ArticlePDF Available

Bad Is Stronger than Good

Authors:

Abstract

The greater power of bad events over good ones is found in everyday events, major life events (e.g., trauma), close relationship outcomes, social network patterns, interpersonal interactions, and learning processes. Bad emotions, bad parents, and bad feedback have more impact than good ones, and bad information is processed more thoroughly than good. The self is more motivated to avoid bad self-definitions than to pursue good ones. Bad impressions and bad stereotypes are quicker to form and more resistant to disconfirmation than good ones. Various explanations such as diagnosticity and salience help explain some findings, but the greater power of bad events is still found when such variables are controlled. Hardly any exceptions (indicating greater power of good) can be found. Taken together, these findings suggest that bad is stronger than good, as a general principle across a broad range of psychological phenomena.
Review of General Psychology
2001.
Vol. 5. No. 4. 323-370Copyright 2001 by the Educational Publishing Foundation
1089-2680/O1/S5.O0 DOI: 10.1037//1089-2680.5.4.323
Bad Is Stronger Than Good
Roy F. Baumeister and Ellen Bratslavsky
Case Western Reserve UniversityCatrin Finkenauer
Free University of Amsterdam
Kathleen D. Vohs
Case Western Reserve University
The greater power of bad events over good ones is found in everyday events, major life
events (e.g., trauma), close relationship outcomes, social network patterns, interper-
sonal interactions, and learning processes. Bad emotions, bad parents, and bad feedback
have more impact than good ones, and bad information is processed more thoroughly
than good. The self is more motivated to avoid bad self-definitions than to pursue good
ones.
Bad impressions and bad stereotypes are quicker to form and more resistant to
disconfirmation than good ones. Various explanations such as diagnosticity and sa-
lience help explain some findings, but the greater power of bad events is still found
when such variables are controlled. Hardly any exceptions (indicating greater power of
good) can be found. Taken together, these findings suggest that bad is stronger than
good, as a general principle across a broad range of psychological phenomena.
Centuries of literary efforts and religious
thought have depicted human life in terms of a
struggle between good and bad forces. At the
metaphysical level, evil gods or devils are the
opponents of the divine forces of creation and
harmony. At the individual level, temptation
and destructive instincts battle against strivings
for virtue, altruism, and fulfillment. "Good" and
"bad" are among the first words and concepts
learned by children (and even by house pets),
and most people can readily characterize almost
any experience, emotion, or outcome as good or
bad.
What form does this eternal conflict take in
psychology? The purpose of this article is to
review evidence pertaining to the general hy-
Roy F. Baumeister, Ellen Bratslavsky, and Kathleen D.
Vohs,
Department of Psychology, Case Western Reserve
University; Catrin Finkenauer, Department of Psychology,
Free University of Amsterdam, Amsterdam, the Netherlands.
Ellen Bratslavsky in now at the Department of Psychol-
ogy, Ohio State University.
We thank the many people who have contributed helpful
comments and references. This work is dedicated to the
memory of Warren.
Correspondence concerning this article should be ad-
dressed to Roy F. Baumeister or Kathleen D. Vohs, Depart-
ment of Psychology, Case Western Reserve University,
10900 Euclid Avenue, Cleveland, Ohio 44106-7123. Elec-
tronic mail may be sent to either rfb2@po.cwru.edu or
kdv3@po.cwru.edu.
pothesis that bad is stronger than good (see also
Rozin & Royzman, in press). That is, events
that are negatively valenced (e.g., losing
money, being abandoned by friends, and receiv-
ing criticism) will have a greater impact on the
individual than positively valenced events of
the same type (e.g., winning money, gaining
friends, and receiving praise). This is not to say
that bad will always triumph over good, spelling
doom and misery for the human race. Rather,
good may prevail over bad by superior force of
numbers: Many good events can overcome the
psychological effects of a single bad one. When
equal measures of good and bad are present,
however, the psychological effects of bad ones
outweigh those of the good ones. This may in
fact be a general principle or law of psycholog-
ical phenomena, possibly reflecting the innate
predispositions of the psyche or at least reflect-
ing the almost inevitable adaptation of each
individual to the exigencies of daily life.
