Tài liệu tâm lý học sức khỏe: Nhân cách và các yếu tố nguy cơ

9
Personality Traits as Risk Factors for
Physical llness

The belief that stable patterns of thought, emotion, and behavior contribute
to the development of physical illness has been present throughout the
history of medicine (McMahon, 1976). Hippocrates, for example, argued
that four basic temperaments or personality types reflected excesses of
specific humors and caused corresponding medical disorders. Many
centuries later, Sir William Osler (1892) suggested that coronary heart
disease befell “not the neurotic, delicate person …but the robust, the
vigorous in mind and body, the keen and ambitious man, the indicator of
whose engine is always at full speed ahead” (p. 839). The descriptions of
personality, disease, and the nature of their relation have varied widely, but
the essence of this psychosomatic hypothesis has remained unchanged.
Earlier in this century, the hypothesis was refined by the psychoanalytic
school in psychosomatic medicine (Alexander, 1950; Dunbar, 1943). These
models assigned a pathophysiological role to unconscious personality
dynamics, and suggested a correspondence between specific emotional
conflicts

and

medical

conditions.

Unlike

previous

psychoanalytic

formulations of hysteria or hypochondriasis (Freud, 1933), these models
identified causes for actual disease, rather than unfounded physical
symptoms. For example, an unconscious conflict between aggressive
impulses and anxiety concerning the consequences of their expression was
described as a cause of essential hypertension. Although a weak scientific


foundation limited the impact of this approach on the mainstream of either
medicine or psychology (Surwit, R. B. Williams, & Shapiro, 1982), it set the
stage for current research on personality and illness.
During the same period, developments in the physiology of stress provided
an essential, scientifically credible set of mechanisms connecting personality
and disease (Ax, 1953; Cannon, 1939; Seyle, 1936, 1952; Wolff, 1950). Not
surprisingly, the psychophysiology of stress and emotion remains an integral
component of this research area (Contrada, Leventhal, & O'Leary, 1990).
The immediate predecessor of the current interest in the issue is undoubtedly
the seminal work of M. Friedman and Rosenman (1959) on the Type A
coronary prone behavior pattern. Although M. Friedman and Rosenman
actively avoided describing their work in the language of personality traits,
their work is now recognized as involving personality characteristics (Suls &
Rittenhouse, 1987). Friedman and Rosenman's version of the centuries- old
psychosomatic hypothesis was a major force in the early development of the
larger fields of behavioral medicine and health psychology (G. C. Stone, F.
Cohen, & Adler, 1979; Weiss, Herd, & Fox, 1981).
An often overlooked forerunner to current research on personality traits as
risk factors for illness are early studies that used psychometrically sound
measures of personality in large, prospective designs (e.g., Ostfeld, Lebovits,
Shekelle, & Paul, 1964). Effects of personality variables on subsequent

disease were examined while attempting to control statistically the possible
confounding medical or demographic variables. Studies of this type
provided important evidence of the merit of the hypothesis and the outlines
of a methodology for constructing a credible epidemiological foundation for
the field.


The current state of research on the hypothesis that personality traits can
influence physical health comprises notable achievements and clear
limitations. On the one hand, several literatures have matured to the point
that the evidence is compelling; specific personality characteristics are
indeed associated with increased risk of serious illness and premature death
(e.g., T. Q. Miller, T. W. Smith, Turner, Guijarro, & Hallet, 1996). Further,
plausible mechanisms accounting for this association have been articulated
and evaluated, at least in a preliminary manner (S. Cohen & Herbert, 1996;
Manuck, 1994). On the other hand, a steady climate of skepticism persists in
much of the medical community (e.g., Angel, 1985), and the empirical
support for the health relevance of some personality traits discussed in this
literature is quite limited. Further, the implications of this work for the
treatment and prevention of illness are largely unknown. Fortunately,
conceptual, methodological, and analytic tools in personality psychology
and behavioral medicine have evolved to the point where future studies will
address these limitations in an increasingly compelling manner.
This chapter provides an overview and critique of the literature concerning
personality traits as risk factors for physical disease. It begins by addressing
some basic issues regarding the nature of personality, disease, and their
potential association. After reviewing models of this association, it turns to
theory and research on the major personality attributes in the field. Finally, it
concludes with a critical evaluation of the state of the literature and issues to
be addressed in its future.

BASIC ISSUES
What Is Personality?


