Toward Better Research on Stress and Coping
Richard S. Lazarus
University of California, Berkeley
In commenting in considerable detail on the four
main articles in the special section on stress and
coping, the author comes to two main conclusions:
First, there is an increasing amount of high quali-
ty research on stress and coping that suggests' the
field is finally maturing, and this research may help
reduce the long-standing gap between research and
clinical practice. Second, this research is increas-
ingly using badly needed research designs that have
not hitherto been sufficiently emphasized, such as
longitudinal or prospective designs, focused on
observations that are day-to-day, microanalytic,
and in-depth, and that are compatible with a ho-
listic outlook. The author also addresses the role
of positive emotion in coping, the concept of de-
fense as it is dealt with nowadays, and the task of
evaluating coping efficacy.
For more than 50 years, my
research and theoreti-
cal efforts have centered on the topics of stress,
the emotions, and the coping process. For this
reason and the fact that in their introduction Somer-
field and McCrae (2000, this issue) cite me as a lead-
er in this field, I am very happy to have been asked to
comment on the four articles in this section (Coyne &
Racioppo, 2000; Cramer, 2000; Folkman & Moskow-
itz, 2000; Teimen, Affieck, Armeli, & Carney, 2000).
I am pleased that the section brings an overview of
the current state of the literature before the entire
field, and I feel a strong obligation to serve this area
of the field in which I have been a pioneer.
A discussant of others' work should at the outset
indicate the biases that inform comments to come. In
my case, I have been plowing and seeding the field of
stress, coping, and the emotions since the late 1940s.
My frame of reference has always been an epistemo-
logical, ontological, and theoretical approach that
emphasizes individual differences, the cognitive-
motivational-relational concepts of appraisal and
coping, and a process-centered holistic outlook
(Lazarus,1999c, in press). The conceptual bottom line
of my approach is the relational meaning that an indi-
vidual constructs from the person-environment rela-
tionship. That relationship is the result of appraisals of
the confluence of the social and physical environment
and personal goals, beliefs about self and world, and
resources.
My commentary is organized around five topics,
most of which were raised by the contributors to the
section. Those topics are as follows: (a) the quality of
coping research; (b) the gap between clinical practice
and research; (c) research designs in coping research,
where I also discuss two research programs included in
this special section, namely, those of Tennen et al. (2000)
and Folkman and Moskowitz (2000); (d) the concept of
ego defense and Cramer's (2000) discussion of it; and
(e) the task of evaluating coping efficacy.
The Quality of Coping Research
A premise that appears again and again in this section
is that for quite a few years research has disappointed
malay who had high hopes it would achieve both fun-
damenlal and practical knowledge about the coping
process and its adaptational consequences. Although I
recently expressed such disappointment (Lazarus,
1993a, 1997, 1998a), I am now heartened by positive
signs that there is a growing number of sophisticated,
resourceful, and vigorous researchers who are dedi-
cated to the study of coping.
In light of the recent accomplishments by those
researchers, some of my previously dour, pessimis-
tic thoughts about the state of the art now seem
overstated. Therefore, one of the main themes of
my commentary is the presence of much greater
hope for the future prospects of this field. This more
optimistic position stands in contrast with what Coyne
and Racioppo (2000) have written, which seems to be
unduly negative. Although I agree with many of the
points they made, their methodological analysis dis-
appoints me for three main reasons.
First, they offer mainly a series of criticisms, some
of them unfair. This endeavor does more harm than
Editor's note. Mark R. Somerfield and Robert R. McCrae developed
this Psychology in the Public Forum section.
Author's note. Richard S. Lazarus, Department of Psychology, Uni-
versity of California, Berkeley.
Cor'respondence concerning this article should be addressed to
Richard S. Lazarus, 1824 Stanley Dollar Drive #3B, Walnut Creek, CA
94595-2833.