This pattern has already been recognized in
certain research domains. This is probably most
true in the field of impression formation, in
which the positive-negative asymmetry effect
has been repeatedly confirmed (e.g., Anderson,
1965;
Peeters & Czapinski, 1990; Skowronski
& Carlston, 1989). In general, and apart from a
few carefully crafted exceptions, negative infor-
mation receives more processing and contrib-
323
324BAUMEISTER, BRATSLAVSKY, FINKENAUER, AND VOHS
utes more strongly to the final impression than
does positive information. Learning something
bad about a new acquaintance carries more
weight than learning something good, by and
large.
In other spheres, the effect seems present but
not recognized. For example, nearly every psy-
chology textbook teaches that propinquity
breeds attraction. This conclusion is based on
the landmark study by Festinger, Schachter, and
Back (1950) in which the formation of friend-
ships in a married students' dormitory was
tracked over time. Contrary to elaborate hypoth-
eses about similarity, role complementarity,
values, and other factors, the strongest predictor
of who became friends was physical propin-
quity: Participants who lived closest to each
other were most likely to become friends.
Yet a lesser known follow-up by Ebbesen,
Kjos,
and Konecni (1976) found that propin-
quity predicted the formation of disliking even
more strongly than liking. Living near one an-
other increased the likelihood that two people
would become enemies even more strongly than
it predicted the likelihood that they would be-
come friends. Propinquity thus does not cause
liking. More probably, it simply amplifies the
effect of other variables and events. Because
bad events are stronger than good ones, an
identical increase in propinquity produces more
enemies than friends.
The relative strength of bad may also be
relevant to the topics studied by research psy-
chologists. As president of the American Psy-
chological Association, Martin Seligman (1999)
called for a "positive psychology" movement to
offset the negative focus that he saw as domi-
nating most of psychology's history. The nega-
tive focus was first documented by Carlson's
(1966) survey of psychology textbooks, in
which he found twice as many chapters (121 vs.
52) devoted to unpleasant as to pleasant emo-
tions,
and a similar imbalance was found in
lines of coverage and use of specific words.
More recently, Czapinski (1985) coded more
than 17,000 research articles in psychology
journals and found that the coverage of negative
issues and phenomena exceeded positive, good
ones 69% to
31%,
a bias that was fairly strong
across all areas of psychology (although weak-
est in social psychology). Seligman is probably
quite right in proposing that psychologists have
focused most of their theoretical and empirical
efforts on understanding the bad rather than the
good.
Why has this been so? Undoubtedly, one
hypothesis might be that psychologists are pes-
simistic misanthropes or sadists who derive per-
verse satisfaction from studying human suffer-
ing and failure. An alternative explanation,
however, would be that psychology has con-
sisted of young researchers trying to obtain pub-
lishable findings in a relatively new science that
was characterized by weak measures and high
variance. They needed to study the strongest
possible effects in order for the truth to shine
through the gloom of error variance and to
register on their measures. If bad is stronger
than good, then early psychologists would in-
evitably gravitate toward studying the negative
and troubled side of human life, whereas the
more positive phenomena had to wait until the
recent emergence of stronger methods, more
sensitive measures, and better statistical
techniques.
The goal of this review is to draw together the
asymmetrical effects of bad and good across a
deliberately broad range of phenomena. Even in
topic areas in which this asymmetry has been
recognized (as in impression formation), re-
searchers have not generally linked it to patterns
in other topic areas and may therefore have
overlooked the full extent of its generality. The
present investigation is intended to provide
some perspective on just how broadly valid it is
that bad is stronger than good. We certainly do
not intend to claim that the greater power of bad
things overrides all other principles of psychol-
ogy. Other relevant phenomena may include
congruency effects (good goes with good; bad
goes with bad) and self-aggrandizing patterns
(bad can be avoided or transformed into good).
Nevertheless, the general principle that bad is
stronger than good may have important impli-
cations for human psychology and behavior.
Definition implies rendering one concept in
terms of others, and the most fundamental ones
therefore will resist satisfactory definition.
Good,
bad, and strength are among the most
universal and fundamental terms (e.g., Cassirer,
1955;
Osgood & Tzeng, 1990), and it could be
argued that they refer to concepts that are un-
derstood even by creatures with minimal lin-
guistic capacity (such as small children and
even animals). By good we understand desir-
able,
beneficial, or pleasant outcomes including
BAD IS STRONGER THAN GOOD 325
states or consequences. Bad is the opposite:
undesirable, harmful, or unpleasant. Strength
refers to the causal impact. To say that bad is
stronger than good is thus to say that bad things
will produce larger, more consistent, more mul-
tifaceted, or more lasting effects than good
things.
