A meta-analytic review of help giving and aggression

COGNITION AND EMOTION, 2004, 18 (6), 815±848
A meta-analytic review of help giving and aggression
from an attributional perspective: Contributions to a
general theory of motivation
Udo Rudolph, Scott C. Roesch, Tobias Greitemeyer,
and Bernard Weiner
Technical University Chemnitz, Germany
The present review syntheses 64 investigations on the determinants of helping and
aggression involving more than 12,000 subjects, providing empirical tests of
Weiner's (1986, 1995) theory of social conduct. A meta-analytic test of the proposed causal cognition-emotion-behaviour sequence reveals that judgements of
responsibility determine the emotional reactions of anger and sympathy, and that
these emotional reactions, in turn, directly influence help giving and aggression.
Results are highly consistent across several potential moderator variables including
type of culture, sample characteristics, publication year, and publication status.
Moreover, the present analyses suggest that the hypothesised model holds true for
real events as well as for simulated data. Exploratory comparisons between the
helping versus the aggression domain suggest that comparable results are obtained
for these two domains, except that perceptions of responsibility are more likely to
exert an additional proximal role in aggressive retaliation as compared to help
giving. The implications of these findings for a general theory of motivation in the
interpersonal and the intrapersonal domains are discussed.
This paper pursues what might be regarded as an anachronistic issue: The
creation of a general theory of motivation. For over four decades, the ``grand''
theories of motivation, as represented in the writings of Freud and Lewin and in
the controversies between Hull and Tolman, have been regarded as outmoded. It
is sometimes thought that viable mini-theories that pertain to a particular
motivational area have arisen in their place, such as theories of achievement
striving, or helping, or aggression. Unfortunately, the models that have been
Correspondence should be addressed to Dr Udo Rudolph, General and Biological Psychology,
Department of Psychology, Technical University Chemnitz, 09107 Chemnitz, Germany.
Work on this paper was supported by a grant from the Deutsche Forschungsgemeinschaft (AZ Ru
599/3-1). We would like to thank Wulf-Uwe Meyer and Rainer Reisenzein for their helpful comments on an earlier version.
# 2004 Psychology Press Ltd
http://www.tandf.co.uk/journals/pp/02699931.html
DOI:10.1080/02699930341000248
816
RUDOLPH ET AL.
proposed have not comprehensively captured the motivational domain that they
address. One reason for this lack of inclusiveness in, for example, the helping
domain, is that there are numerous sufficient but not necessary causes of help
giving, such as genetic relatedness, moral beliefs, role demands, and so on.
Conversely, help may be withheld because others are present to diffuse
responsibility, personal costs, inferences about the lack of deservingness of aid,
and so on. In a similar manner, lists of factors that instigate aggression (e.g.,
high testosterone level, aggressive cues, outgroup prejudice), or that prevent its
expression (fear of retaliation, moral inhibition, ingroup norms) can also be
readily generated. It is manifestly difficult to develop a theory of helping or
aggression that could incorporate all these divergent antecedents of behaviour.
Hence, researchers understandably focus (often quite successfully) on documenting the importance of one of the many determinants of these behaviours.
A more general theory across domains then seems even less possible inasmuch
as the sources of action or inaction between these fields of motivation are not
entirely comparable.
This paper adopts a different perspective to motivational theoryÐone that
neither proposes a ``complete'' theory of behaviour across areas of study nor
attempts to explain all actions within a content domain. Rather, it is suggested
that an indeterminant number of motivational domains may be examined with
the same theoretical lens, but accounting for only a restricted set of phenomena
within each of these domains. Nonetheless, by exhibiting applicability within
more than one field of study, the seeds for a more general theory may be sown.
At the least, a distinctive, alternative approach to theory building is proposed,
one that advocates theories at an ``intermediate'' level of generality (e.g., see
Kelley, 1992; Vollmer, 1984).