June 2000
•
American Psychologist
Copyright 2000 by the American Psychological Association, Inc. 0003-066X/00/$5.00
Vol. 55. No. 6, 665 673 DOI: 10.1037//0003-066X.55.6.665
665
good. Second, they provide little in the way of solu-
tions. For example, to describe research in this field,
they use the extraordinarily denigrating image of a
slum that should be razed rather than reconstructed.
Analyses in which authors are intent on damning the
whole coping research enterprise by not clearly distin-
guishing the wheat from the chaff are not needed;
plenty of good material can increasingly be found.
Coyne and Racioppo's (2000) overzealousness is
exemplified in their discussion of coping question-
naires. I speak as one who has developed a coping
questionnaire (Folkman & Lazarus, 1988). I agree that
questionnaires should be thought of as an initial rather
than a final step toward understanding and that they
do not allow psychologists to go below the surface to
identify goals and situational intentions, especially
those of which the individual is unaware. Neverthe-
less, they permit the study of large samples and the
quantification of the coping process, which under some
circumstances is useful and important.
Coyne and Racioppo's (2000) overzealousness is
also exemplified by their discussion of confounding.
Although this can, indeed, be a problem in coping
research, they oversimplify and overemphasize the
problem of confounding. For example, they present a
strong warning about the risks of confounding, but
those risks are hardly as great as portrayed. The dan-
ger of confounding is that measures of coping could
contain some of the same variables--for example, dis-
tress or psychopathology--as the outcome measure of
mental health. Thus, if the antecedent and consequent
measures are essentially the same, any correlation be-
tween them would represent some degree of tautology
rather than a causal explanation. Without specifying
where the confounding lies in any particular study and
without data demonstrating confounding, however, any
assessment of risk is much too facile and misleading.
What Coyne and Racioppo (2000) fail to mention
in their warning is that confounding is a difficult epis-
temological issue about which scientifically oriented
scholars disagree. If a simple answer exists, it is that
one should not draw implications about the risk of
confounding without data to demonstrate that risk,
and one should do so only if there is little or no room
remaining for independent variance after the so-called
confound has been eliminated. Coyne and Racioppo's
heavily emphasized warning about this risk is greatly
overdrawn. It is little more than supposition, based on a
superficial impression that could easily be erroneous.
The warning, which at first blush may sound like a wise
defense of science, merely masquerades as wisdom.
The issue of confounding has always been trou-
bling for psychology from a measurement standpoint.
This was evident some years ago in a criticism made
by Dohrenwend, Dohrenwend, Dodson, and Shrout
(1984) that the Berkeley Stress and Coping Project's
Hassles Scale, which first appeared in 1981 and was
later given to a test publisher (see Lazarus & Folkman,
1989), contained the same variable, psychopathology,
that was also present in the outcome measure, psycho-
logical symptoms.
Lazarus, DeLongis, Folkman, and Gruen (1985)
responded to this criticism by initially examining the
issue of confounding from a broad scientific perspec-
tive. For example, they noted that correlational re-
search involving social and psychological processes
is likely to be characterized by some degree of circu-
larity, which is very difficult to avoid. This circularity
should not be fatal to the validity of the research
conclusions, however, as long as a reasonable amount
of the variance remains after the confound has been
eliminated. In effect, confounding is relative and an
empirical question.
Lazarus et al. (1985) also reported a new study in
which many of the items of the Hassles Scale were
rewritten to free it of the supposed confounding. The
substantial correlations between hassles and psycho-
logical symptoms were not greatly changed by this
manipulation, and they remained significant at the
.001 level. Therefore, the impression of confounding
suggested by Dohrenwend et al. (1984) was not borne
out by the data.
Second, Coyne and Racioppo (2000) ignore what
many others have said on the topic of coping research.