A Brief Discussion: Why Should Bad Be
Stronger Than Good?
Offering an explanation for the greater power
of bad than good is likely to be an inherently
difficult enterprise. The very generality of the
pattern entails that there are likely to be few
principles that are even more broad and general.
Meanwhile, researchers will have found lower
level explanations that help explain why bad
may be stronger than good with regard to spe-
cific,
narrowly defined phenomena.
From our perspective, it is evolutionarily
adaptive for bad to be stronger than good. We
believe that throughout our evolutionary his-
tory, organisms that were better attuned to bad
things would have been more likely to survive
threats and, consequently, would have increased
probability of passing along their genes. As an
example, consider the implications of foregoing
options or ignoring certain possible outcomes.
A person who ignores the possibility of a pos-
itive outcome may later experience significant
regret at having missed an opportunity for plea-
sure or advancement, but nothing directly terri-
ble is likely to result. In contrast, a person who
ignores danger (the possibility of a bad out-
come) even once may end up maimed or dead.
Survival requires urgent attention to possible
bad outcomes, but it is less urgent with regard to
good ones. Hence, it would be adaptive to be
psychologically designed to respond to bad
more strongly than good. After we review the
evidence for the phenomenon of bad being
stronger than good, we present a more complete
discussion of the theoretical reasons for the
strength of bad over good and also review other
theories that have been proposed in the context
of specific subareas (e.g., impression formation).
Evidence
The purpose of the following sections is to
review evidence pertaining to the central hy-
pothesis that bad is stronger than good. To
establish the breadth of the pattern, we try to
identify many seemingly different and diverse
spheres in which bad is stronger than good.
Given the breadth of the hypothesis, it is prob-
ably not possible to cover every study that has
ever found bad to be stronger than good in any
sphere. We have, however, tried to cover as
much as possible and to provide evidence for
the effect in as many different spheres as pos-
sible.
How did we accomplish this? Unlike
more focused narrative reviews or meta-analy-
ses,
we were unable to conduct a systematic
search using keywords such as good or bad.
Instead, we made an effort to cast as broad a net
as possible and then focus our search on several
research areas. As part of this process, we made
a request via e-mail to the members of the
Society for Personality and Social Psychology
list-serve. The roughly 100 responses received
from these members served as a starting point
for our search. After dividing our review into
several topic areas, we then set out to uncover
those studies that compared the relative strength
of good and bad effects, especially those that
also included a neutral control group.
The central goal of this review is to establish
convergence across multiple areas. The consis-
tency of conclusions across each area is more
important than the robustness or methodologi-
cal strength of evidence in each specific area.
We attempt to do justice to each area, but our
emphasis is on breadth (and on the quest for any
patterns in the opposite direction), so it seemed
desirable to cover as many different areas as
possible.
Reacting to Events
All lives contain both good and bad events. If
bad is stronger than good, then the bad events
will have longer lasting and more intense con-
sequences than good events. In particular, the
effects of good events should dissipate more
rapidly than the effects of bad events. This
should occur despite the mechanisms described
by Taylor (1991), by which many people strive
to minimize bad events and distance themselves
from them, although those minimizing pro-
cesses should limit the impact of bad events and
possibly produce some contrary findings.
A widely accepted account of the impact of
life events was put forward by Helson (1964)
326BAUMEISTER, BRATSLAVSKY, FINKENAUER, AND VOHS
and called adaptation level theory. In this view,
the impact of substantial changes in life circum-
stances is temporary. People (and animals) react
more to changes than to stable conditions, so
they are most sensitive to new conditions.
Change, therefore, produces strong reactions,
but the circumstances that result from the
change gradually cease to elicit a reaction and
eventually become taken for granted. Applying
this theory to human happiness, Brickman and
Campbell (1971) postulated a "hedonic tread-
mill" by which long-term happiness will remain
roughly constant regardless of what happens
because the impact of both good and bad events
will wear off over time.
In testing the hedonic treadmill, however, it
emerged that bad events wear off more slowly
than good events. Brickman, Coates, and
Janoff-Bulman (1978) interviewed three groups
of respondents: people who had won a lottery,
people who had been paralyzed in an accident,
and people who had not recently experienced
any such major life event. The lottery wins and
accidents had occurred about 1 year before the
interview. Confirming the hypothesis for posi-
tive events, the lottery winners did not report
greater happiness than the two other groups.