The conceptual framework adopted here is attribution theory, which examines phenomenal causality, or beliefs about why a particular event or outcome
has occurred (Heider, 1958). In this paper, we first introduce some of the tenets
of an attributional approach, drawing a distinction between intrapersonal and
interpersonal theories of behaviour. Then the implications of this framework for
an understanding of help giving and aggression are developed, with the same
concepts used in both domains. This is followed by two meta-analyses that test
the conceptual framework in each domain. A meta-analytic approach seems
justified for two reasons. First, prior to this writing, meta-analytic overviews of
data pertinent to an attributional approach to motivation have not been reported.
Second, instead of merely being a statistical method for detecting effect sizes,
meta-analytic techniques have become a promising tool for the exploration of
causal mediating processes (for an overview, see Shadish, 1996). As the mediating role of emotionsÐmediating between thoughts and actionsÐis one of the
central tenets of attribution theory as interpreted here, a meta-analytic approach
enables us (1): to comprehensively synthesise the existing data in this field, and
(2) to employ specific techniques to test for causal characteristics of the proposed cognition-emotion-behaviour sequence. Before presenting this specific
HELP GIVING AND AGGRESSION
817
meta-analytic approach in more detail, let us first summarise the theoretical
background of these analyses.
ATTRIBUTION THEORY
To define the empirical focus of this paper, a distinction between intrapersonal
versus interpersonal attribution-based theories of behaviour is useful (see
Weiner, 2000). Imagine, for example, a person is in financial need because of
the loss of a job. Assume that the laid-off employee makes an ascription of the
job loss to a lack of aptitude. It has been documented that this attribution results
in lowered self-esteem, low expectancy of future success, affective experiences
of shame and humiliation and, in turn, behavioural withdrawal (e.g., see Weiner
1985b). These self-directed and personally regulating thoughts, feelings, and
behaviours provide the components for an intrapersonal theory of motivation.
Now consider that, following the lay-off, this person asks someone for
financial help. The potential help giver might view the needy other as able or
unable, good or bad, likely or unlikely to get another job, and responsible or not
responsible for the present plight. The potential helper also could experience
emotions, such as irritation and anger, and/or pity and sympathy, which are
directed toward the person in need. And they may praise and help, or reprimand
and turn away. These other-directed cognitions, affects, and behaviours are the
components of an interpersonal (social) theory of motivation.
Of course, the boundaries between the intrapersonal and the interpersonal
systems are fuzzy, and both are evident and interact in many behavioural
episodes. Nevertheless, this distinction bears upon this paper, where we
empirically examine only an interpersonal theory related to help giving and
aggression. It will be suggested, however, that intrapersonal and interpersonal
approaches guided by attribution theory share similar conceptual properties (see
Weiner, 2000).
Now we turn to help giving and aggression from an attributional perspective.
Going toward (altruism) and going against (aggression) form the very heart of
social psychology. Hence, we believe that bringing these two areas of study
together within a common theoretical framework is a valuable endeavour. To the
best of our knowledge, with the exception of evolutionary psychology, these two
social domains have been addressed by different theorists, proposing incommensurate theories that are supported (or not) by noncomparable data. As
already intimated, it will be argued here that these fields of study can be
incorporated within the same theoretical system, and this contention will be
tested with meta-analytic techniques.
Help giving
First we turn to an attributional explanation of helping behaviour. Inasmuch as
this theoretical framework has been described in detail elsewhere (e.g., see
Weiner, 1995), only the key conceptual assumptions are reviewed briefly here.
818
RUDOLPH ET AL.
Imagine that one is approached with a request for aid. Our attributional
perspective proposes that this act tends to give rise to a causal search to discover
why help is needed, and this causal search eventuates in a causal inference. This
aspect of the attribution process, which concerns the antecedents of causal
understanding (e.g., Kelley & Michela, 1980), is neglected here. That is, how
one comes to know is not addressed, whereas the social consequences once
causal knowledge has been attained are focal.