In their introduction, Somerfield and McCrae (2000)
note a number of such critiques, including two of
mine (Lazarus, 1993a, 1997). Although Coyne and
Racioppo cite the 1997 article, they fail to acknowl-
edge that I had expressed many of the same method-
ological concerns they present here. I could easily
agree with many of the concerns if they were directed
at only the poorest research on coping. These con-
cerns remind us of what psychologists should take into
account when looking at coping research. Stated as they
are, however, they often fail to constitute a properly
balanced rendition of the methodological issues.
Third, and far more damaging to an accurate por-
trait of today's coping research, is that what Coyne
and Racioppo (2000) say about an overreliance on
coping questionnaires ignores the growing number of
important current coping research programs that do not
depend on such questionnaires. Their failure to cite this
research can mislead readers who are not familiar with
the literature on stress, coping, and emotion. Consider,
for example, the current and continuing research on the
caregivers of partners suffering and dying of AIDS by
Folkman and her colleagues, which is referred to in
Folkman and Moskowitz's (2000) article in this sec-
tion. Although these researchers used the Ways of
Coping Questionnaire (Folkman & Lazarus, 1988), their
main source of data included detailed and in-depth
interviews and observations.
Their longitudinal research design and careful
observations reveal in great detail powerful positive
and negative emotional reactions of the caregivers
before and after the tragic death of the men they cared
666 June 2000 • American Psychologist
for. These methodological virtues also made it possi-
ble to evaluate defensive and motivational features of
the coping process. My high regard for this research is
reinforced by widespread international praise for Folk-
man's research and ideas. It is, in my view, a first-class
example of what can be done in coping research, but
Coyne and Racioppo (2000) make no mention of it.
Coyne and Racioppo's (2000) criticism about
questionnaires also does not apply to the research of
Tennen et al. (2000), whose research I consider later in
this article along with that of Folkman and Moskowitz
(2000). In both these cases--Folkman and Moskowitz
and Tennen et al.--their research has been ongoing
for some years and published previously, so Coyne
and Racioppo could easily have referred to it favor-
ably. I found it easy to praise a number of other studies
(Lazarus, 1999c) that were described in two recent
edited books: one by Gottlieb (1997) that deals with
chronic stress, the other by Eckenrode and Gore (1990)
on the spillover of work to the family.
My main point is that although there is plenty
of unhelpful research to complain about, there is
also a substantial amount of promising solid--even
creative--work on stress and coping. The fact that
Coyne and Racioppo (2000) ignore these worthy ex-
amples and the published statements of other serious
scholars seems unfortunate and misleading.
Gap Between Clinical Practice and
Research
Coyne and Racioppo (2000) criticize the failure of
stress and coping research to be useful in clinical
practice. Many psychologists in this field did, indeed,
hope that their research would have practical clinical
value, and the studies I have been applauding are, in
fact, quite relevant to clinical practice. However, as
Somerfield and McCrae (2000) point out in their intro-
duction, there has also been a failure on the part of
clinicians to acknowledge the complexity of the
task of making the research applicable. Many vari-
ables are involved in how people cope and the out-
comes of coping. These include the different kinds
of stress and their details, such as whether a loved
one died in dragged-out misery or quickly and un-
expectedly; individual differences in personality
traits and resources; the interpersonal and cultural
context; and the diverse criteria that should be used for
evaluating the success of coping efforts, such as subjec-
tive well-being, somatic health, and criteria based on
societal values. Given the great scope and difficulty of
the task, can anyone really believe that the task of under-
standing what people are like and how change can be
promoted is a simple one to be achieved overnight, so to
speak, or in a few research programs?