Allport (1937) succinctly argued that “personality is something and
personality does something” (p. 48, emphasis added). Personality traits are
stable patterns of thought, emotion, and behavior that characterize an
individual across time and situations. Traits are presumed to be based in
psychological and/or biological structures within the individual, and they
form a dimensional basis for comparing individuals. For example, some
people are generally friendly and warm, whereas others are cold and
disagreeable, presumably because of differences in their biologic and/or
psychologic “make-up.” Thus, from this perspective, personality traits are
things that people “have” (Cantor, 1990).
In Allport's other, more active meaning, personality refers to the processes
through which an individual's thoughts, emotions, and behavior cohere into
meaningful patterns over time and across situations. These processes include
the ways in which individuals select and interpret the contexts and situations
of their lives, the goals they pursue, the strategies and tactics they employ in
doing so, and the ways in which they evaluate and react to the outcome of
these activities. These more circumscribed and dynamic psychological
processes are closely associated with the stable patterns of thought, emotion,
and behavior that are indicators of traits. Yet, this other sense of personality
is obviously much more concerned with how traits operate, rather than their
description. Thus, the study of personality as “doing” rather than having
(Cantor, 1990) focuses on describing both the psychological mechanisms
underpinning more broadly defined, static personality traits and the ways in
which these “middle units” of personality are dynamically interrelated and
expressed.

Current personality psychology reflects both of Allport's meanings, and
recent developments of both types have the potential to make enormously
valuable contributions to the study of personality and health (T. W. Smith &
P. G. Williams, 1992). In the classic trait perspective, a far-reaching
development is the emergence of the five-factor model of personality as an
adequate taxonomy of basic personality characteristics (Digman, 1990; John,
1990; McCrae & John, 1992). Although descriptions vary across versions of
this model, and despite several notable critics (e.g., Block, 1995), there is
general consensus regarding the traits listed in Table 9.1- Extraversion,
Agreeableness,

Conscientiousness,

Neuroticism,

and

Openness

to

Experience. These traits have been recovered in factor analyses of self- and
other- ratings, and several reliable and valid measures of these broad
dimensions and their subcomponents have been developed (Digman, 1990;
John, 1990).
The validity and potential impact of this taxonomy are evident in the fact
that these traits are clearly not simple mental abstractions or linguistic
conveniences used by raters (e.g., Funder & Colvin, 1991; Moskowitz,
1990). Rather, they reflect verifiable, general dimensions of individual

functioning. Further, these broadly defined traits are stable (McCrae &
Costa, 1990), show patterns of variability consistent with genetic influences
(Bouchard, Lylkken, McGue, Segal, & Tellegen, 1990), and predict behavior
in many circumstances (Kendrick & Funder, 1988).
The elements of this taxonomy can provide a useful guide for organizing the
growing array of otherwise conceptually isolated traits suggested as risk
factors for physical illness (Costa & McCrae, 1987b; Marshall, Wortman,
Vickers, Kusulas, & Hervig, 1994; T. W. Smith & P. G. Williams, 1992).


Traits studied as risk factors are often described and studied individually,
without attention to their overlap or even redundancy with other traits. One
important application of the five-factor taxonomy is the conceptual and
empirical description of traits suggested as potential risk factors. This use of
the five-factor taxonomy might reveal similarities and differences among
otherwise isolated traits. In addition, the five traits themselves might be
viable candidates as risk factors (e.g., Costa, McCrae, & Dembroski, 1989).
In either of these applications of the model, the well-validated assessment
devices are likely to be useful to health researchers.
One important variation on the five-factor model substitutes the dimensions
of Friendliness versus Hostility and Dominance versus Submissiveness for
Agreeableness and Extraversion, respectively (Trapnell & Wiggins, 1990).
This permits the integration of the five-factor approach with the
interpersonal approach to personality (Carson, 1969; Kiesler, 1983; Leary,
1957; Wiggins, 1979). As depicted in Fig. 9.1, the dimensions of dominance
and friendliness define a two-dimensional space, or circumplex. The
circumplex model has been used, conceptually and empirically, to describe a
variety of personality characteristics, interactional behaviors, and social
stimuli (Kiesler, 1991; Wiggins, 1991; Wiggins & Broughton, 1991). As a
result, it has considerable potential for facilitating the integration of

personality and social risk factors for disease (Gallo & Smith, 1998; T. W.
Smith, Gallo, Goble, Ngu, & Stark, 1998; T. W. Smith, Limon, Gallo, &
Ngu, 1996). That is, personality traits and aspects of the social environment
can be described and even assessed through a common framework.
Although the five-factor model and its variants are potentially invaluable in
identifying and organizing traits in the study of personality and health, they