It is next contended that the selected cause is given meaning by its placement
within a dimensional space, a space that represents the underlying characteristics
or properties of causes. To understand the motivational effects of causal beliefs,
it is necessary that qualitative differences between causes are transformed into
quantitative differences, and for this to occur the causes must be rendered
comparable on some psychological dimension(s).
A great deal of research has been directed to ascertaining what these causal
properties might be (e.g., Weiner, 1986). One characteristic that has repeatedly
been identified is that of controllability. Some causes, such as lack of effort as
the cause of a job lay-off, are subject to volitional alteration, whereas other
causes of the loss, including low aptitude or poor economic conditions, cannot
willfully be changed by the person in need. Again restricting the scope of the
discussion in this article, personal controllability is the only property of causes
that is distinguished here.
If the cause of another's need is controllable by that person, that is, if /she
``could have done otherwise'', then the person may be perceived as responsible
for the plight (for greater detail regarding the link between causal controllability
and personal responsibility, and for a discussion of mitigators that negate this
association, see Weiner, 1995). According to appraisal theories of emotion (e.g.,
Lazarus, 1991; Reisenzein, 1986, Smith & Ellsworth, 1985), responsibility for a
negative state of another gives rise to anger. Thus, one is angry when a
dependent other has been dismissed because of laziness, intentional absence
from work, and the like. Anger, in turn, is believed to be a goad to action that
results in antisocial responses, including the decision not to help the needy other
(for a review, see Weiner, 1995).
Now consider this motivational sequence if the job had been lost because of
lack of aptitude. Aptitude is construed an uncontrollable cause in that it cannot
be volitionally changed. The laid-off worker therefore is not held accountable or
personally responsible for the financial need. Again guided by appraisal theories
of emotion, lack of controllability and nonresponsibility for a negative outcome
often elicit sympathyÐwe feel sorry for the mentally handicapped who cannot
perform cognitive tasks and for the physically handicapped who cannot perform
motor tasks. Sympathy, in turn, is thought to give rise to prosocial reactions,
including helping. The motivational processes from an attributional perspective
that bear upon help giving are shown in Figure 1, initiating with ``another in
need of help''.
HELP GIVING AND AGGRESSION
819
Figure 1. A combined model of a cognition-emotion-action sequence for pro- and antisocial
behaviour.
It is evident from this model that the vast array of variables that influence
helping behaviour, including costs and benefits, genetic relatedness, and the like,
are not embraced. Rather, attributions and their influence on judgements of
responsibility and emotions are proposed as mediators between the perception of
the need of another and action. This restrictive approach is quite similar to what
researchers in the field of helping do, employing other concepts and mediators.
What makes this analysis distinct is that we now apply the same conceptual
framework to aggressive behaviour.
Aggression
The basic argument to be put forth regarding aggression follows closely that
concerning help giving, and thus is presented in even more abbreviated form (for
a more complete statement, see Weiner, 1995). It is contended that, if a person is
a victim of a harmful act, then that individual seeks to determine the cause of the
infraction (see also Averill, 1982). Hence, the type of aggression incorporated
within this framework is reactive, following the harm that was done by another.
If the harmful behaviour is perceived as subject to the volitional control of
another, and particularly if committed with the intent to do harm, then the
perpetrator of this initial misdeed is inferred to be responsible for the act and the
damage done. That is, as in the proposed explanation of help giving, the
motivational process proceeds from an event (request for aid, a harmful deed) to
a causal inference and then to a judgement regarding personal responsibility. If
820
RUDOLPH ET AL.
the individual is judged to be responsible, then anger is elicited, followed by the
tendency to engage in hostile retaliation. On the other hand, if the offender is
perceived as not responsible for the prior harm (e.g., it was an accident, or there
were mitigators, such as the inability to distinguish right from wrong), then
anger is lessened (or reduced to zero). There might even be a sympathetic
reaction, although this is equivocal depending on whether the harm doer is also
harmed by the action. In turn, the tendency to respond aggressively will be
weak, or perhaps not at all elicited. This sequence is also depicted in Figure 1,
initiating with ``personal harm''.
