter 2 and to

discuss the application of appropriate statistical tech-

niques. The first part of the analysis describes the

formal testing of hypotheses treating component measures

of WRLCE with the HOS, the PDS, demographic and resource

variables. The second part of the analysis describes re-

lated tests of secondary work behavior measures.

Primary Methods of Analysis

Chi square tests of significance and one-way analysis

of variance are employed in the crosstabular presentation

of data. In addition, multiple regression analysis

includes F tests of significance. While most of the

important relationships are longitudinal, synchronic

relationships are also examined. In cases where few var-

iables are being considered, emphasis is placed on the

crosstabular display of data. When a greater number of

independent variables enter the analysis, however, multiple

regression is favored. Multiple regression offers the

advantage of permitting the measurement of many diverse

variables as they relate to a given dependent variable.

Further, it does not require any process of elaboration

(Kerlinger, 1973).

Multivariate Analysis of Change Scores

Crosstab displays of HOS and PDS scores permit

simple comparisons of case and non-case categories for

different demographic and social resource groups. For

example, HOS scores above 30 and PDS scores above 5 put

respondents into probable cause categories. Examining

change in caseness over time as it relates to WRLCE and

and other control factors is one of the primary concerns

of the present study. Crosstab displays offer the

advantage of easy partitioning of continuous variables

into distinct categories.

Multiple regression analysis of change scores offers

a powerful interval level measure which has even greater

utility for purposes of this study since it produces a

percentage of total variance explained by the impact of

each independent variable on the dependent variable.

However, Cronbach and Furby (1970) and Bohrenstedt (1968)

are among those who caution that statistical regression

towards the mean is particularly a problem with this

variety of change score analysis. More specifically,

multiple regression change scores for those high in

symptomatology at time 1 will tend to be negative (lower)

at time 2. Hence, there is a natural regression towards

the mean for high and low scores over time.

Fortunately, this difficulty can be corrected for

in a multiple regression equation which calculates HOS

scores (TI) from a model containing HOS (T2) as the

predictor variable. The following model, based on the

work of Holzer (1977) contains such a correction factor.

More importantly, it provides the foundation for the

present multiple regression analysis.

HOS, T, = 8 + 8 HOS, T +
HOS, T 0 l


In the above, HOS, T2 is the dependent variable; Bo is a

constant which is derived from the shift in the mean HOS

score from Ti to T2; HOS, T1 is the independent variable;

and is the symbol for random error which accounts for

individual lack of fit in the model (Holzer, 1977:62).

Expanding the above model to include estimation of

the influence of WRLCE on change in HOS scores over time

results in the following equation:

HOS, T2 = c0 + B1 HOS, Ti + B2 WRLCE + E (3.2)

It is a relatively simple matter to expand the above

formula as additional controls are required.

As pointed out at the beginning of the plan of analysis,

terms in multiple regression equations can be interdependent

to varying degrees. The introduction of crossproducts as

shown in equation 3.3, centered by subtracting the means

of each component variable before inclusion in the

equation, results in the avoidance of the distorting effects

of such variable overlap on multicolinearity:

HOS, T2 = 8, + B HOS, T1 + 2 WRLCE + B3 Income

+ 4 WRLCE x Income + E (3.3)

A complete discussion of this and related matters is in

Holzer (1977). Finally it should be noted that the p<.05

level of probability will be used throughout as the level

for rejecting the null hypothesis.

Plan of Analysis: Formal Hypotheses

This section consists of an outline of the various

means whereby the formal hypotheses will be tested. By

way of review, the general analytic framework upon which

the following hypotheses are constructed is illustrated

in Figure 3.1.

Providing a foundation and introduction for the

formal hypothesis testing described below are simple

crosstabular breakdowns of HOS and PDS caseness scores

by different social and demographic variables. Statis-

tical tests include chi square and one-way analysis of

variance (ANOVA).

Hypothesis #1

The first hypothesis investigates the relationship

between events and symptoms:

It is hypothesized that if life crisis events
have a deleterious effect on mental health,
then those with a large number of WRLCE will
have higher symptom and dysfunction scale
scores than those experiencing few or no
events, controlling for Ti scores.