Brickman et al. proposed that this result was
due to habituation, as the adaptation level phe-
nomenon would predict: The euphoria over the
lottery win did not last, and the winners' hap-
piness levels quickly returned to what they had
been before the lottery win. Ironically, perhaps,
the only lasting effect of winning the lottery
appeared to be the bad ones, such as a reduction
in enjoyment of ordinary pleasures.
In contrast to the transitory euphoria of good
fortune, the accident victims were much slower
to adapt to their fate, Brickman et al. (1978)
found. They rated themselves as significantly
less happy than participants in the control con-
dition. The victims continued to compare their
current situation with how their lives had been
before the accident (unlike lottery winners, who
did not seem to spend much time thinking how
their lives had improved from the bygone days
of relative poverty). Brickman et al. called this
phenomenon the "nostalgia effect" (p. 921).
The seeming implication of these findings is
that adaptation-level effects are asymmetrical,
consistent with the view that bad is stronger
than good. Adaptation-level effects tend to pre-
vent any lasting changes in overall happiness
and instead return people to their baseline. After
a short peak in happiness, people become ac-
customed to the new situation and are no more
happy than they were before the improvement.
After a serious misfortune, however, people ad-
just less quickly, even though many victims
ultimately do recover (Taylor, 1983).
Comparison of unanticipated financial out-
comes can equate the objective magnitude of
events. Kahneman and Tversky (1984) had par-
ticipants perform thought experiments in which
they either gained or lost the same amount of
money. The distress participants reported over
losing some money was greater than the joy or
happiness that accompanied gaining the same
amount of money. Put another way, you are
more upset about losing $50 than you are happy
about gaining $50.
In a prospective study of stress in pregnant
women, Wells, Hobfoll, and Lavin (1999) ex-
amined gains and losses of resources early in
pregnancy and measured postpartum outcomes
including depression and anger. Gains in re-
sources had no significant effects, but losses
produced significant effects on postpartum an-
ger (even after controlling for anger at the time
of initial measurement, which included anger at
the loss of resources). Wells et al. also found
that effects of subsequent losses of resources
were significantly higher among women who
had experienced the previous losses; whereas if
they had not had the initial loss, the effect of the
later loss was muted. These findings point to a
snowballing effect of consecutive bad out-
comes. Good outcomes did not produce any
such effects.
Developmental and clinical observations
likewise suggest that single bad events are far
stronger than even the strongest good ones. Var-
ious studies reveal long-term harmful conse-
quences of child abuse or sexual abuse, includ-
ing depression, relationship problems, revictim-
ization, and sexual dysfunction, even if the
abuse occurred only once or twice (Cahill,
Llewelyn, & Pearson, 1991; Fleming, Mullen,
Sibthorpe, & Bammer, 1999; Silver, Boon, &
Stones, 1983; Styron & Janoff-Bulman, 1997;
Weiss, Longhurst, & Mazure, 1999). These ef-
fects seem more durable than any comparable
positive aspect of childhood, and it also seems
doubtful (although difficult to prove) that a sin-
gle positive event could offset the harm caused
by a single episode of violent or sexual abuse;
BAD IS STRONGER THAN GOOD327
whereas the single negative event can probably
undo the benefits of many positive interactions.
Sexuality offers a sphere in which relevant
comparisons can perhaps be made, insofar as
good sexual experiences are often regarded as
among the best and most intense positive expe-
riences people have. Ample evidence suggests
that a single bad experience in the sexual do-
main can impair sexual functioning and enjoy-
ment and even have deleterious effects on
health and well-being for years afterward (see
Laumann, Gagon, Michael, & Michaels, 1994;
Laumann, Paik, & Rosen, 1999; Rynd, 1988;
note,
however, that these are correlational find-
ings and some interpretive questions remain).
There is no indication that any good sexual
experience, no matter how good, can produce
benefits in which magnitude is comparable to
the harm caused by such victimization.
Turning from major experiences to everyday
actions, we find the same pattern of greater
power for the unpleasant than the pleasant
events. A diary study by David, Green, Martin,
and Suls (1997) examined the effects of every-
day good and bad events, as well as personality
traits.