GENERAL THEORY
It is now straightforward to combine the social domains of helping and
aggression within the same theoretical framework. The more general conceptual
scaffold is depicted in Figure 1. As previously indicated, neither a theory of
helping nor a theory of aggression in an inclusive sense is being proposed.
Rather, within each field of study there is a restricted set of behaviour that falls
within the range of convenience of this conception. On the other hand, it also is
the case that the phenomena that can be incorporated within the theory are not
trivial and cannot be dismissed as rare or atypical instances. It will also become
evident that the two areas of study, helping and aggression, are considered
because only these two areas lend themselves to a meta-analytical approach. In
the discussion, additional motivational areas (including what we have labeled as
intrapersonal domains) are considered, but without accompanying empirical
reviews.
At a more macro level, the theory just outlined suggests that a motivational
sequence progresses from thinking to feeling to acting. That is, emotions provide
the bridge between cognition and behaviour, being determined by thoughts and
giving rise to action. Obviously, many other motivated sequences can be postulated. For example, perhaps emotions give rise to thoughts (``I am angry; that
person must be responsible for the need'') and thinking produces action (``I
think it is his fault, I will not help''). Or perhaps thinking gives rise to both
emotions and behaviour, so that affects are epiphenomena in terms of their
relation to behaviour. And these are just a few of the permutations that are
possible, particularly if bidirectionality is admitted. Examining the specific
content areas that have been discussed will provide evidence regarding the
temporal sequence for a more general motivational formulation.
OVERVIEW OF METHODOLOGY
Two meta-analyses (one for helping and one for aggression) and an exploratory
comparison of the results of these separate analyses follow. As outlined above,
an attributional explanation of helping and aggression proposes a causal
sequence from thoughts to emotions to behaviour. In contrast to the most
common use of meta-analysis, which has focused on summarising the magnitude
HELP GIVING AND AGGRESSION
821
of a relationship between two variables, our main interest here is in detecting the
causal relationship between the motivational variables proposed by attribution
theory. A comprehensive account of methods for the exploration of casual
mediating processes is given in Cook et al. (1992) and Shadish (1996). The
present meta-analyses combine several characteristics of the possible metaanalytic research methods described thus far; similar techniques have been
employed by Shadish and Sweeney (1991) and Becker and Schram (1994).
When looking at the present theoretical account from a methodological point
of view, emotions have the status of mediators, that is, they are ``the generative
mechanisms through which the focal independent variable is able to influence
the dependent variable of interest'' (Baron & Kenny, 1986, p. 1173). Thus,
perceptions of controllability have the status of an independent variable that
gives rise to two potential emotional mediators (anger and sympathy), which
then cause the behavioural outcome (helping or aggression).
As Shadish (1996) has pointed out, mediators are often confused with
moderators, which represent ``a qualitative (e.g., sex, race, class) or quantitative
variable (e.g., level of reward) that affects the direction and/or strength of the
relations between an independent or predictor variable and a dependent or
criterion variable'' (Baron & Kenny, 1986, p. 1174). In addition to analysing
the proposed causal thought-emotion-behaviour sequence by means of metaanalyses testing of the proposed causal mediation models, we also include a
number of potential moderators in the following analyses. Specifically, we test
whether the proposed causal sequence is influenced by culture, sample characteristics, publication year, and publication status (for details, see below).
Moreover, as tests of attributional theories of motivation have occasionally
been criticised for restricting themselves to data from thought experiments
instead of ``real'' data (e.g. see Enzle & Shopflocher, 1978), we will also
include a test of this potential moderator in the following analyses. For
generalisability, it is crucial that the pattern of results is not limited to a special
paradigm (e.g., to simulational data obtained by thought experiments). As
several studies in our meta-analysis provided data about behaviour in a real
event, it proved possible to test whether there are differences between simulational data on the one hand, and recounting or real event data on the other.
For each of the following meta-analyses, the same process and type of
analysis was conducted. For space considerations, we present the methodology
only as applied to helping, except when specific details regarding the aggression
analysis also are needed.
