1989,

PSYCHOLOGY,

Lawrence

Erlbaum

Associates,

/0(2),

101-117

Inc.

Control,

Perceived

of

An

Examination

It' 1989,

SOCIAL

APPLIED

AND

BASIC

Copyright

Humor,

Irrational

Beliefs,

and

Positive

Stress as Moderators of the Relation Between Negative Stress and Health A.

Anderson

Craig

indic that: (a) nega stre was dire rela to bo ps mode inter appr sign how thr of the w effec were actu Typ I e (d) cop hu pe rro work on the effec of stres on dise (T. H. Ho & R 19 Ra a Requ for repri shou be sent to Cra A. An De of Ps of

Missouri-Columbia

Rice

Arnoult

H.

Lynn

University

University

Coping humor, beliefs about personal control, irrational beliefs, and the occurrence of desirable events (positive stress) have been proposed as moderators of the effects of negative stress on psychological and physical health. The effects of these variables and of negative stressful life events on health were examined in a retrospective study of 159 college students. The results

and physical health as a main effect, but positive stress was not; (b) when statistical corrections designed to hold the overall Type I error rate at .05 were made, there were no significant negative stress by moderator interactions; (c) when less conservative statistical restrictions were used, four negative stress by actually

in the wrong

direction,

further

suggesting

(as in b) that

these

control, and irrational beliefs each yielded at least one significant main effect on health. The prevalence of TypeI errors within studies and between studies as a function of publication and reporting biases is also discussed.

6521

MO

Columbia.

Missouri-Columbia,

of

University

\.

Over the last 2 decades interest in the effects of life experiences on a person's health has blossomed. Much of this interest began with Holmes and Rahe's

AND AND ARN T Meye Smit Kjae & T H. Hol 196 Su . re depre and the com cold (e.g T. S. Ho & T H. Ho 19 . speci stres such as brea can (Ta Lic & W In this artic we addr two mai issu me an as (T. H. Holm & R 196 On the oth han co ah se no types of stres requ life cha (T. H. Ho & R 19 T a theor assu was refl in the ma sca de to as John & S 1978 Vin & S ieg 197 Th res of stu el of little direc effec on heal (Lin Dea & E 19 M Ed n & Y 1977 Ros & M 197 Sar arv et al. 19 V & ir unne simp In a g per life som im ive e w v have prim posi effe othe wil hav pri ne ef still other will have appr equ pos and A Impr asse of stre was one goa of ou res Bu de 102

have investigated

Paykel

the effects of life stress on health measures

et aI., 1969; Rahe,

1968) as well as psychological

as varied as

adjustments

to

ment of positive stress effects, and the premature acceptance of moderator models / of the stress/health relation for several . potentially moderating variables.

POSITIVE

STRESS

Consider first the issue of the existence and effects of positive stress. Because positive events in one's life (e.g., promotions, marriage) necessitate great changes as negative events, they may increase one's health problems (as well

as most

reinforcement

theories)

suggest

that

positive

events

can

have only positive effects on one's health. Initial research ignored this distinction between positive and negative stress, and assumed that both would adversely affect health because both

stress. Subsequent researchers recognized the need to examine the effects of each type of stress separately. and grouped life events into those that are primarily positive versus those that are primarily negative (e.g., Sarason. this type are somewhat

mixed, but generally

Selzer, 1975). There is, unfortunately,

an additional

suggest that positive stress has

methodological

problem

with

this

work. Classifying each life event as either positive or negative may be

is likely to have both positive and to

individual

one

from

will

degree

the

and

consequences,

vary

(e.g., marriage)

negative

given event of importance

another. Thus, the most sensitive tests of positive and negative stress effects require that both the positive and the negative effects of each life event be assessed for (and by) each person individually. the foregoing

argument

that

the failure

to find

positive

,,~ effects

may

ST AN HE resid in past appr to the ass of str we ex lit effec of posi stres for two reas Fir the me pr also exist with mea of neg stre Ye res of ne st Of cour a s case can be ma tron for exp po ev to be meas of stres is two Firs by mo acc as st Seco we prov a m sen test of the eff of po or on MO VA seffec A majo issu con eco sev psy va th m insu a p from the neg eff ers of str (M & The rang of prop mod is bot wid and fas A p a Abra Selig Sem & P 198 soc su (C et & H 1983 Coh W 198 obe Le M & S il 1984 copi hum (Ma L pe co ef be (Folk 1984 John & S 197 et aI. ar 19 an irrati belie (Elli 197 As men ear po str ha al been prop as a m (Co & H 19 La od et aI o these varia rang from wea to mo str W sth e 103

that may have contributed

to the lack of consistent

effects of positive stress

are

events

events.