Multivariate chi square and ANOVA of change in

HOS and PDS case/non-case statuses over time controlling

for "major event" and "minor event" (WRLCE) constitute

the first set of tests of this hypothesis. Constituting

the second set of tests of the hypothesis are multiple

regression analyses of HOS and PDS change scores


[Ti] IT2]

Hi lossHi outcome (HOS, PDS"scores)

H S PDS(WRLCE) ~-- ^Lo outcome (HOS, PDS scores)
scr lossHi outcome (HOS, PDS scores)

(WRLCE).Lo outcome (HOS, PDS scores)

Hi lossHi outcome (HOS, PDS scores)
Hi loss

jLo HOS, PDS (WRLCE)-L outcome (HOS, PDS scores)
SLo loss -- oHi outcome (HOS, PDS scores)

(WRLCLo outcome (HOS, PDS scores)

Figure 3.1 Different Possible Outcomes of HOS* and PDSt Scores
Over Time Controlling for WRLCE#

controlling for three continuous WRLCE measures: job

event total, major job event total, sex, age, race,

marital status, and part-time/full-time work status.

Hypothesis #2

The second hypothesis focuses on the relationship

between symptoms and events:

It is hypothesized that if psychiatric symptoms
produce a greater number of life crisis events,
then those with high symptom and dysfunction
scale scores (Ti) will have a larger number of
WRLCE (T2) than those with low Ti scores.

The major tests of this second hypothesis consist

of multiple regression analyses of HOS and PDS scores,

Ti, with criterion variable WRLCE score measures (T2)

controlling for sex, age, race, marital status, part-time/

full-time job status and income.

Hypothesis #3

The third hypothesis examines the relationship between

events, coping resources and symptoms:

It is hypothesized that if social and economic
resources mitigate the impact of life crisis
events, then the occurrence of WRLCE will
produce a greater increase in symptom scale
scores for those with few social and economic
resources than for those with many such

Because of the large number of variables involved in

this hypothesis, concern for analytic clarity dictates

the creation of two distinct sub-hypotheses:

A. It is hypothesized that if economic
resources mitigate the impact of life
crisis events, then the occurrence of
WRLCE will produce a greater increase
in symptom scale scores for those low
in income and occupation than for those
high in income and occupation.

In order to test the above, multiple regression

equations for the criterion variables HOS (T2) and PDS

(T2) are introduced. Income, WRLCE, occupational level,

and standard demographic controls are included as

independent variables.

B. It is hypothesized that if social
support resources mitigate the impact
of life crisis events, then the
occurrence of WRLCE will produce a
greater increase in symptom scale
scores for those with less social
support than for those with greater
social support.

This sub-hypothesis is also tested in a multiple

regression equation for dependent variables HOS and PDS.

The primary independent variable, support, is a measure

derived from dummy variables for marital status, presence

of family and friends on whom one can depend, and whether

one has a friend of the opposite sex (for single

individuals). These components of one support measure

are also entered individually in various regression

equations. Standard demographic controls are employed

among the independent variables.

Other Measures of Work Behavior

A number of secondary measures of work behavior

were developed for the present research. Although these

measures are not necessary to test any of the hypotheses,

they offer a means of complementing and clarifying the

relationship of WRLCE with other variables in the research

design. First among these is an index based on reasons

given by respondents for having left their last job.

Negative reasons for having left former job include the

following: diffuse job dissatisfaction, specific job

dissatisfaction, discharged/fired/laid off, physical

health, authority relationships, interpersonal relation-

ships and inability to perform work. Positive or external

demand reasons include: plant closing/end of work,

retirement, change for advancement, marriage, and family

problems. This measure is useful inasmuch as it parallels

WRLCE in content and precedes it in time.

In addition, years employed on current job and number

of times unemployed complement the WRLCE focus on short-

term work crisis experiences by providing insight into

long-range stability of work behavior.

Finally, the present study includes indices of

occupational and income mobility. These measures permit

investigation of the relationships between changes in

occupation and income levels over time compared with

changes in HOS and PDS scores.

Tests of Secondary Work Behavior Variables

Three work-related measures tap work behavior variables

which preceded the first interview: reasons left last

job, times unemployed (in the three years preceding the

first interview) and years employed on current job.