Criteria for inclusion
There were three criteria for including a study in the present meta-analysis. (1)
The study had to have at least one attribution variable (i.e., controllability,
responsibility). (2) The study had to have at least one emotion and/or
behavioural variable, with anger and sympathy as possible emotions and some
822
RUDOLPH ET AL.
help- or aggression-related variable as a possible behaviour. (3) First-order
correlations between at least two of these variables were available. Therefore,
when other kinds of statistics were reported (e.g., path-coefficients, partial
correlations, means, or F-values), we contacted the principal author of the study
to obtain first order correlations.
Literature search
Our search for relevant studies included three steps. First, we developed a
comprehensive set of keywords and combinations of keywords to conduct a
computer scan of the PsycLit database. For both help giving and aggression,
three groups of variables exist, namely, cognitive, emotional, and behavioural
measures. For help giving, these keywords included responsibility and controllability for the cognitive variables, anger and sympathy for the emotional
measures, and help, help giving, as well as social support for the behavioural
variables. For the aggression domain, we used the key words aggression,
retaliation, and violence for the behavioural measures; for the cognitive and
emotional variables, the same key words were used as for help giving. Because
studies were needed that included at least two stages of the hypothesised model,
it was required that the database entry contained at least two keywords that stem
from at least two different groups of variables (i.e., at least one cognitive and
one emotional variable, or one cognitive and one behavioral variable, or one
emotional and one behavioural variable). All possible combinations between at
least two groups of variables were searched (e.g., by combining ``controllability'' and ``help'', or ``anger'', and ``retaliation'').
It is evident that in this initial selection stage many investigations regarding
helping and regarding aggression were eliminated from consideration. For
example, in the helping area, research from a sociobiological perspective documenting that helping is related to genetic similarity (e.g., Burnstein, Crandall,
& Kitaymana, 1994), as well as investigations linking help giving to personal
costs and benefits (e.g., Dovidio, Piliavin, Gaertner, Schroeder, & Clark, 1991),
to the number of people available to help (Latane & Darley, 1970), and to mood
(Batson, 1998) were not included in this meta-analysis. In a similar manner, and
now considering aggression, it will be seen that influential investigations documenting that, for example, overt hostilities are affected by temperature level
(Anderson & deNeve, 1992) and violence in the media (see Huesmann & Eron,
1986) also are not included in these analyses. That is, the included investigations
are a small subset of the total research conducted in these areas.
A second step toward locating appropriate studies was to examine the citations provided within the reports and reviews that were identified by the PsycLit
search and subsequently retrieved and scanned for eligibility. About 1800 papers
were identified with this method. The abstracts were manually searched to
extract the relevant studies. By this method, the number of potential studies was
HELP GIVING AND AGGRESSION
823
reduced substantially. For example, less than 20% of these articles were original
empirical contributions; the other articles were theoretical contributions that
contained no original data and hence were not included in the present analyses.
Finally, we contacted all principal authors of papers and studies identified by
these first two steps and asked them for unpublished or submitted studies pertinent to this research field.
As a result of this search process, 39 studies were obtained that met the
above-mentioned criteria for the help giving domain, and 25 studies were
obtained for the aggression domain (these 64 studies are presented with asterisks
in the References). Within the help giving domain, 27 out of 39 studies are
published, 9 are currently submitted for publication, and 3 are unpublished.
Overall, the total number of participants involved in these investigations of the
determinants of help giving was 7945, with an average of 204 participants per
study. With regard to the aggression domain, 17 of the 25 studies are published,
6 are currently submitted for publication, and 2 are unpublished. The total
number of participants involved was 4598, with an average of 184 participants
per study. An overview of these investigations and the respective first-order
correlations are reported in Table 1 (help giving) and Table 2 (aggression).