Negative

positive

for

and

we

optimistic;

expect

plan

(particularly main effects) are fairly consistent. Second, there is a psychological difference between positive and negative stress that parallels success and failure differences in other domains. Basically, people tend to be

generally

unexpected

associated

with

and are typically

improved

health,

more disruptive.

rather

than

with

health

problems,

either

the

in

refining

goal

measures.

stress

from

health

of

predictability

the

to

expect

we

improve

&

our

Folkman,

Kanner,

Lazarus,

Thus,

1980).

directly (as a main effect) or as a moderator of negative life events (e.g.,

been

have

variables

of

number

A

relation.

stress/health

moderate

the

health both as a main effect and as a moderator.

proposed that may attenuate or exacerbate the deleterious effects of stress. For example, it has been proposed that having a particular sense of humor

the

to

sensitive

particularly

be

will

humor

of

sense

this

particular

coping

Lefcourt, 1983). That is, this sense of humor helps the person to cope with stressful events, so that they do not adversely affect health. Those who lack

deleterious effects of negative stress (according to the theory). In essence, these moderating theories predict that it is the interaction of stress and some second (moderating) variable that adversely affects health. listing

includes

hardiness

(Kobasa,

1979), attributional

style (Metalsky,

1980).

Theoretical

and empirical

support

for the moderator

view of each of

AND AND ARN T weak (and beca we coul not stud all the abo var at on M speci we were conc that var ha be ac as of nega stres All thre stud wer we con an valid In only one stud thou was the Co Hu Sc us as a no publi repo repl thes find Th we fel itw be sugg such a r How elat the two ma res art pu contr the relat was not diff from zer Al thi su a 104

four for study

primarily

because

having the predicted interactive adequate empirical support.

support

effects on the stress/health

is rather

relation without

Humor

control

personal

moderators

the

Of

four

potential

under

study

Coping

their empirical

(humor,

latter

convergent

in

different

a

in

assessed

Humor

was

each,

providing

way

beliefs, irrational beliefs, and positive stress), humor appears to have the most support. In all three of their studies, Martin and Lefcourt (1983) found significant negative stress by humor interactions in the prediction of mood disturbances. That is, the presence of high levels (relative to low levels) of humor effectively insulated subjects from the deleterious effects

measure of the proposed useful

to attempt

Personal

moderator.

to replicate

Furthermore,

the coping

human

to date there have been

results.

Control

1985b) and performance

edly demonstrating

1985a,

Arnoult,

&

Anderson

as

such

of

attributional

areas,

5tyle

(e.g.,

The support for personal control beliefs as a moderator of the negative stress/health relation is weaker. There is good reason to expect perceptions of control to moderate the stress/health relation. That is, people high in perceptions of control should show weaker relations between negative stress and health than people low in perceptions of control. Research in a variety under

moderator

aversive

effects

conditions

of personal

(e.g.,

control

Averill,

both

1973),

suffer

from

(for

significantly

both depression

and state anxiety);

for people

with an internal

zero

from

different

was

relation

stress/health

negative

the same statistical weakness. Johnson and Sarason (1978) showed that for people with an external locus of control (Le., low personal control) the locus of

moderating effect, it does not test it. The proper statistical test is whether the slope relating stress to health differs as a function of locus of control (Le., the interaction of locus of control and negative stress). Unfortunately, the authors did not report this test. Inspection of their reported data anlyses (the partial correlations) suggests that this interaction test would not have

ST AN HE TH been signi but with the app sta tes we ca be Lefc et al. (198 repo thre stu in wh the str relati was supp mod by con Ag int w An artic by Cald Pea and Ch (19 rep th ap this expe Thu we incl a m of per co be in e life even (e.g. Ellis 197 Wo & R 19 On m ic expe that peop with man irra bel suc as "E l o me, or I'm a f as a h ailu bein wo be mo um su to th How we were unab to loca any stu tes irr be as Kann & F198 The idea is tha olk peo ex lo of stress was less distr for thos who had bee ins to pe th 105

certain.

not tested. Inspection of the reported analyses suggests that these interaction tests also would be nonsignificant.

to

beliefs

control

expect

to

reason

theoretical

good

is

In

there

sum,

priate control by stress interaction results for both physical and psychological health. Because of other variables included in the study, 10 such interactions were tested; none were significant. moderate

our

the negative stress/health

relation,

but little empirical

support

for

study.