Each of these variables is tested for degree of

correlation with Ti HOS, PDS, and subsequent WRLCE scores.

Controls for age, race, sex, income, part-time job

status, and social support are included in the analysis.

Finally, the variables, income difference and

occupational difference, constitute intervening measures

between T, and T2 psychiatric scale scores. The effect

of these social resource-related measures of social

mobility on criterion HOS and PDS change scores is

determined by multiple regression analysis employing

standard demographic controls.




This chapter presents a discussion of the employed

sample composition and the general distribution of HOS

and PDS scores, followed by a presentation of the tests

of hypotheses described in Chapter 3. These statistical

tests, appropriate for crosstabular data presentation

and multiple regression analysis, provide a framework

for the presentation of data. The chapter concludes

with an analysis of findings of ancillary measures of

work-related behavior.

Employed Sample Characteristics

Individuals reporting full or part-time employment

at T2 constitute the sample of employed people (N=267);

thirty-nine of these were part-time workers. They were

included in order to learn more about their social demo-

graphic and scale score differences with full-time workers.

It should also be understood that respondents who

were no longer employed at T2 (N=43) were retained in the

sample so that factors contributing to the attrition

of the working population could be considered. Their

loss of employment is considered an important outcome

which should not be overlooked.

Table 4.1 introduces the analyses which follow

with a breakdown of the primary demographic and resource

variable characteristics of the full sample of 1645

respondents compared with the sub-sample of employed

people interviewed at Time 1 (Tj) and Time 2 (T2) (N=267).

Tables 4.2 through 4.5 display social and demographic

variable breakdowns of HOS and PDS scores for the full

original and longitudinal employed sample of respondents.

Because the central research questions of this study all

deal with the reciprocal effects of WRLCE with HOS and

PDS scale score levels, no direct analysis of the formal

hypotheses is involved at this point. However, a founda-

tion for the analyses which follow is given in these

tables. For example, breakdowns of income and occupa-

tional level variables provide a framework for under-

standing how social resource variables modulate HOS and

PDS scale score levels. This helps clarify the later

analysis of Hypothesis 3 which examines the interrelation-

ship between WRLCE, psychiatric scale scores and social

resource variables.

Table 4.2 displays full sample (N=1645) HOS case-

ness scores by sex, race, age, and personal income.

Females, the elderly, Blacks and those under $3000 in

income all had significantly higher mean scores and

Demographic Characteristics of the Original and
Follow-up Worker Samples.

Full Sample Sample of Sample of
1970 Data Workers (Ti) Workers (T2)
(N=1645) (N-299) (N=299)

1al e

60 +


Under $3,000
15,000 +

Common Law

Grade School
Some High School
High School Graduate
Some College
College Graduate

736 44.7
909 55.3











N %

147 55.1
120 44.9

N %

147 55.1
120 44.9


77 28.8
189 70.8
1 0.4
















Table 4.1

Table 4.2 Full Sample HOS Scale Scores
By Sex, Age, Race, Income

Number Mean Case

TOTAL 1645 27.5 17.3

Male 736 26.7 21.6
Female 909 28.0 33.8
*** ***
16-22 270 27.1 25.6
23-29 315 25.9 6.5
30-44 411 26.7 23.8
45-59 332 28.5 34.9
60 + 317 29.2 41.3
*** ***
Black 336 29.0 39.3
White 1267 26.7 25.4
*** ***
INCOME (Personal)
Under $3000 119 28.6 32.7
3,000-5,999 291 27.1 27.0
6,000-9,999 240 25.2 12.0
10,000-14,999 96 25.1 13.6
15,000 + 70 25.3 12.5
*** ***

p<.05 S.D.
** p<.01 4.3
*** p<.001 to
Chi square for Caseness 6.4
Anova for Means

percentages in the case range than did their social and

demographic opposites. The inverse relationship between

income and caseness is consistent with findings in the

literature of an inverse relationship between SES and

psychiatric scale scores (Warheit et al., 1973; B.P. and

B.S. Dohrenwend, 1969).

Table 4.3 presents HOS scores for the employed sample

at TI and T2 for key social and demographic variables.

The relatively low N of this sample is likely to result

in the underestimation of some values; caution should

therefore be employed in their interpretation.