Study categorisation
For the primary studies in both the helping and aggression meta-analysis, five
descriptive variables were coded and used in a moderator analysis: (1) type of
culture (independent cultures such as the US or Germany versus interdependent
cultures such as Japan, Nigeria); (2) type of investigation (simulation,
recounting or participation in a real event); (3) type of sample (children, community, students); (4) year the study was conducted (based on a median split);
and (5) publication status (published, unpublished). The detailed classifications
for these moderator variables are presented in Tables 1 and 2 (see below). For
the helping meta-analysis only, one additional moderator was identified: group
in need of help (stigmatised, nonstigmatised).1 Finally, it is worth noting that we
did not include gender as a possible moderator. The majority of the studies
included in the analyses did not differentiate between female and male subjects
and target person, and if they did, they did not find any systematic differences.
Model testing method and procedure
To test the judgements of responsibility models, zero order correlations and
path-analytic models were evaluated. First, all relevant statistical data not
already in the form of a correlation coefficient (r) were converted to r. As noted
1
The authors of the present manuscript coded all of the above-mentioned variables from the
relevant studies. Given the objective nature of the coding, it is not surprising that reliability was very
high (r = .96). Disagreements were resolved by discussion.
824
Meyer & Mulherin (1980)
Weiner (1980a)
Weiner (1980b)
Reisenzein (1986)
Schmidt & Weiner (1988)
Weiner et al. (1988)
Weiner & Graham (1989)
Betancourt (1990)
Betancourt (1990)
Sharrock et al. (1990)
2.
3.
4.
5.
6.
7.
8.
9.
10.
Author(s)
1.
No.
1
2
1
1
1
1
1
2
3
1
1
1
1
1
1
1
1
1
1
1
Study Status
34
61
156
370
59
496
138
116
28
80
N
US medical
staff
US students
US students
US citizens
5±95 years old)
US students
US students
US students
US students
US students
Canadian
students
Subjects
Helping behaviour
toward patients
Problem in school
Problem in school
Various situations
with controllable vs.
uncontrollable reasons
for needing help
10 different stigmas
(kinds of illness)
Needing class notes
Class notes, subway
Needing class notes
Drunk vs ill person in
subway
8 causes for being in
need for money
Eliciting stimulus
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Measure of
help giving
TABLE 1
Overview of the helping studies included in the present analysis
.52
.51
.51
.36
.55
.64
7.31 7.23
7.33
7.40
7.56
7.55
7.64
7.49
7.54
7.77
7.37
.29
7.15
7.43
7.26
7.39
7.29
7.44
7.41
7.37
7.14
.39
.45
.38
.63
.47
.45
.59
.46
.37
S- -A
7.35
7.58
7.43 7.71
7.49
7.71 7.17
7.65
G- -S C- -A C- -H S- -H A- -H
825
Graham & Weiner (1991)
Karasawa (1991)
Kojima (1992)
Matsui & Matsuda (1992)
Matsui & Matsuda (1992)
Zucker & Weiner (1993)
Zucker & Weiner (1993)
Zucker & Weiner (1993)
Zucker & Weiner (1993)
Betancourt et al. (1995)
Menec & Perry (1995)
Sunmola (1994)
11.
12.
13.
14.
15.
16.
17.
18.
19.
20.
21.
22.
1
1
2
2b
2a
1b
1a
2
1
1
1
1
1
1
1
1
1
1
1
2
2
1
1
1
414
249
126
47
47
122
122
80
100
112
180
370
Nigeria comm.