Undesirable (bad) events had more per-
vasive effects on subsequent mood than desir-
able (good) ones. Although each type of event
influenced the relevant mood (i.e., bad events
influenced bad mood, and good events predicted
good mood) to similar degrees, bad events had
an additional effect on the opposite-valence
mood that was lacking for good events. In other
words, bad events influenced both good and bad
moods, whereas good events influenced only
good moods. Similar findings emerged when
David et al. compared neuroticism (associated
with distress and negativity) and extraversion
(associated with positivity). Neuroticism influ-
enced both good and bad moods, whereas ex-
traversion affected only good moods.
Further evidence of the greater power of bad
events emerged from a 3-week longitudinal
study by Nezlek and Gable (1999). Their par-
ticipants furnished multiple measures of adjust-
ment each day, as well as recording daily
events. Bad events had stronger effects on ad-
justment than good events on an everyday basis.
The superior strength of bad events was consis-
tent across their full range of measures of ad-
justment, including self-esteem, anxiety, causal
uncertainty, perceived control over the environ-
ment, and depressogenic cognitions about the
future, the
self,
and life in general.
How long the impact of everyday events lasts
was studied by Sheldon, Ryan, and Reis (1996).
Bad events had longer lasting effects. In their
data, having a good day did not have any no-
ticeable effect on a person's well-being the fol-
lowing day, whereas having a bad day did carry
over and influence the next day. Specifically,
after a bad day, participants were likely to have
lower well-being on the next day. Although the
results are technically correlational, something
must cause them, whether it is the bad day itself
causing the subsequent bad day or some other
cause producing the consecutive pair of bad
days.
Either way, the bad has stronger power
than good because only the bad reliably pro-
duced consecutive bad days.
Even at the sensory level, bad events seem to
produce stronger reactions than good ones. Ex-
pressive reactions to unpleasant, pleasant, and
neutral odors were examined by Gilbert, Frid-
lund, and Sabini (1987). Participants smelled
various odors while alone, and their facial ex-
pressions were videotaped. Raters then watched
the tapes and tried to infer the odor from the
facial reaction. Unpleasant odors were most ac-
curately classified, partly because more facial
movement was perceived in the unpleasant odor
trials.
Pleasant odors elicited more facial move-
ment than neutral odors, but the neutral ones
were still rated more accurately than the posi-
tive ones. Thus, responses to unpleasant odors
were apparently stronger, at least to the extent
that they could be accurately recognized by
raters.
Perhaps the broadest manifestation of the
greater power of bad events than good to elicit
lasting reactions is contained in the psychology
of trauma. The very concept of trauma has
proven broadly useful, and psychologists have
found it helpful in many different domains.
Many kinds of traumas produce severe and last-
ing effects on behavior, but there is no corre-
sponding concept of a positive event that can
have similarly strong and lasting effects. In a
sense, trauma has no true opposite concept. A
single traumatic experience can have long-term
effects on the person's health, well-being, atti-
tudes,
self-esteem, anxiety, and behavior; many
such effects have been documented. In contrast,
there is little evidence that single positive expe-
riences can have equally influential conse-
328BAUMEISTER, BRATSLAVSKY, FINKENAUER, AND VOHS
quences. It is possible that such events have
simply eluded psychological study, but it seems
more likely that the lack of an opposite concept
for trauma indicates the greater power of bad
than good single events to affect people.
Although the findings from Brickman et al.
(1978) and others reviewed in this section pro-
vide vivid and well-known indications that bad
events produce stronger, more lasting reactions
than good ones, some of the studies suffer from
a possible asymmetry in the objective magni-
tude of the event. There is no way to ascertain
objectively that winning a lottery is comparable
in magnitude to becoming paralyzed by an ac-
cident. The diary studies have the advantage of
having used all the events of the day, so these
are methodologically more useful. Most con-
vincing are the studies that attempted to ensure
equal objective magnitudes (such as when peo-
ple gain vs. lose the same amount of money)
because these permit the firmest conclusions
that bad events produce stronger reactions.
Therefore, throughout the rest of
this
review, we
emphasize studies that either did manage to
equate the good and the bad events in terms of
their objective magnitude or that took some
broad, representative or exhaustive sample of
events.
In summary, most findings indicate that peo-
ple react more strongly to bad than good events.
The evidence covers everything from minor ev-
eryday events and brief experimental exposure
to aversive odors to major life events and trau-
mas.
Bad events produce more emotion, have
bigger effects on adjustment measures, and
have longer lasting effects.