The same general pattern of relationships found in

Table 4.2 are similar for sex, race, and income in Table 4.3,

although their level of significance is diminished and

has disappeared altogether in several cases. The relation-

ship for age differs, though, in that there is relatively

low caseness for those sixty and over. This last relation-

ship suggests that those among the elderly who remain

employed form a cadre of "healthy survivors." Their

subsequent experience of higher caseness, T2 can per-

haps be attributed to difficulties in adjusting to

retirement, and to the development of physical disabilities.

Regarding this last comment, it should be noted that

physical symptomatology has been associated with higher

HOS caseness levels (Schwab et al., 1970).

Table 4.3

Employed Sample HOS Scale Scores By
Key Social and Demographic Variables

Time 1 Time 2

% %
N Mean Case N Mean Case

267 26.68 21.3

147 25.98
120 27.53





77 27.57 27.3
189 26.32 19.0

INCOME (Personal)
Under $3,000
15,000 +

Upper White Collar
Lower White Collar
Upper Blue Collar
Lower Blue Collar




228 26.02
39 29.23




267 27.07 23.6

147 26.34
120 27.98


77 29.49
189 26.10



197 26.69
26 29.42

** p<.01
*** p<. 001
Chi square for Caseness
Anova for Means


Ma le

60 +









Of particular interest is the significant relation-

ship between occupational level and HOS caseness. Lower

(unskilled) blue-collar workers experience the highest

symptom level in both surveys for caseness (p<.05, p<.01)

and mean scores (p<.01). Surprisingly, the upper (skilled)

blue-collar workers have the lowest HOS caseness and mean

scores of any group at T1 and the second lowest at T2.

Their scores are consistently lower than those of lower

(clerical) white-collar workers, suggesting that skilled

blue-collar workers are selected for their higher level

of adjustment and/or that they experience a comparatively

healthy work environment.

Another important relationship is the higher case-

ness and mean score levels of part-time workers (N=39)

(p<.05, p<.01). That high level of caseness should be

related to part-time work status indicates that work-

related symptomatology and work dysfunction are to some

extent related. Further supporting this relationship

is the finding that T, part-time workers have approximately

three times the employment drop-out rate of full-time

workers (p<.001).

As described earlier, individuals no longer employed

at T2 (N=43) were retained in the longitudinal sample

so that factors contributing to the attrition of the

working population might be considered. Since only 11.6%

of the respondents in this group stated that they were

"in between jobs," the group is composed almost entirely

of people who decided to leave the labor market, 55% of

whom are under sixty years of age. Statistical analysis

of this group demonstrates a distincly disadvantaged

profile, with significantly higher HOS caseness and

mean score levels both at Ti (p<.01) and at T2 (p<.001).

In addition, lower educational (p<.001) and income

levels (TI), were reported. Finally, individuals in

this group are older (p<.001) and more commonly female

(p<.001). These relationships are consistent with

repeated research findings describing a strong on-going

association between socioeconomic and psychiatric symptom-

related disadvantage (B.S. Dohrenwend, 1973; Warheit,


The patterns of association for PDS scores with

social and demographic factors in Tables 4.4 and 4.5 are

basically the same as those for HOS with several exceptions.

First, PDS scores are lower for the total population than

are HOS scores. It has been suggested that this is

because PDS taps a deeper level of psychiatrically

diagnosed disability which is less commonly found in the

population than factors tapped by HOS (Arey, 1976).

Second, PDS score level is inversely related to age.

Warheit et al., (1973) comment that this and other

Table 4.4

Full Sample PDS Scale Scores
By Sex, Age, Race, Income

Number Mean Case

TOTAL 1645 2.5 14.3

Male 736 5.5 10.5
Female 909 7.3 17.5
**** ***
16-22 270 3.2 24.8
23-29 315 2.2 17.5
30-44 411 2.5 15.8
45-59 332 2.5 16.3
60 + 317 1.9 13.6
Black 336 8.6 24.0
White 1267 5.9 11.7
*** ***
INCOME (Personal)
Under $3,000 119 3.3 23.5
3,000-5,999 291 2.5 16.8
6,000-9,999 240 1.2 7.4
10,000-14,999 96 1.9 13.5
15,000 + 70 1.7 9.7