have less to say about the mechanisms through which traits influence
behavior, emotion, and ultimately health. It is here that the second major
emphasis in personality psychology is of use. Unfortunately, less agreement
exists regarding an adequate taxonomy of the “middle units” of personality
processes (Cantor, 1990). However, several overlapping sets have been
articulated, and clear themes have emerged in the related research. These
approaches follow from the cognitive social learning tradition in personality
psychology (Kelly, 1955; Mischel, 1973; Rotter, 1954), and they share many
conceptual similarities to interpersonal approaches in personality and
clinical psychology (Kiesler, 1996; Westen, 1991). Examples of the
constructs described in this literature are mental representations (i.e.,
schemas) of the self, others, relationships, and social interaction sequences
(i.e., scripts); life tasks, motives, and goals; appraisals, values, and beliefs;
strategies, tactics, and competencies in goal-directed behavior; and coping
styles and behaviors (Cantor, 1990; McAdams, 1995; Mischel & Shoda,
1995, 1998; Oglevie & Rose, 1995; Westen, 1995).
An underlying premise in this tradition is that characteristics of the person
are reciprocally related to the social environment. Intentionally or not,
people choose to enter some situations and not others, and their actions and
overt expressions of emotion elicit responses from their interaction partners
in ways that reflect their personality traits (Asendorpf & Wilpers, 1998).
These selected, evoked, and intentionally manipulated features of the

individual's social environment in turn influence the individual (Bandura,
1977; Buss, 1987; Ickes, Snyder, & Garcia, 1997). Thus, an individual's
thoughts, emotions, and behavior are seen as highly responsive to
characteristics of the specific situation, and many situations are modified by


the individual's actions. Through these recurring, reciprocal patterns,
individuals foster social environments that maintain central features of their
personalities over time (Caspi et al., 1989; Kiesler, 1996; Wachtel, 1994;
Wagner, Kiesler, & Schmidt, 1995). A further implication of this view is that
personality processes are best understood in the contexts that comprise and
surround these reciprocal interactions between people and social
environments (Revenson, 1990), such as characteristics of the physical
environment,

subculture,

and

socioeconomic

factors.

Personality

descriptions are likely to be more accurate and informative to the extent that
they consider individuals, their recurring social circumstances, and the
context in which they are embedded.
Current versions of the cognitive-social approach to personality offer the
potential for a comprehensive description of broad traitlike characteristics

and recurring patterns of situationally specific responding (e.g., Mischel &
Shoda, 1995, 1998). That is, the approach has the potential to describe the
mechanisms through which traits, such as those in the five-factor taxonomy,
influence thought, emotion, and behavior in interaction with social
situations. Another important advantage of the cognitive-social perspective
is its overlap with current stress and coping theory, given their mutual
emphasis on cognitive appraisal processes, self- regulation of emotional
responses, and strategies for managing situational threats and demands
(Contrada, 1994). The general stress and coping model (e.g., Lazarus &
Folkman, 1984; Lazarus, 1991) has become a cornerstone of health
psychology and behavioral medicine, and it provides an important
conceptual and empirical connection to the psychophysiological responses
hypothesized to link personality traits and subsequent disease.


Another benefit of the cognitive-social approach is its relevance for
interventions. Although the general trait approach is useful in identifying the
personality characteristics that might be useful foci in interventions intended
to prevent or manage illness, it has less to say about specific targets for
change. With its increased attention to specific psychological mechanisms
and dynamic patterns, the cognitive-social approach is likely to aid in the
articulation and refinement of intervention techniques. For example, whereas
the five-factor taxonomy might identify neuroticism and (low) agreeableness
as useful targets for change, the cognitive-social perspective could suggest
specific patterns of appraisals, beliefs, interaction tactics, and coping
behaviors to be included in such interventions.
What Are the Appropriate Indications of Illness?
A revolutionary difference between psychoanalytic writing on hysteria and
hypochondriasis as opposed to the later work of Alexander, Dunbar, and
their colleagues lies in the nature of the health endpoint under considerationabnormal illness behavior versus actual illness. This distinction was clearly

drawn more recently in conceptual discussions of the potential effects of
personality on health (e.g., F. Cohen, 1979). Outcomes such as symptom
reports, utilization of health care resources (e.g., physician visits), taking
medication, or receiving other treatments typically reflect the presence of
illness, but are fallible indicators. In evaluating the role of personality in
physical illness, care must be taken to avoid mistaking an association
between personality traits and illness behavior for an association with actual
illness. The former may or may not reflect the latter.