Close Relationships
One of the central tasks and goals of human
life is to sustain a network of close relationships
characterized by mutual caring and pleasant,
supportive interactions (e.g., Baumeister &
Leary, 1995). Unfortunately, many relation-
ships fail to last, and others are sometimes less
than satisfactory. In this section, we review
evidence about good versus bad patterns that
contribute to the long-term relationship out-
comes. Obviously, one would expect that bad,
destructive characteristics of the relationship
will hasten its demise; whereas good, construc-
tive ones will preserve it. The relevant predic-
tion goes beyond that, however: The harmful
effects of the bad characteristics will exert more
influence over the relationship outcome than the
beneficial effects of the good characteristics.
People commonly believe that positivity of
communication (as opposed to negativity) is
associated with high relational satisfaction (e.g.,
friendships, marriages, partnerships, and fami-
lies).
In general, research findings are consistent
with this assumption. People satisfied with their
relationships communicate with more positive
verbal behaviors (e.g., agreement, confirmation,
constructive problem solving, politeness, ex-
pressing forgiveness) and nonverbal behaviors
(e.g., smiling, head nodding, caring, or con-
cerned voice; for more detailed descriptions of
these behaviors, see Gottman, 1979; Riskin &
Faunce, 1970; Rusbult, Johnson, & Morrow,
1986;
Stafford & Canary, 1991; Ting-Toomey,
1983;
Wills, Weiss, & Patterson, 1974). On the
contrary, people dissatisfied with their relation-
ships communicate with more negative verbal
behaviors (e.g., insults, threats, or criticisms)
and nonverbal behaviors (e.g., frowning or
speaking in a cold hard voice).
More important, however, positive and neg-
ative communication have different impacts on
relational satisfaction, and the negative are
more decisive. To show this, John Gottman and
his colleagues (Gottman, 1979, 1994) video-
taped married couples in the laboratory and at
home as they talked about a wide variety of
topics such as how their day went, the nutri-
tional value of certain foods, marital problems
in general, and specific conflicts in their rela-
tionship. They then coded the couple's behav-
iors in categories (e.g., verbal, nonverbal, pos-
itive,
and negative). The findings indicated that
the presence or absence of negative behaviors
was more strongly related to the quality of cou-
ples'
relationships than the presence or absence
of positive behaviors. Positivity and negativity
were independent, in the sense that increasing
one did not necessarily decrease the other. The
important implication is that increasing positive
behaviors in a relationship will not affect the
relationship as much as decreasing negative be-
haviors. In another study in which videotaped
marital interactions were used, Gottman and
Krokoff (1989) found that negative interactions
predicted marital satisfaction more strongly
than positive interactions.
The effects of emotional interactions on
changes in relationship satisfaction were exam-
BAD IS STRONGER THAN GOOD329
ined by Gottman and Levenson (1986; Leven-
son & Gottman, 1983, 1985). They made vid-
eotapes of couples interacting, then showed the
interaction tapes to the individuals and obtained
ongoing ratings of affect through the interac-
tion. Of particular interest were data on reci-
procity, defined in terms of one person express-
ing a similar emotion or change in emotion right
after the partner had indicated similar feelings.
Reciprocity of negative affect was especially
potent and in particular was more influential
than reciprocity of positive affect. The greater
influence of negative affect reciprocity was
found with regard to differentiating happy ver-
sus distressed marriages (Levenson & Gottman,
1983).
In a longitudinal follow-up 2 years later,
the couples who had initially shown higher rates
of negative affect reciprocity reported greater
declines in relationship satisfaction, whereas
reciprocity of positive affect had no significant
effect (Levenson & Gottman, 1985). In sum-
mary, relationships are most affected by pat-
terns in which one person responds negatively
to the other's negative act or feeling.
On the basis of these results, Gottman (1994)
has proposed a revealing diagnostic index for
evaluating relationships: He proposed that in
order for a relationship to succeed, positive and
good interactions must outnumber the negative
and bad ones by at least five to one. If the ratio
falls below that, the relationship is likely to fail
and breakup. This index converges well with
the thrust of our argument: Bad events are so
much stronger than good ones that the good
must outnumber the bad in order to prevail.
Gottman's index suggests that bad events are on
average five times as powerful as good ones, at
least with regard to close relationships.