Chi square for Caseness
Anova for Means


Employed Sample PDS Scale Scores By
Key Social and Demographic Variables

Time 1 Time 2
o/ o/
/oN Mean High N Mean High
N Mean High N Mean High

267 2.13 15.4

147 1.40
120 3.02




77 2.31 19.5
139 2.06 13.8

267 2.58 18.4

147 2.05 15.6
120 3.22 21.7



77 4.06 24.7
189 1.97 15.9

INCOME (Personal)
Under $3,000
15,000 +

Upper White Collar
Lower White Collar
Upper Blue Collar
Lower Blue Collar


228 1.70 12.7
39 3.65 30.8

197 2.05 14.2
26 4.08 30.8

p .05
** p .01
*** p .001
Chi square for Percent High
Anova for Means

Table 4.5


Mal e

60 +











measures of anxiety are more associated with youth than

most other commonly employed psychiatric measures of


Description of WRLCE

Table 4.6 lists the various measures which comprise

the primary work behavior variables in this study. It

should be noted that the five events with the highest

"upset" ratings define the Major Event variable and that

the three events at the bottom of the scale define the

Minor Event variable employed in this chapter. In all,

54(20.2%) of the total number of respondents experienced

one or more major work-related event and 76(28.5%)

experienced one or more minor event during the time

period between the two interviews. Among those reporting

work events, 51 (52%) experienced two or more events.

Only 13 (13.3%) claimed they sustained two or more major

WRLCE, however.

Tests of Hypotheses

Hypothesis 1: Stability of HOS and PDS over time

controlling for WRLCE. An initial test of Hypothesis 1

is based on a simple crosstabular dichotomization of

respondents into case and non-case categories. PDS scores

are divided into high and low categories depending on

whether scores are one standard deviation above or below

Table 4.6 Paykel's Work-Related Items, Their Weights
and Frequency of Occurrence (N=267).

Upset Respondents
Event Rating Reporting Event

N %

1. Business Failure 16.46 4 1.5

2. Fired 16.45 5 1.9

3. Unemployed for
one month 15.26 34 12.7

4. Demotion 15.05 2 0.7

5. Troubles with
Boss or Co-Workers 12.21 22 8.2

6. Change in
Work Hours 9.96 50 18.7

7. Change in
Work Conditions 9.23 37 13.9

8. Change in
Line of Work 8.84 34 12.7

Table 4.7

Transition of Respondents Between Case
Categories Controlling for WRLCE

Transition from Transition from
Non-Case to Case Case to Non-Case

Number of Percentage Number of Remaining in
Non-Cases Becoming a Cases the Case
TI Case T2 Category

HOS Scores

Event Type

No Event 136 10.1% X2=4.11 31 61.3% X2=2.79
Minor 35 8.6 N.S. 9 55.6 N.S.
Major 37 21.6 17 82.4

PDS Scores

Transition from Transition from
high to low low to high
score status score status

Event Type

No Event 153 8.5% X2=6.90 16 56.3% X2=0.76
Minor 35 11.4 p<.05 9 44.4 N.S.
Major 38 23.7 16 62.5

the mean. Again, Hypothesis 1 states that the experience

of WRLCE is related to higher levels of symptoms and

dysfunctions at T2.

The findings presented on the left-hand side of

Table 4.7 show that those experiencing major work events

are more than twice as likely to fall into the case/high

scale score groups at T2 than those experiencing no

events or only minor events. This gives qualified

support to the hypothesis.

The statistics on the right-hand side of the table

give only limited support to Hypothesis 1, however, in

that they reveal only a trend for greater retention of

those with major WRLCE in the HOS case and PDS high

categories. It should be considered that small N's

might be devaluating the significance of the values


Regression Analysis of HOS and PDS Change Scores

A regression model for analysis of change in HOS

and PDS scores has been created for the methods chapter.

This model, as applied to Hypothesis 1, predicts T2 HOS

and PDS scores from T, scores and measures of WRLCE.

Additional control variables include age, income, sex,

race, marital status and full-time/part-time work status.

The last four of these are treated as dummy variables.

For example, part-time employed individuals are coded

as "1" and full-time employed individuals are coded as

"0" in the part-time variable.