Irrational

Beliefs

There is considerable discussion in the clinical literature about the debilitating effects of irrational beliefs, especially in conjunction with negative

negative effects of negative stressors, such as failed social relationships.

a moderator

of the stress/health

relation.

Positive Stress

Finally, the positive stress as moderator view is a fairly recent one (Lazarus,

positive stress should be relatively unaffected by negative stress events. Only one study has been reported that is directly relevant. In a well-conducted correlational study, Cohen and Hoberman (1983) found a significant negative stress by positive stress interaction in predicting depression. Their results failed to find a significant interaction in predicting physical symptoms (though it approached significance). One other study is sometimes cited as support for the positive stress as moderator position. Reich and Zautra (1981) instructed a random subset of their subjects to engage in certain pleasurable activities over a 2-week period. Among other findings, these authors reported that engaging in pleasurable activities interacted with negative life events, such that negative pleasant

activities.

Although

this study is important

in many respects,

it is

AND AND AR T OV We selec for stud four pote mo va ba on th impo We hope to prov the nee add su fo stres Of cour at the outs of suc a p on mu rea (a w ro d i again repo null resu ("If we run it rig nex tim it wi an IAnde Type e (Gre 197 rror Tha is, on occ em re w and such error are muc mor like to res in pu th ar false we haste to ackn that the sam bia int on ou ow 1983 And New & S 19 An & S pe e Pratk Leip & B 198 sug tha su au Ty I e r exac these prob in psy Th in a d of ou ow is in the stres rela but we wer ope to the po th willi to fairl test hyp and to acc and rep nu re w that gene exis and that tend to pro Ty Ieac strr 106

However,

it

also

does

be

To

it

sure,

events.

life

positive

manipulate

not a solid test of the positive stress as moderator model. One obvious interpretational problem concerns the pleasurable activities manipulation.

manipulates perceptions of control or personal efficacy. Thus, it is not clear to what the moderating effect of the activity instructions should be attributed.

weakness

of empirical

moderator

support

view of coping

that the relevant

theories

for them

humor,

as well as their

control,

may be wrong.

irrational

obvious

beliefs,

potential

and

positive

Given the bias of most authors

the corresponding bias of editors and reviewers against their publication ("Maybe this author doesn't know how to conduct studies properly" or "What do null results tell us?"), the few supportive studies could be simple

suggest rejecting the null hypothesis when in fact the null hypothesis is true, Type II errors (Le., acceptance of null results when in fact the null hypothesis is

decisions

from

time to time).

1986) as well as philosophy

are

extremely

difficult

to

Research

on the perseverance

of science considerations

correct.

Greenwald's

of theories

(e.g.,

(1975)

(e.g.,

Greenwald,

discussion

and

models

are

models of humor, wrong.

Further,

control,

because

irrational theory

the

to

for

support

provide

interactive

hoped

we

hypothesis

null

the

biases

against

simulation model of prejudice against the null hypothesis pointed out the magnitude of TypeI error problems and clarified the various factors that

beliefs, and positive stress

development

depends

on

decided to attempt to publish our results regardless of the outcome, as long as certain methodological requirements (e.g., scale reliability) were met (cf. Greenwald, 1975).