Despite the early and clear articulation of this issue, many influential
empirical reports on the association between personality and physical illness
relied heavily on these less definitive indices (e.g., Haynes, Feinleib, &
Kannel, 1980; Kobasa, 1979; Scheier & Carver, 1985). Similarly, several
important reviews of this literature (e.g., H. S. Friedman & Booth- Kewley,
1987) have been criticized for the potential misinterpretation of associations
between personality traits and illness behaviors-especially somatic
complaints-as reflecting the effects of personality on actual physical health
(e.g., Matthews, 1988; Stone & Costa, 1990; T. W. Smith & Rhodewalt,
1991).
In the most notable example of this issue, several investigators have
demonstrated that neuroticism is consistently related to somatic complaints,
even in the absence of actual illness (Costa & McCrae, 1985a, 1987; Watson
& Pennebaker, 1989). If the personality characteristic under consideration is
associated with neuroticism and if the disease endpoint studied wholly or
even partly reflects illness behavior rather than objectively documented
disease, then this interpretive ambiguity arises; the association observed
might involve personality traits and actual illness, or personality and illness
behavior. Given this concern and its potential negative impact on the
identification of robust causal influences on actual illness, symptom reports

and other illness behaviors are no longer considered an acceptable
operational definition of illness. Although illness behaviors are important
topics for research, cumulative progress in the study of personality traits as
risk factors requires less ambiguous methodologies.
A second major development in this literature is the recognition that the
pathophysiology of the major diseases studied varies considerably across


their progression. For example, the beginning stage of coronary heart disease
(CHD) is characterized by microscopic injuries to the endothelial lining of
the coronary arteries. The potential psychophysiological influences on this
process might differ from those of the next stage-the slow progressive
buildup of fatty deposits at the injury sites. Further, the factors that
precipitate the emergence of overt symptoms of CHD (i.e., angina,
myocardial infarction, sudden cardiac death) may differ still from those that
hasten the progression of previously asymptomatic coronary artery disease
(CAD; S. Cohen, J. R. Kaplan, & Manuck, 1994; Kamarck & Jennings,
1991). The natural histories of many forms of cancer and the course of HIV
infection and AIDS include similar possibilities for heterogeneous
psychophysiological influences over time. This poses two challenges for
researchers in the area. First, biologically plausible mechanisms linking
personality and disease must be articulated, and second, mechanisms must
be tied to identifiable points in the etiology of the illness.
How Do We Evaluate the Personality-Disease Hypothesis?
In order to reach valid conclusions about the presence or absence of a
hypothesized relation between a personality trait and illness, several
common methodological issues must be addressed. These can be grouped
into the classic, four categories of validity in research design specified by
Cook and Campbell (1979)-statistical conclusion validity, internal validity,
construct validity, and external validity (or generalizability).

Statistical associations between personality traits and illness are generally
small, at least by the standards typical in behavioral research (J. Cohen,
1990). These effects are often at least as large as those involving more


conventional risk factors for disease. Given the prevalence and cost of the
diseases studied, even small effects can have important public health
implications. However, small effects, multifactorial etiologies, and changing
influences across the course of disease can make it difficult to detect
significant covariation between personality characteristics and disease. As a
result, large epidemiological studies and quantitative combination of
independent results (i.e., meta-analysis) are essential in the development and
evaluation of this literature. The results of individual studies must be
considered with caution, especially if they employ small, select samples. T.
Q. Miller and his colleagues (1991) demonstrated that the use of high risk
samples and the associated restriction of range in disease prevalence and
severity (i.e., disease-based spectrum bias) can mask the statistical
association of personality traits and other risk factors with disease.
Given that personality characteristics are rarely manipulated experimentally,
the issue of internal validity is central in this literature. A common strategy
in early stages of investigation of a risk factor is the concurrent case control
design, in which individuals with the disease (or more severe disease) are
compared with controls on the trait of interest. However, given the many
possible consequences of serious illness, the alternative interpretation that
cognitive, emotional, and behavioral correlates of a disease might reflect
consequences rather than causes of the condition often seems equally
plausible(S. Cohen & Rodriguez, 1995). Prospective designs eliminate this
ambiguity to a large extent, but because of their cost are often
underrepresented in the literature on specific traits. Further, the possibility of
unmeasured third variables a& counting for statistical associations between