As stated in the last chapter, it is important to

avoid multicolinearity among key variables when employing

regression techniques. To recapitulate, multicolinearity

is a situation in which the degree of association between

the independent and any one of the control variables

changes with the value of a second control. Because of

an expected association between part-time work status

and the experience of (fewer) work events, a cross-

product term for the interaction between these variables

has been introduced into the regression equations for

Hypothesis 1. Additional interaction terms have been

added for Hypothesis 3 and will be pointed out later.

In the regression equation presented in Table 4.8

it can be seen that T, HOS scores accounted for more

explained variance in T2 HOS scores than any other

factor (F=178.206, p<.001, RSQ change = 32.886%). This

finding is as expected since the stability of HOS over

time has been established in other studies, a number

of which are concerned with these data (Warheit, 1979;

Holzer, 1977).

While the F score for major WRLCE was relatively

high (3.265) and explained approximately 3.1% of the

variance in T2 HOS scores, it failed to achieve a level

Table 4.8

Regression Analysis of T2 HOS Scores as Change from Ti
HOS Scores, Using Major WRLCE and Sociodemographic
Controls as Predictors.

Regression Coefficients

Variable B Beta Error Beta F Significance

HOS, Ti 0.660 0.614 0.050 178.206 p<.001
Major WRLCE 0.836 0.084 0.463 3.265 N.S.
Race-Black 2.263 0.190 0.428 18.370 p<.001
Sex-Female -0.164 -0.015 0.506 0.105 N.S.
Part-time 1.604 0.102 0.739 4.711 p<.05
Age 0.023 0.054 0.019 1.386 N.S.
Not Married 0.051 0.004 0.573 0.008 N.S.
Part-time x WRLCE 0.004 0.000 0.533 0.000 N.S.
(Constant) 7.319

Analysis of Variance

Sum of Mean
df Squares Square F Sianificance

Regression 8 4137.918 517.239 36.582 p<.001

Residual 255 3605.539 14.139

Multiple R

R Square

Standard Error




of significance in the equation. It should be noted

that major WRLCE was selected over total WRLCE for this

and other tests of relationship because preliminary

analysis showed it was a better predictor of T HOS

and PDS scores than the other two WRLCE measures.

Although they have considerably smaller F scores

than HOS, several other variables produced levels of

significance in the total equation: part-time work

status (F = 4.711, p<.05) and race (F = 18.370, p<.001).

The high significance for race should be interpreted

cautiously because controls for income and other social

resource variables are not called for in this hypothesis.

It will be recalled that scale score differences for

race repeatedly have been reduced to non-significance

when controlled for SES (Warheit et al., 1975b). *The same

result is anticipated when income is used as a control.

Finally, the significant positive association between

part-time work status and higher levels of HOS scale

scores is in keeping with the crosstabular findings

already presented.

The independent variables in this regression to-

gether explain 53.4% of the variance in T2 HOS scores.

T1 HOS scores accounted for 36.172% of the variance,

WRLCE and other work-related variables for 13.287% with

sex, race, age and marital status combined explaining

the remaining 3.978%.


In Table 4.9 the same basic equation is introduced

for T, and T2 PDS scores. Beacuse PDS is a less stable

measure over time than HOS, the total variance explained

in this regression by Ti scale scores is comparatively

lower (27.992%). Another difference is the greater

significance for major WRCLE (F = 5.555, p<.05). This

finding is not surprising inasmuch as the PDS scale

was designed to gauge psychiatric dysfunction syndromes

and therefore would relate more closely to work dys-

function measures like WRLCE than does the,HOS.

Race is less significant in this regression (F = 6.937,

p<.01) while work status remains at approximately the same

level (F = 6.155, p<.05). In addition, T2 PDS demonstrates

much greater sensitivity to the interaction between WRLCE

and work status (F = 8.375, p<.001). Sex, age and

marital status retain the same general pattern of non-

significance in this equation, however.

The combined independent variables in this equation

explain 37.0% of the variance in T2 PDS scores. Again,

T, PDS scores alone account for 27.992%. Work-related

variables define 6.616% of the total and sociodemographic

variables together explain the remaining 2.343%.