Constructive and destructive problem-solv-
ing behavior patterns for relationships were
studied by Rusbult et al. (1986; see also Rusbult
& Zembrodt, 1983). They were able to classify
couples as to the degree to which they used
constructive and destructive approaches to
problems, and these were assessed indepen-
dently so that a given couple might use both,
either, or none. In a longitudinal design, Rusbult
et al. showed that the destructive patterns were
more predictive of relationship outcomes than
constructive ones were. In particular, destruc-
tive responses to the partner's destructive re-
sponses showed a greatly increased predictive
power. This finding confirms Levenson and
Gottman's (1985) conclusion that reciprocity of
bad responses is an especially potent predictor
of relationship outcomes (and is stronger than
reciprocity of good responses). The implication
is that the long-term success of a relationship
depends more on not doing bad things than on
doing good things.
A similar conclusion emerged from a recent
longitudinal study by Huston, Caughlin, Houts,
Smith, and George (2001). By following cou-
ples for more than a decade, they were able to
ascertain what features of early marital relation-
ships predicted divorce (and other unhappi-
ness) 10 to 12 years later. Huston et al. found
that levels of negativity and distress early in the
marriage were higher among the later divorcing
couples than among the happily married ones.
Positive relations during the early years of mar-
riage, including love and affectional communi-
cation, did not differ significantly between the
ones who ended up divorcing and those who
ended up happily married. Without random as-
signment to early marital conflict (a technical
and ethical impossibility), it is difficult to draw
a firm causal inference from these data. It also
remains possible that the early distress reflected
some underlying conflict or even personality
problem among the later divorcing spouses, but
even that conclusion would fit the view that bad
is stronger than good.
Even stronger results emerged from a 2-year
longitudinal study by Huston and Vangelisti
(1991).
They measured three types of socio-
emotionally expressive behavior among newly-
wed couples: affectionate communication, sex-
ual interest, and negativity. Sexual affection had
no relation to marital satisfaction, and giving or
receiving affection had only weak and inconsis-
tent relationships to satisfaction. In contrast,
negativity had strong and consistent links to
global marital satisfaction. Thus, people's sat-
isfaction with their marriage depended much
more heavily on the bad parts (negativity) than
on the good parts (affection and sex).
In support of this idea, sexual dysfunction
was found to have a greater effect on the marital
bond than good sexual functioning. McCarthy
(1999) reported that when sexuality functions
well within a marriage, it accounts for 15-20%
of the variance in the marital bond, but when
sex functioning is bad or nonexistent (which
most married couples would consider a bad
state),
it accounts for 50-75% of the variance.
330BAUMEISTER, BRATSLAVSKY, FINKENAUER, AND VOHS
The power of bad sexual experiences, then, far
outweighs the benefits of good sexual experi-
ences within a marriage.
One factor that may contribute to some of
these effects is that destructive behaviors are
understood better than constructive ones.
Acitelli, Douvan, and Veroff (1993) found that
during the early years of marriage, couples per-
ceived and understood each other's destructive
behaviors better than the constructive ones.
Acitelli et al. interpreted this as based on the
greater visibility and recall of bad, destructive
behaviors. As Kellermann (1984) has noted,
however, such explanations are not theoretically
complete because they fail to say why bad
events are more readily noticed and recalled.
Daily reports of spousal behaviors and mari-
tal satisfaction were made for 2 weeks by par-
ticipants in a study by Wills et al. (1974). Of the
amount of variance (in marital satisfaction) that
the predictor variables were able to explain in
regression analyses, the majority (65%) was
captured by the aversive, displeasurable behav-
iors.
This was significantly greater than the
amount explained by supportive, pleasurable
behaviors (25%). This was true despite the fact
that there were about three times as many pos-
itive behaviors as aversive ones. The greater
power of the bad behaviors had to overcome
their lesser number in order to produce a stron-
ger effect.
Reciprocation patterns were also examined
by Wills et al. (1974). Interspouse correlations
indicated that negative, displeasurable behav-
iors were reciprocated to a significant degree,
whereas the reciprocation of positive, pleasur-
able behaviors was weaker and not significant.
This is an important step toward explaining the
greater power of bad events to affect relation-
ship outcomes: The couples' subsequent inter-
actions are apparently more directly and consis-
tently affected by bad than good behaviors. As
with the daily events reviewed in the preceding
section, couple interactions continue to be af-
fected by bad more than good.
The relative contributions of stress (negative
factors), social support (positive), and resources
(positive) to the quality of family life were
assessed in an extensive telephone survey by
Pittman and Lloyd (1988). Both marital satis-
faction and parental satisfaction were more
strongly affected by the bad events (i.e., the
stresses) than by the positive (i.e., support and
resources). Thus, negativity and stress added
more than 20% to the amount of variance in
marital satisfaction that was explained, whereas
positive support and resources added only 5%.