personality and disease remains as a threat to internal validity even in


prospective designs. Biologic, psychologic, and sociodemographic variables
could exert simultaneous effects on personality and health, thereby
producing a noncausal association.
As previously noted, many studies have relied on symptom reports or
diagnoses in which somatic complaints are primary criteria (e.g., angina to
operationalize CHD). This poses the threat to construct validity described
earlier; the statistical associations might involve illness behavior rather than
illness itself. The field has shown an increasing recognition of this concern
over time.
However, the construct validity of the personality variables studied as
predictors of illness is still often underemphasized. Scales are sometimes
specifically developed to assess an individual trait described in a newly
proposed model of personality and health. Sometimes such scales are
employed prior to thorough construct validation. As a result, it is often
uncertain if the statistical associations examined actually involve the
personality trait(s) under study, as opposed to some other characteristic(s)
assessed unintentionally. Similarly, the distinct versus overlapping nature of
the growing array of traits in the field is largely neglected (H. S. Friedman et
al., 1995). Thus, it is often unknown whether or not scales intended to assess
similar characteristics actually do so (i.e., convergent validity), and whether
or not scales with dissimilar names and conceptual descriptions actually
assess distinct traits (i.e., discriminant or divergent validity). Careful
attention to the issue, possibly using the five-factor model as an integrative
tool and source of validated indicators (Gallo & Smith, 1998; Marshall et al.,
1994; T. W. Smith & P. G. Williams, 1992), could improve the quality of this
literature considerably.

Finally, as with much medical research, the literature on personality traits as
risk factors can be rightfully criticized as employing samples that are
disproportionately White, middle- class, and male (N. B. Anderson, 1989; N.
B. Anderson & Armstead, 1995; Stanton, 1995). Thus, the generalizability of
effects across sexes, races, and socioeconomic status is often unknown.
Once apparently robust relationships are identified, their generalizability to
more diverse groups should be examined. Given that many of these
personality traits vary systematically as a function of age, race, sex, and
socioeconomic status (e.g., Siegler, 1994), and that these demographic
characteristics are themselves risk factors, the issue is likely to be important
in the future of the research area.
MODELS OF THE ASSOCIATION BETWEEN PERSONALITY AND
DISEASE
Even if reliable and valid associations between personality traits and
subsequent illness can be established, the processes or mechanisms that
underlie these effects have yet to be determined. Several models have been
proposed to describe the association of personality characteristics and
subsequent health (F. Cohen, 1979; S. Cohen & Rodriquez, 1995; Krantz &
Glass, 1984; Suls & Sanders, 1989). The interactional and transactional
stress moderation models suggest that physiological reactivity is the critical
link underlying this statistical association. Likewise, the constitutional
vulnerability model posits a physiological influence on disease, but
identifies personality as a noncausal correlate or epiphenomenon of this
responsivity. The health behavior model suggests that personality influences
health by affecting health practices. Finally, the illness behavior model
indicates that personality impacts the subjective experience of illness and the


behavioral responses to perceived symptoms. The subsequent section

elaborates on each of these models. It is important to note that they are not
necessarily mutually exclusive, and that for any given personality
characteristic, several models may explain health effects in a complementary
manner.
Stress Moderation Models
Stress moderation models are based on the premise that stress is a
fundamental component of the relation between personality and disease
(e.g., F. Cohen, 1979; S. Cohen & Rodriquez, 1995; Contrada, Leventhal, &
O'Leary, 1990; Houston, 1989; Suls & Rittenhouse, 1987). To explain the
historically modest association of stress and illness (Rabkin & Struening,
1976), the interactional stress moderation model goes beyond a direct or
main effect model by suggesting that individuals will differ in their degree of
vulnerability to the detrimental effects of stress, as a consequence (in part) of
their personality characteristics. Thus, personality traits are seen as
moderators (Baron & Kenney, 1986) of the stress-illness relation, in that
illness is more accurately predicted by the statistical interaction of
environmental stress and personality traits. Further, the interactive effect of
stress and personality on illness is, in turn, seen as mediated by differential
psychophysiological responses to stressors as a function of an individual's
standing on the personality dimension(s).
Given its central role in this model, it is important to describe the current
status of the hypothesis that the psychophysiology of stress can influence
disease. The prevailing theory suggests that psychological stress elicits
increases in activity in the sympathetic adrenomedullary and hypothalamic,


pituitary adrenocortical axes. Over time, pronounced, repetitive, or
prolonged physiological responses are thought to contribute to the etiology
of illness. Cardiovascular illnesses (e.g., CHD, hypertension, stroke) are
believed

to

be

fostered

by

activation

of

neuroendocrine

(e.g.,

catecholamines, cortisol) and cardiovascular (e.g., blood pressure) responses
(Barnett, Spence, Manuck, & Jennings, 1997; Kamarck & Jennings, 199 I;
Manuck, 1994; Markovitz, Raczynski, Wallace, Chettur, & Chesney, 1998).
Infectious illnesses and cancer are presumed to be influenced by the effects
of stress on the immune system (Herbert & S. Cohen, 1993; Kiecolt-Glaser
& Glaser, 1995). Although definitive evidence is lacking, these pathways are
biologically plausible and the research to date provides considerable, albeit
preliminary, support for their role in pathophysiology.
Personality may serve to attenuate or exacerbate (i.e., moderate) the
connection between stress and pathophysiology at several places in the stress
and coping sequence (see Fig. 9.2 ). First, the degree to which a given event
will be experienced as stressful depends on an individual's subjective
appraisal of harm or loss, as well as the resources believed to be available