The lack of collaboration and communication be-
tween researcher and clinician, another criticism made
by Coyne and Racioppo (2000), is a familiar and pain-
ful topic for most psychologists. This highly complex
and political issue has been previously addressed with
wisdom by others (e.g., Beutler, Williams, Wakefield,
& Entwistle, 1995). It is disheartening that so few
researchers accept the responsibility of making the
relevance of their research clear to the practitioner,
and so few clinicians pay attention to such research
even when it has implications for clinical practice. The
gap between practitioner and researcher has a long
and un~brtunate history, with some psychology de-
partments refusing to offer clinical training and oth-
ers, such as my own, having experienced bitter strug-
gles over this issue in the past (for a more detailed
look at the Berkeley struggles, see Lazarus, 1998a,
1998b). It is also illustrated by the separate existence
of the American Psychological Association and the
American Psychological Society. This history undoubt-
edly undermines what could be useful collaboration
between researchers and clinicians in advancing our
understanding of the stress and coping process.
Research Designs
In the upcoming text, I highlight certain research
methods--for example, longitudinal, prospective,
and microanalytic approaches; in-depth observa-
tion; and holism--all of which I consider essential,
if not just advantageous, for the study of stress, emo-
tion, and coping. I also draw on Tennen et al.'s (2000)
and Folkman and Moskowitz's (2000) research as con-
crete illustrations. I may seem to digress briefly as I set
the stage by first speaking generally about research.
Analysis and Synthesis (and Holism)
The prime objective of basic research is to understand
life and the world in which it exists. This understand-
ing can then be validated, in part, by its ability to help
psychologists predict human reactions under diverse
conditions and demonstrate its utility in practical ap-
plications. There are two main ways to understand a
phenomenon. One is the standard scientific approach,
which these days depends almost entirely on analysis,
and in which complex phenomena are broken down
into smaller, presumably more elemental explanatory
variables, each of which is only a part of the whole.
This canon has been revered as the standard cause-
and-effect research style of reductive science, which
includes psycholOgy. This approach has allowed re-
searchers to achieve impressive knowledge in the
physical and biological sciences, and it provides a
considerable degree of practical control over the
world, although I am not convinced it is ideal for
many of the most important topics of psychology.
Although an analytic, cause-and-effect episte-
mology aims at exploring the functional connections
among the component parts, it also fractionates the
phenomena with which the researchers began, thereby
limiting understanding, especially when the parts are
treated as if they were the whole. An important step is
often missing, namely, the effort to reconstruct the
June 2000 • American Psychologist 667
whole so that the phenomena under study are restored
to the form in which they appear in nature. This deficit
has led many psychologists to advocate another ap-
proach to understanding, not as a substitute for the
analytic approach but as a necessary complement. Its
essence is the full and accurate
description
of phenome-
nal wholes, which is what is observed in nature and
conceptualized with abstract categories constructed by
researchers themselves; detailed description is as im-
portant to science as is the search for causal variables.
Here psychologists might follow the lead of John
Dewey (1894, 1896; Dewey & Bentley, 1949), the
philosopher-psychologist, who argued for the impor-
tance of
synthesis.
When psychologists study stress,
emotion, and coping, they want to accurately portray
the behavioral display and the experience of emotions
and to say what they are like. To understand stress,
emotion, and coping fully, psychologists need to take
into account both levels of abstraction, that is, the
parts they are treating as causes and the organized (or
synthesized) wholes that comprise the parts.
Thus, psychological stress can be thought of as
part of a complex, organized biosocial-psychological
entity or whole, which psychologists refer to as an
emotion,
such as anger, fear, shame, joy, or love. Emo-
tions are broader, more inclusive concepts than stress
in that they comprise both positive and negative emo-
tional states. Coping is an integral part of an emotion,
but it is not the whole. The methods one uses to study
emotion and its component parts depend on which of the
two modes of scientific understanding one favors, anal-
ysis or synthesis. Although they complement each other,
their scientific tasks are sufficiently distinctive to re-
quire different lines of thought and research methods.