All in all, the evidence is fairly clear and
unanimous in indicating that relationships are
more affected by bad events than good ones. As
seen in daily interactions, broad patterns, affect
of problem solving, and marital communica-
tion, bad events have stronger effects than good
events. Reciprocation of bad responses appears
to be especially powerful for leading to deteri-
oration and breakup of close relationships.
Other Relationships and Interactions
Although close relationships have received
the greatest amount of study, there is also some
relevant information regarding not-so-close re-
lationships and other forms of interpersonal in-
teraction. We report several such studies here,
although the research on formation of initial
impressions is covered in a separate section
later in the article.
Sociometric studies have examined how in-
dividuals perceive each other within established
groups or social networks. If bad is stronger
than good, then dislike and social rejection
should be more pronounced, which would be
reflected in higher agreement throughout the
social network. A meta-analysis of sociometric
studies of children recently confirmed this con-
clusion (Newcomb, Bukowski, & Pattee, 1993).
In particular, the two social extremes were rep-
resented by the highly popular children and the
rejected children, and these were about the same
proportions (9% and 12% of the groups, respec-
tively, on average). Consistency of reports
across the children, as well as for self-reports
and for ratings by teachers and parents, was
higher for the rejected than for the popular
children. In other words, all perspectives agree
more about who is rejected than about who is
popular.
Another approach to the same problem is to
examine the link between naming someone as
"best friend" or "worst enemy" and overall rat-
ings of ability. An ambitious study by French,
Waas,
and Tarver-Behring (1986) obtained ex-
tensive sociometric data from 250 third- and
fourth-grade children. All children listed their
three most and least desired friends, as well as
listing the three best and worst peers at sports
BAD IS STRONGER THAN GOOD331
and at schoolwork. All children then rated ev-
eryone in their class in terms of friendship,
sports, and schoolwork. The proportion of vari-
ance shared by the two methods was consis-
tently higher for the bad than for the good; that
is,
low ratings led to more frequent nominations
as "undesirable friend," "bad at sports," and
"bad at schoolwork" more reliably than high
ratings led to more frequent nominations as
"desirable friend," "good at sports," and "good
at school."
Undoubtedly the initial acts in an interaction
create expectancies and set the tone for further
ones,
and if subsequent acts differ, the expect-
ancies are violated. The impact of these was
assessed by Afifi and Burgoon (2000), who had
participants observe a videotaped interaction.
Their research design included changes from
initially negative to positive and from initially
positive to negative interactions. These violated
expectancies produced strong reactions, but the
violations in the negative direction had stronger
effects on attraction. Their finding that changes
produced stronger reactions than consistent in-
terpersonal behavior replicated an earlier dem-
onstration by Aronson and Linder (1965), but
Afifi and Burgoon's work clearly showed that
the shift toward the negative has a more pow-
erful effect than the comparable shift toward the
positive. That is, the largest change in attraction
occurred when the stimulus person on the vid-
eotape started off by behaving in a friendly,
interested manner and then turned aloof and
unfriendly.
A further investigation by Reyes et al. (1999)
showed that both social attraction and sexual
desire were more strongly influenced by nega-
tive,
unpleasant social interactions than by pos-
itive,
pleasant ones. Participants who behaved
in the unpleasant style produced clear (negative)
reactions that were stronger than the reactions to
people who showed a positive, friendly interac-
tion style.
The notion that bad is stronger than good in
social interactions received a fairly explicit test
in a recent study. Exline and Baumeister (1999)
had people play prisoner's dilemma against a
simulated opponent who was randomly pro-
grammed to start off with either a cooperative
(good) or a defensive/exploitative (bad) move.
When participants were asked to rate their op-
ponent, the one who started off with the bad
move was rated as stronger than the one who
started off with the cooperative move. Partici-
pants also rated the bad opponent as stronger
than they themselves were, unlike the coopera-
tive opponent.
In fact, when Baumeister and Leary (1995)
reviewed the evidence in support of a need to
belong, they concluded that that need was for
nonnegative interactions, rather than positive
ones as they had originally theorized. The rea-
son was that neutral interactions seemed ade-
quate to satisfy the need to belong in many
cases.
This too confirms the greater power of
bad: The effects of positive, good interactions
were not consistently different from the effects
of neutral interactions, whereas