for managing the situation (Lazarus and Folkman, 1984). Certain personality
characteristics are thought to influence this subjective appraisal. For
example, Type A individuals frequently appraise situations as involving
more threat or demand than do Type B persons-a tendency thought to link
TABP to cardiovascular disease (Houston, 1989). In contrast, hardy
individuals are believed to perceive life events to be challenging, rather than
threatening, which may decrease their vulnerability to disease (e.g., Kobasa,
1979). Through these appraisals, personality influences the emotional and
motivational responses to events (Houston, 1992), and emotions and aroused


motives influence psychophysiological response (Wright, 1996). Second,
personality may influence the coping mechanisms that the individual applies
in managing the stressor. Specific coping mechanisms probably cannot be
categorically regarded as adaptive or maladaptive (e.g., Lazarus, 1990).
However, personality may influence the likelihood that the individual will
employ strategies that are adaptive in a given circumstance (Bolger &
Zuckerman, 1995).
Although the succinctness of the interactional stress moderation approach is
appealing, the model is somewhat limited. Fundamentally, it is a model of
individual differences in responses to potentially stressful circumstances.
These responses are viewed as the result of the static or statistical interaction
of personality traits and aspects of the situation. Several researchers have
advocated a more process-oriented approach (e.g., Bolger, 1990; Contrada et
al., 1990, Houston, 1989; Lazarus, 1990; Revenson, 1990) that emphasizes
the ongoing interplay between personality, coping, and contextual factors.
Such models move beyond the static or statistical interactional approach by
acknowledging the type of reciprocal transactions between persons and their
environments typical of the cognitive-social learning and interpersonal
models of personality already described. Thus, transactional views of the

stress moderation process emphasize the ways in which people influence the
objective features of their environments by actively choosing situations and
subsequently responding to them in characteristic ways (Bandura, 1977;
Buss, 1987; Cantor, 1990; Mischel, 1973). For example, because of their
antagonistic interactional style, hostile persons are likely to evoke frequent
interpersonal strain. Such conflicts probably confirm hostile expectations


and increase the likelihood of future antagonistic behavior (Wagner et al.,
1995).
The transactional approach submits that personality influences the
stress/illness relation at three points in the stress and coping cycle (see Fig.
9.3 ). As in the interactional stress moderation approach, personality is
thought to alter the appraisal of events and to influence the choice of coping
responses. In addition, personality is thought to affect exposure to stressful
events, through the individual's selection, evocation, and intentional
provocation of characteristics of the situations they encounter (Buss, 1987).
Personality characteristics that expose the individual to increased stress
through these processes will also elicit the increased psychophysiological
reactivity and subsequent potential for disease discussed earlier. Thus, from
this perspective, psychosomatic processes are not simple consequences of
specific personality characteristics, but reflect a dynamic process emerging
from the recurring transactions between people and the social contexts they
inhabit (Revenson, 1990; T. W. Smith, 1995).
Health Behavior Model
The stress moderation and transactional models rest on the assumption that
the physiological components of stress mediate the association between
personality and disease. In contrast, the health behavior model posits that the
effects of personality traits on health are indirect, mediated by the
intervening effects of health practices (see Contrada et al., 1990; F. Cohen,

1979). (See Fig. 9.4 ). This model is derived from research suggesting that
certain behaviors (e.g., smoking, leading a sedentary lifestyle, and practicing
poor nutrition habits) are reliably associated with disease risk (e.g., Blair et


al., 1989; Holroyd & Coyne, 1987; Paffenbarger & Hale, 1975). Further, the
model draws on research suggesting that personality traits, including
hardiness (Wiebe & McCallum, 1986), neuroticism (Costa & McCrae,
1987a; McCrae, Costa, & Bosse, 1978), and hostility (Leiker & Hailey,
1988; Siegler, 1994) affect the likelihood that one will practice negative
health habits.
Personality might influence the choice of health practices in several ways.
First, psychological factors presumed to guide lifestyle choices may be
components or correlates of personality constructs. Examples include
variables such as locus of control, health beliefs and values, and self-efficacy
(Bandura, 1989; Lau, 1988; Strickland, 1978). Alternatively, negative health
practices may reflect ineffectual coping practices. That is, personality
characteristics, such as hostility, may not only increase the likelihood that
subjective stress will be experienced, but also that maladaptive behaviors,
such as smoking or substance use, will be utilized as emotion-focused
coping strategies. Research suggesting that individuals often adopt more
negative health habits when exposed to stress is consistent with this
hypothesis (Horowitz et al., 1979; Langlie, 1977; Shachter, Silverstein,
Kozlowski, Herman, & Liebling, 1977).
As noted earlier, the health behavior model does not suggest that a direct
physiological pathway connects personality to health. However, there may
be physiological correlates of health behaviors that operate through the set of
psychophysiological responses described in the stress moderation models.
As portrayed in the lower panel of Fig. 9.4, many health practices produce
physiological changes similar to those generated by stress. For example,