sample
Canadian
students
US students
US students
US students
US students
US students
Japanese
students
Japanese
students
Japanese
students
US students
US community
sample
Offer help to
government
Various stigmas
(kinds of illness)
Needing help at
school
Welfare decision in
case of poverty
Personal financial
help in case of poverty
Welfare decision in
case of poverty
Personal financial
help in case of poverty
Lending class notes
Lending class notes
Lending class notes
Falling behind in
school
Waiting in line
situation
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
7.29
7.46
7.20
7.53
7.31
7.42
7.51
7.71
7.51
7.56
.27
.78
.06
.17
.44
.57
.61
.61
.33
.52
7.27
7.35
7.56
7.61
7.39
7.28
7.38
7.52
7.47
7.26
.31
.34
.40
.79
.77
.43
.60
.54
.53
.51
.32
.38
(Continued)
7.35 7.27
7.31 7.22
7.09
7.32
7.19 7.26
7.34
7.45 7.44
7.65
7.52 7.36
7.61 7.70
7.34
7.40
826
Menec & Perry (1998)
George (1997)
Menec & Perry (1998)
Dagnan et al. (1998)
George (1998)
Dijker & Koomen (1999)
Steins & Weiner (1999)
Watson & Higgins (1999)
Yamauchi & Lee (1999)
24.
25.
26.
27.
28.
29.
30.
31.
Author(s)
23.
No.
1
1
1
1
1
1
2
1
1
1
2
1
2
1
1
1
2
1
Study Status
171
217
281
143
537
40
137
279
133
N
Japanese
students
Canadian
students
US/German
students
Dutch students
Community
sample
US health care
staff
Canadian
students
US students
Canadian
students
Subjects
Moral dilemmas
Various stigmas
(kinds of illness)
Infection with HIV
Various stigmas
(kinds of illness)
Academic problem of
friend
Persons with learning
disabilities
Various stigmas
(kinds of illness)
Academic problem of
friend
Various stigmas
(kinds of illness)
Eliciting stimulus
TABLE 1
(Continued)
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Actual
behaviour
Self-reported
intention
Self-reported
intention
Actual
behaviour
Self-reported
intention
Measure of
help giving
7.27
7.38
7.39
7.57
7.29
7.42
7.31
7.26
7.60
.40
.59
.21
.63
.49
.52
.63
.45
.62
7.11
7.20
7.20
7.58
7.19
7.25
7.16
.27
.33
.51
.56
.32
.25
.51
.25
.46
S- -A
7.13 7.28
7.06 7.26
7.15 7.32
7.47 7.34
7.05 7.22
7.53 7.40
7.13 7.18
7.09 7.21
7.19 7.25
G- -S C- -A C- -H S- -H A- -H
827
Rudolph & Greitemeyer
(2001)
Rudolph & Greitemeyer
(2001)
Rudolph & Greitemeyer
(2001)
Greitemeyer & Rudolph
(2002)
Greitemeyer & Rudolph
(2003)
Greitemeyer & Rudolph
(2003)
Greitemeyer et al. (2003)
33.
34.
35.
36.
37.
38.
39.
1
2b
2a
1
2b
2a
1
1
2
2
2
2
2
2
2
1
649
150
150
204
210
210
766
161
German
students
German
students
German
students
German
students
German
community
German
community
German
students
US community
Personal help
following a car
accident
Family members
being in need of help
in various social vs.
achievement settings
Strangers being in
need of help in
various social vs.
achievement settings
Being in need of help
in various social vs.
achievement settings
Autobiographical
recall of helping
Autobiographical
recall of helping
Autobiographical
recall of helping
Unwanted pregnancy
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Recalled
behaviour
Recalled
behaviour
Recalled
behaviour
Self-reported
intention
7.42
7.48
7.29
7.32
7.36
7.32
7.15
7.23
.60
.71
.35
.61
.63
.58
.27
.20
7.11
7.29
7.01
7.16
7.15
7.19
7.11
7.05
.54
.47
.39
.40
.34
.35
.35
.37
7.02 7.30
7.35 7.42
7.12 7.12
7.22
.01 7.32
7.20 7.26
7.04 7.03
7.15 7.33
Note: For (publication) status; 1 = published; 2 = unpublished or submitted. C = controllability/responsibility; S = sympathy; A = anger; H = helping/prosocial
behaviour. First order correlation coefficients are reported.
Zucker (1999)
32.
828
Zumkley (1981)
Johnson & Rule (1986)
Vala et al. (1988)
Betancourt & Blair (1992)
Graham et al. (1992)
Graham & Hoehn (1995)
Higgins & Watson (1995)
Ho & Venus (1995)
2.