Although one cannot protect against the reporting and publication biases

ST AN HE rejec the null hypo whe it is true at som rea lev (e .05). For exam if there are 20 inde test of som eff an th null hypo is true, it is quit like that one or mo the tes wi be signi at .05 leve simp by cha On pro to ad fo perfo (e.g. 20) to esta a m rea sig o The varia in this stud wer stre (po and neg ne irrati belie All varia asse by sel On iss th becau they share the sam affe tone not bec str aff m We feel that this is not a p seri pro art in ou stu fo prima reaso First our mai goa was to test mo mo of str and healt using type of mea (alt im v in e more susce to such abTha is, if suc aff ias re In sum, our attem to conc rep awit ncur of mo um relati in the stres dom is con kn 107

one can reduce the likelihood of such errors within a particular data set by adjusting significance levels to maintain the overall p level (probability of

this is to divide the criterion

p level (.05) by the number

of tests of the

criterion (e.g., .0025). This procedure, known . as the Bonferroni solution, is used throughout our analyses. mood,

physical

symptoms

(e.g.,

colds,

headaches),

coping

humor,

and

frequently arises in this kind of research concerns the accuracy of such self-reports. In particular, it could be that observed relations between negative stress and depressed mood (for instance) obtain artifactually

health

stress/psychological

the

expect

should

we

at

biases

were

work,

several cases) as have been frequently reported in the literature. Thus, if one accepts the dominant methodology in stress research, then one cannot fault our replication/extension study. Second, physical health symptoms as requested in this study should not be very susceptible to affect/memory response biases; one doesn't forget a cold experienced several weeks ago just because of a present positive mood state. Thus the stress/physical health relations should not be very susceptible to any such bias. Third, one might expect the stress/psychological health relations to be

relations to be stronger than the stress/physical health symptoms relations. Our data show just the opposite, despite the above response bias prediction and despite the higher reliabilities of the psychological health measures (to be reported later). Fourth, and most convincing to us in the design phase of this work, is empirical evidence demonstrating that such biases do not occur in this type of research. Specifically, Lakey and Heller (1985) showed that this frequently proposed response bias criticism of self-report stress/health research does not stand up empirically.

and methodologies

in the field.

AND AND ARN At the begin of an intro psy cla su co Stres We adap the Life Exp Su (Sa et al 19 nega impa The follo exa wa giv to illu th po bility that an even mig hav both pos and neg im Each item was rated twic onc for pos im an on fo ne colum for even not expe Thi col wa inc so th in th case of an on the imp sca we cou de w & C 1984 Thes wer Item 35, "M ost per illn or in an 108

T

METHOD

College students completed measures of life events, health, mood states, and three personality variables: perceived control, sense of humor, and irrational beliefs.

Subjects

Undergraduates at Rice University participated in this study as an in-class demonstration of stress-health relations. There were 89 male and 70 female students in the sample.

Procedure

the

following

measures.

instructed

we

this

In

lives.

their

on

or

positive

negative

impact

study,

This instrument consists of 60 life events that lead to change in the lives of those who experience them. Our adaptation chiefly concerns the instructions given subjects for responding to the items. The original instrument instructs subjects to check items that they experienced during the last 6 months and those experienced between 7 months and I year ago. Subjects also are asked to rate the degree to which each event experienced had either

subjects to rate each event experienced during the past year for positive and

When spend their both

people retire from work, they usually enjoy the freedom they have to their time as they like. That's positive. In many cases, though, they miss work or their friends at work. That's negative. So retirement can have positive and negative impact.

impact. extreme

the

subject

Ratings were on 4-point scales anchored at no impact (0) and impact (3). Subjects also were instructed to check a "didn't occur"

had

simply

overlooked

the

item.

Four of the original items were omitted, due to the possibility of confounding with the concurrent measures of health and mood (Schroeder

ST AN HE (MA Zuck 196 Zuc Lub Vo & V 19 a asses depr host and anx res Su w resul in preli anal The a s z-s co um MAA score was crea for use in all sub an (BOI Beck & B 197 whi con of 13 ite eck me cu summ z-sco com sick me wa cre fo su analy Inso was not rela to sick (r = . an so wa 06 ke as Hum The Cop Hum Sca (M & L 19 w e Note that the last two item mig pro an art co w the depe mea How wh ana we 109

"Other

recent

experiences

which

have had an

Adjective

List

Check

the

administered

We

Affect

Multiple

Affect.

open-ended Items 48-50, impact on your life."