stress-related alterations in nutrition or sleep habits appear to attenuate


immune functioning (Hall et al., 1998; O'Leary, 1990). Furthermore, certain
health practices, such as smoking, leading a sedentary lifestyle, and
consuming caffeine, appear to intensify laboratory-induced stress responses
(e.g., Blumenthal et al., 1990; MacDougall, Musante, Castillo, & Acevedo,
1988; M. D. Davis & Matthews, 1990). Thus, it is possible that the effects of
stress on health behaviors produce pathophysiological responses similar to
those described in the stress moderation models.
The health behavior model suggests that the common practice of controlling
traditional risk factors (e.g., smoking, exercise, etc.) in epidemiological
research might lead to an underestimate of the effects of personality on
illness, as some of the impact of personality traits on health may occur
through unhealthy lifestyles. Contrada et al. (1990) described several other
methodological concerns with evaluations of this model. In particular, they
emphasized the problems associated with utilizing self-report indices to
assess health practices. Methodological artifacts such as social desirability
may be particularly problematic in these studies, given the widespread
publicity about the detrimental effects of health practices like smoking. In
addition, Contrada and colleagues noted that the health behaviors commonly
assessed in this research often exhibit only modest intercorrelations (e.g.,
Leventhal, Prochaska, & Hirshman, 1985; Norris, 1997), and that health
behaviors appear to be inconsistent over time (Mechanic, 1979). Failure to
recognize the limitations of most measures of health behavior could lead to
an underestimate of their role as a mediator of the association between
personality traits and subsequent illness.
Constitutional PredisposXon Model

Regardless

of

whether

or

not

the

connection

is

direct,

via

psychophysiological responses, or indirect, via intervening effects on health
behavior, the stress moderation and health behavior models share the
common assumption that the statistical association between personality and
health reflects a causal relation. Several researchers have suggested that it
may not be causal, but instead may reflect the existence of a third variable
(Krantz & Durel, 1983; Suls & Sanders, 1989; R. B. Williams, 1994). As
depicted in Fig. 9.5, this model proposes that an underlying constitutional
vulnerability causes a predisposition for autonomic lability, which
subsequently influences both personality processes (e.g., emotional
responses, etc.) and health problems. Thus, this model considers statistical

associations between personality and subsequent health to be artifacts
resulting from the existence of a biologic third variable.
Given the growing body of evidence suggesting that certain personality
-factors and physiological stress responses may be at least partially
determined by genetic factors (Bouchard et al., 1990; Smith et al., 1987;
Turner & Hewitt, 1992), this model may be p, articularly important. This
approach has been applied to the relation of Type A behavior and coronary
heart disease (Krantz & Durel, 1983)) and more recently, to the association
of hostility and disease (R. B. Williams, 1994). Future research is necessary
to clarify possible genetic influences on other personality-disease relations.
Illness Behavior Model
In contrast to the previous models, the illness behavior approach suggests
that personality does not actually affect illness, but that it influences
behaviors related to the subjective perception of physical health. This model


is derived from evidence indicating that objective health versus illness does
not fully explain illness behaviors such as health care utilization, symptom
reporting, work absenteeism, and self-medication (e.g., G. A. Kaplan &
Camacho, 1983; Kaplan & Kotler, 1985; Maddox & Douglas, 1973). On the
contrary, psychological variables strongly affect the likelihood that
individuals will attend to physiological sensations and perceive that they are
indicative of illness (Cioffi, 1991; F. Cohen, 1979; Pennebaker, 1982;
Watson & Pennebaker, 1989).
Fig. 9.6 depicts the potential effects of psychological variables on various
manifestations of illness behavior. Symptom reports, which are reliably but
weakly predictive of objective health outcomes (Idler, Kasl, & Lemke,
1990), provide the clearest example of the less than perfect relation between
illness behaviors and disease. Self-reports of physical symptoms are
influenced by various psychological factors, including health beliefs and

differences in focus of attention (Pennebaker, 1982). For example,
individuals higher in neuroticism are more likely to be concerned with
somatic sensations, and subsequently, to report symptoms (Costa & McCrae, 1985a, 1987; Watson & Pennebaker, 1989). Furthermore, behavioral
health indices such as health care visits may be influenced by somatic
perceptions and other psychological processes. This phenomenon may be
particularly relevant to studies comparing samples selected through medical
clinics to control groups solicited from the community. These samples may
be biased by psychological characteristics that relate to self-selection into
health care settings. Any observed association between a targeted personality
trait and illness may be confounded by the relation of the trait to health care
utilization behaviors.