The task of analytic research--that is, to identify
causal variables and show how they work--requires a
timeline of antecedents and consequences, as in John
Stewart Mill's (1843/1949) logical canons of experi-
mentation. If the appropriate logic is followed, it be-
comes possible to demonstrate that variable A is a
cause of variable B by showing that if A is not present
B does not occur. Nevertheless, from a broader point
of view, the timeline must not be viewed as rigidly
fixed; rather, a more modem recursive frame of refer-
ence should be adopted, in which any of the variables
can serve as an antecedent, a mediator, a moderator, or
a consequence, although not at exactly the same mo-
ment. Cross-sectional research can only demonstrate a
correlation between A and B; it cannot prove causali-
ty. For causation in stress and coping to be under-
stood, longitudinal research is needed.
Longitudinal (or Prospective) Research
Prospective longitudinal research allows researchers
to try to predict later events from measures obtained
earlier. This requires repeated measurements on the
same persons who are observed from Time I to Time N
and across diverse circumstances. What distinguishes
longitudinal from cross-sectional research is not the
duration or size of the study. Rather, the research
design must be within-subjects and prospective (Laz-
arus, 1999c). This kind of research allows researchers
to identify psychological structures, that is, stable per-
sonality dispositions (or traits). It also allows the iden-
tification of changes (or processes) in psychological
reactions over time and diverse conditions. Both are
important.
The dilemma created for the researcher is that so
much time and energy must be spent obtaining repeat-
ed measures with the same persons that the size of the
participant sample is inevitably limited by the cost of
obtaining those repeated measures. This limitation
makes it all but impossible to select a representative
sample that would permit secure generalizations about
the average person or particular classes of persons.
With an
ipsative-normative
version of longitudi-
nal research, researchers seek a solution to this dilem-
ma by selecting a particular sample, albeit a limited
one, while recognizing that generalizations to other
samples cannot be depended on without additional
studies using different samples. There is no way out of
the dilemma: To adhere to one value, one has to sacri-
fice the other, at least temporarily. Ipsative-normative
research designs permit the delineation of what the
persons being studied are like individually and how
they feel, think, and act, which is revealed by repeated
measurement and an examination of the ways and
conditions under which they relate to the world.
Ipsative
refers to within-person comparisons, and
normative
refers to comparisons between- or among-
persons. Either perspective can be dominant, and they
complement each other. For those who are especially
interested in the rationale involved here, I know of no
clearer discussion of this than was published long ago
by Broverman (1962).
Exemplary Research Designs
Tennen et al. (2000) summarize research that uses
both a within-persons and an across-persons research
design--in effect, an ipsative-normative version of
longitudinal research--to study the clinical problems
of pain, depression, and alcohol consumption in a
programmatic effort to link theory, research, and clin-
ical practice. A highly desirable feature of their re-
search is the emphasis on day-to-day variations, that
is, changes (processes) that take place over time and
conditions. Their research is also microanalytic in that
they look closely at the details of what is happening
intrapersonally as well as interpersonally.
Tennen et al.'s (2000) article provides an excel-
lent example of how to conduct research that follows
the epistemological and ontological positions I identi-
fied earlier. The focus is on individual differences,
cognitive-motivational-relational meaning-centered
mediation (e.g., appraising), a longitudinal (or pro-
spective) research style, and an effort to obtain
668 June 2000 • American Psychologist
microanalytic data in a framework that is process-
centered and holistic. Notice that research of this
kind is most like the assessment and treatment proce-
dures used in clinical practice, but it also draws on a
sufficient number of participants to pen-nit generaliza-
tions beyond a sample size of one. I consider these
outlooks and methods to be hallmarks of the best short-
term research on stress, coping, and the processes
whereby emotional reactions occur and affect social
functioning. Long-term research of this sort is needed
for the study of stress, coping, and emotions as factors
in health and illnesses, such as cancer and heart dis-
ease, which take a long time to develop and emerge.
The logical and empirical case Tennen et al. make for
an emphasis on within-persons comparisons and an
intensive focus on day-to-day processes is impressive.
Their data also confirm the utility of both within-
persons comparisons and day-to-day analyses.