This measure

consists

of lists of adjectives

grouped

into three subsca1es

instructed to indicate whether or not each adjective described how they had been feeling over the last month, in general. These three affective measures

correlated

highly with each other (average

Subjects

degree

also completed

the short

form

r

=

.71) and

produced

similar

of the Beck Depression

Inventory

of depression.

they

indigestion,

following:

the

of

each

experienced

had

week

the

often

during

past

Physical health. Subjects indicated how many colds they had had so far that semester (approximately 10 weeks). Subjects also were asked how headache,

and insomnia.

Incidences

of colds, indigestion,

headaches were significantly but moderately related (average r

and

= .27); a

was

stressful

with

cope

to

humor

use

person's

a

measures

of

events,

a separate dependent measure. Finally, subjects rated their overall health, ranging from very poor (I) to excellent (7), during the last 6 months as a measure of wellness.

administered.

Perceived control. Subjects indicated on 7-point scales, ranging from no control (1) to total control (7), the degree of control they had had during the last year over the following six aspects of their lives: living environment, relations with others, financial state, academic program, health, and mood.

was created,

based

on the work

beliefs irrational

of

measure

Irrational

A

dysfunctional)

(or

beliefs.

these items removed from the measure of control, virtually no change in results or conclusions appeared. Therefore, all items were retained in the analyses reported in this article.

of Ellis (1977) and of Woolfolk

and

AND AND ARN T items were (a) in orde to be trul hap I m pro tha I a th m u awere sOne in my chos care ucce or "I do me up" (d) on hig ta week after com of thes me the res of th st share with the subj acco by a d is sex; it was there drop from sub ana evide that the effec of neg stre on hea we mo by meas and for each the four pro mo as th Beca there are 20 resu inte tes (i.e tes of m rate was .05 requ adju sign lev to .00 Th re Ther no evid that any of the fou pro mo af One could argu our crite (.00 wa too sti giv th is some a p reas to exp neg rior stre by in to occu But, even if the truth is that mo eff do no oc w intera at .05. Furt if the nul hyp is tru we wo ex 110

Richardson (1978). Subjects 7-point scales ranging from

ra~ed their agreement with six such beliefs on totally disagree (1) to totally agree (7). The

adequate and achieving in most things I attempt; (b) everyone is basically equal and as good as anyone else, but in the real world you have a "market value" that determines most of your happiness; (c) in order to feel like a worthwhile person, I must tell myself things such as "I will be recognized as

or superior individuals can get the deepest kind of satisfaction out of their work; (e) I become upset when I fail to live up to the expectations that others have of me; (f) people must achieve their full potential for happy and worthwhile living, else they have little value as humans.

ical relations among the variables, stress-management strategies.

and

of

current

knowledge

about

RESULTS

Preliminary analyses revealed no consistent or reliable effects of subject

was

there

if

see

to

performed

was

of

series

A

regression

any

effects

analyses

Moderator

coping behavior, personal control beliefs, irrational beliefs, or positive stress. Such moderator effects would be seen as interactions between negative stress and the moderating variables. For each of the five health main effects of negative

stress and the moderator,

models),

solution

the Bonferroni

1, were that none of the interactions

stress/health

TypeI error

met this criterion.

relation.

surprised

significant

several

find

to

be

not

should

these

variables,

the negative

in Table

that the overall

we

presented

to insuring

and their interaction.

1

TABLE

Predic of Health Overa ~, Main Effec and Inter of Neg Stre and Fo Po M Copin Humo Note. Only effects sipific at .OS (unad are listed ModerGtor

PersonGI

Model

IrrGtionGI

Control

Beliefs

Positi"e

Stress

Heolth

R2

MetlSUTt!

N'"

.188

.001 .001

Sickness Wdlness

.187 .14S

.001 .001

Insomnia

.086

.03

BDI

MAACL

.181

clI"

R2

NxCH'

N'"

PC'

~N'"

P~

NxPS'

~N'"

.01

.219

.001

.001

.147

.001

.01

.02

.31S .223 .283

.001 .001 .01

.001 .01 .001

.162 .201 .109

.001 .001 .001

.01

.070

.OS

NxlB'

III'

NxPC"

.03

.03

.120

.001

.IS4 .192 .IOS

.001 .001 .001

.049

.04

.04

aMain effects of neaative stress, p < the value listed. bMain effects of proposed moderators, p < the value listed.