The preceding discussion illustrates the distinction between actual illness
and illness behavior, and suggests the necessity of carefully evaluating the
methodology utilized in studying personality and disease. As discussed
previously in the section on validity threats, before the mechanism(s) by
which personality contributes to disease can be classified, the research must
demonstrate that illness-and not simply illness behavior-is influenced by
personality.
TRAITS LINKED TO HEALTH
The following sections review literature regarding several specific
personality traits studied as risk factors. The list is not exhaustive; some
traits discussed in this literature are not included. The criteria for including
traits are that research published in refereed outlets shows considerable
evidence of a prospective association of the trait with objective indicators of
serious illness, or that despite the lack of such evidence, the trait is widely
studied in the field. For each trait, a description of its usual measurement,
the findings linking it to illness outcomes, and theory and research regarding
the processes underlying this association are presented.

Type A Behatior
As noted earlier, the Type A behavior pattern (TABP) occupies a central
place not only in the modern literature on personality and health, but also in
the evolution of the larger fields of behavioral medicine and health
psychology. The recent history of research on the topic illustrates most of the
central conceptual and methodological issues in this area of research.


Assessment. Following M. Friedman and Rosenman's (1959) description of
an inaction-emotion complex” consisting of achievement striving,
competitiveness, excessive job involvement, time urgency, and easily
provoked hostility, two assessment procedures quickly achieved widespread
use. The Structured Interview (SI; Rosenman, 1978) is a semistandardized
interview intended to elicit a behavioral sample of the behaviors comprising
the TABP, or their relative absence (i.e., the Type B pattern). With sufficient
raining in administration and scoring, reliable ratings can be achieved. A
variety of studies indicate that valid ratings can be made with the procedure,
with the caveat that the style of administration can affect the quality of
ratings (Scherwitz, 1988). The second principal measure is the Jenkins
Activity Survey (JAS; Jenkins, Rosenman, & Zyzanski, 1974). This selfreport measure primarily assesses achievement striving, competitiveness, job
involvement, and hard-driving behavior. Unlike the SI, it does not
adequately sample individual differences in hostility. However, many years
of research with various versions of the JAS have indicated that it is a
reliable and valid measure of the other aspects of the TABP (Rhodewalt & T.
W. Smith, 1991).
Because of its availability in a large, prospective study of corontly risk, the
Framingham Type A scale (Haynes et al:, 1980) is also recognized as a
potentially important measure. However, it is poorly correlated with the SI
and is more closely associated with both neuroticism and symptom reporting
than either the SI or JAS (T. W. Smith, O'Keeffe, & Allred, 1989; Suls &

Marco, 1990). Thus, some of the association between the FTAS and illness
endpoints involving symptom reports (e.g., angina vs. myocardial infarction


or cardiac death) might involve the association between neuroticism and
symptom reports already described.
Notably, the convergence among the three principal measures is more
modest than would be expected if they are to be interpreted as reflecting a
single construct. Thus, a basic measurement concern-poor convergent
validity-limits much of the literature on the TABP (Rhodewalt & T. W.
Smith, 1991).
Assocition With Diseage. After nearly 20 years of cross-sectional and
prospective research, a panel of experts convened by the American Heart
Association concluded that the TABP was a robust risk factor for CHD
(Cooper, Detre, & Weiss, 1981), with Type As having about a twofold
greater risk than Type Bs. Several notable failures to replicate this relation
appeared soon after the expert panel's conclusion. These included not only
several prospective, multicenter studies (Shekelle, Gale, & Norusis, 1985;
Shekelle, Hulley, et al., 1985), but a long-term follow-up from the original
prospective study of the TABP (Ragland & Brand, 1988).
These and other negative endings (e.g., Barefoot, Peterson, et al., 1989)
called into question the status of the TABP as a risk factor, and prompted a
more fine-grained analysis of the broadly defined pattern. The negative
reports also prompted skepticism in both the medical and popular health
literature. However, skepticism about the TABP may have been both
premature and too general. In a carefully rendered quantitative review of this
literature, T. Q. Miller and his colleagues (1991) demonstrated that when it
is assessed via the SI as opposed to self-report methods, the TABP is indeed
a reliable risk factor for the subsequent development of CHD, even when the