Tennen et al. (2000) point out that the questions
addressed by between-persons research designs are
entirely different from those addressed by within-
persons designs. Their own findings on coping and
alcohol consumption differ when analyzed from
one or the other of these two perspectives, echoing
Epstein's (1983) earlier observation that correlations
among emotions vary depending on whether they are
studied in the same person across occasions or in
different persons on the same occasion.
Because these two levels of analysis address very different
questions, we are not surprised by this discordance. What
does surprise us, however, is that investigators continue
their attempts to answer inherently within-person ques-
tions regarding stress, coping, and psychological adapta-
tion with between-person research designs and analytic
strategies. (Tennen et al., 2000, p. 628).
This principle is also instantiated in Tennen et
al.'s (2000) findings about the influence of depression
on the experience of pain. If depression and pain are
studied with a between-persons research design, per-
sons who are depressed report more pain than do those
who are not depressed. They also engage in more pain
catastrophizing and believe that their coping strate-
gies were relatively inefficacious. So far so good from
a normative standpoint.
However, within-person analysis shows that pa-
tients with a history of depression are less able to
inhibit pain catastrophizing the day after a good night's
sleep compared with those who had never been de-
pressed. What is striking, in effect, is that a day-to-day
examination of pain reveals more accurately than be-
tween-person analysis the extent to which depression
affected the pain experience. It also shows that pain
varies greatly depending on what takes place in the
coping process from day to day. The conclusion Ten-
nen et al. (2000) reach, that "a pain diary may yield a
different picture of the pain experience than would a
summary obtained at an office visit" (p. 631), is very
important to the future development of the stress and
coping literature.
But what about coping? A tendency has emerged
in coping research to pit the problem-focused and
emotion-focused coping functions against each other
to compare their respective efficacy. However, this is
a strategic mistake, as I said in Lazarus (1993b, 1999c).
It is misleading to separate these two functions of
coping and compare their efficacy because although
conceptually distinguishable, both strategies are in-
terdependent and work together, one supplementing
the other in the overall coping process. The distinc-
tions between them should not be taken too literally by
comparing their individual capacities to influence ad-
aptational outcomes. For a picture of how people cope,
psychologists need to study how both functions, and
perhaps the balance between them, work and affect
each other and the adaptational outcome: in effect,
how they operate as a single coping unit.
The studies of daily coping with rheumatoid ar-
thritis by Tennen et al. (2000) confirm the utility of
this ontological position. Their data show that the two
functions of coping, problem- and emotion-focused,
usually occur together as they did in the early research
of Lazarus and Folkman (1984). What is even more
revealing in their data is that when they compared the
relative probability of the occurrence of emotion-
focused coping, the likelihood was 4.4 times greater
on a day when problem-focused coping had occurred
than on a day without it. Here, intraindividual analysis
of day-to-day coping provides a far richer and more
provocative picture of how coping works than when
the two functions of coping are treated as independent
and in competition with each other with respect to
their adaptational consequences. Referring to these
findings, they conclude:
Not only was today's emotion-focused coping predicted by
yesterday's problem-focused coping, but this cross-day asso-
ciation was itself a function of the change in pain from
yesterday to today. An increase in today's pain over yester-
day's pain increased the likelihood that problem-focused cop-
ing yesterday would be followed by emotion-focused coping
today. In other words, when efforts to directly influence pain
are not successful (as evidenced by an increase in next-day
pain), the next day people may try harder to adjust to that
which cannot be readily changed. At the risk of belaboring
the obvious, we stress that these associations cannot be ascer-
tained through cross-sectional or several wave longitudinal
designs. We suspect that they would evade in-depth inter-
views . . . or, worse, that participants would reconstruct
associations on the basis of unexamined heuristics, personal
theories, or a single recent or particularly distressing coping
encounter. (pp. 632 633)
Given that psychologists have been presented with
mainly cross-sectional coping research in the past,
this analysis becomes important methodologically as
June 2000 • American Psychologist 669