Hypothesis 2.: The Relationship between WRLCE and

antecedent HOS and PDS scores. If WRLCE are significant

predictors of T2 PDS, what is the nature of the relationship

Table 4.9

Regression Analysis of T2 PDS Scores as Change from Ti
PDS Scores, Using Major WRLCE and Sociodemographic
Controls as Predictors.

Regression Coefficients

Variable B Beta Error Beta F Significance

PDS, T1 0.540 0.427 0.069 60.401 p<.001
Major WRLCE 1.181 0.127 0.501 5.56 p<.05
Race-Black 1.505 0.136 0.571 6.9 p<.05
Sex-Female -0.151 -0.014 0.551 0.075 N.S.
Part-time 1.955 0.133 0.788 6.15 p<.05
Age 0.010 0.025 0.021 0.211 N.S.
Not Married 0.543 0.047 0.634 0.735 N.S.
Part-time x WRLCE 1.702 0.153 0.588 8.375 p<.01
(Constant) -0.060

Analysis of Variance

Sum of Mean
df Squares Square F Significance

Regression 8 2479.337 309.917 18.680 p<.001

Residual 255 4230.568 16.590

Multiple R

R Square

Standard Error




between T, PDS and HOS scores with subsequent WRLCE? The

results of multiple regression tests of Hypothesis 2

found in Tables 4.10 and 4.11 provide an answer to this


A clear pattern of inverse association between age

and WRLCE emerged in these two tables. Age was the most

significant predictor of WRLCE both in the equation in-

cluding HOS (F = 45.182, p<.001) and in the equation in-

cluding PDS (F = 38.929, p<.001) as one of the independent

variables. Furthermore, similar high levels of signifi-

cance were found in preliminary analyses using major and

minor WRLCE, respectively, as the dependent variable.

That younger workers experience more LCE in their work

environment than older workers probably relates to their

unsettling search for an occupational niche.

The second strongest predictor of WRLCE was antece-

dent HOS (F = 15.333, p<.001) in Table 4.10 and antecedent

PDS (F = 27.499, p<.001) in Table 4.11. These findings

support Hypothesis 2 in demonstrating a strong relation-

ship between WRLCE and antecedent symptoms and dys-

functions. Furthermore, the larger F score for PDS

supports the theoretical assumption that PDS surpasses

HOS as a predictor of work dysfunction. Finally, as

with age, preliminary analyses revealed similarly high

levels of significance for HOS and PDS employing major

and minor WRLCE each as the independent variable.

Table 4.10

Regression Analysis of HOS and Sociodemographic
Predictors of WRLCE Scores.

Regression Coefficients

Variable B Beta Error Beta F Significance

HOS, TI 0.050 0.224 0.013 15.333 p<.001
Income, Ti -0.000 -0.077 0.000 1.309 N.S.
Sex-Female -0.322 -0.140 0.145 4.973 p<.05
Race-Black -0.118 -0.047 0.156 0.576 N.S.
Not Married 0.404 0.155 0.154 6.873 p<.01
Age -0.033 -0.374 0.005 45.182 p<.001
(Constant) 0.958

Analysis of Variance

Sum of Mean
df Squares Square F Significance

Regression 6 72.654 12.109 11.534 p<.001

Residual 260 272.971 1.050

Multiple R

R Square

Standard Error




Table 4.11

Regression Analysis of PDS and Sociodemographic
Predictors of WRLCE Scores.

Regression Coefficients

Variable B Beta Error Beta F Significance

PDS, Ti 0.079 0.294 0.015 27.499 p<.001
Income, Ti -0.000 -0.102 0.000 2.486 N.S.
Sex-Female -0.372 -0.162 0.142 6.849 p<.01
Race-Black -0.084 -0.033 0.153 0.306 N.S.
Not Married 0.295 0.113 0.153 3.729 N.S.
Age -0.030 -0.339 0.005 38.929 p<.001
(Constant) 2.076

Analysis of Variance

Sum of Mean
df Squares Square F Significance

Regression 6 84.205 14.034 13.958 p<.001

Residual 260 261.420 1.005

Multiple R

R Square

Standard Error




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID EF7G1XIIN_B8UFY1 INGEST_TIME 2012-03-08T19:06:45Z PACKAGE AA00003876_00001