3.
4.
5.
6.
7.
8.
Author(s)
1.
No.
1
1
3
1
1
1
1
1
1
2
1
1
1
1
1
1
Study Status
203
56
86
88
154
258
100
75
N
US community
sample
US students
US children
US children
US students
Portuguese
students
US students,
males only
German
students
Subjects
Reactions to a
battered woman who
killed her spouse
Intention to retaliate
Perception of
ambiguous events
Perception of
ambiguous events by
aggressive/
unaggressive children
Perception of a stonethrowing competition
Perception of violence
Reactions on an insult
by a coworker
Intention to retaliate
Eliciting stimulus
Recalled
behaviour
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Self-reported
intention
Measure of
help giving
7.44
7.33
.51
.28
.51
.27
.34
S- -V
.48
.31
.27 7.51
.33
.32 7.44
.48
G- -S C- -A C- -V
TABLE 2
Overview of the aggression studies included in the present analysis
S- -A
.39
.63 7.51
.41
.47 7.35
.27
A- -V
829
Stiensmeyer-Pelster (1995)
Thompson et al. (1995)
Feather (1996)
Feather (1996)
Allred et al. (1997)
Byrne & Arias (1997)
Graham et al. (1997)
Graham et al. (1997)
Stiensmeyer-Pelster &
Gerlach (1997)
9.
10.
11.
12.
13.
14.
15.
16.
17.
1
2
1
1
1
1
1
1
1
1
1
1
1
2
1
2
1
1
219
166
177
132
132
181
220
130
465
German
children
US students
African Am. &
Whites
66 US couples
66 US same-sex
dyads
Austral. comm.
sample
Austral. comm.
sample
65 US couples
German
students
Intention to retaliate
by aggressive vs.
nonaggressive
children
Reactions to O. J.
Simpson
Reactions to O. J.
Simpson
Marital satisfaction
and marital violence
Negotiation
performance in
employer vs.
employee role
Reactions to penalties
Reactions to penalties
Cardiac patients and
their wives/husbands
Intention to retaliate
by aggressive vs.
Nonaggressive
schoolchildren
Self-reported
intention
Self-reported
intention
Self-reported
intention
Recalled
behaviour
Self-reported
intention
Self-reported
intention
Self-reported
intention
Recalled
behaviour
Self-reported
intention
7.45
7.42
7.50
7.49
7.32
7.33
.57
.57
.12
.50
.49
.52
.39
.55
.49
.45
7.03
.00
.62
7.52
7.50
7.21
7.30
.31
.46
(Continued)
.46
.47
.15
.74
.50
7.39
.59
830
Wingrove & Bond (1998)
Watson & Higgins (1999)
Rudolph & Greitemeyer
(2001)
Rudolph & Greitemeyer
(2001)
Greitemeyer & Rudolph
(2003)
Greitemeyer & Rudolph
(2003)
Greitemeyer & Rudolph
(2003)
19.
20.
21.
22.
23.
24.
25.
2b
2a
1
2
1
1
1
4
2
2
2
2
2
2
1
1
Study Status
150
150
204
766
56
217
23
190
N
German
students
German
students
German
students
German
students
Austrian
students
Canadian
students
English students
US students
Subjects
Intention to retaliate
concerning family
members
Intention to retaliate
concerning strangers
Intention to retaliate
in a social vs.
achievement situation
Autobiographical
recall of aggressive
events
Autobiographical
recall of aggressive
events
4 kinds of stigmas
Angry reactions to
failure on a
cooperative game
Compliance after a
good vs. negative
achievement outcome
Eliciting stimulus
Self-reported
intention
Self-reported
intention
Self-reported
intention
Recalled
behaviour
Recalled
behaviour
7.56
7.68
7.24
7.06
7.42
7.15
.65
.77
.72
.57
.39
.82
.50
Actual
behaviour
Self-reported
intention
.48
.57
.77
.68
.43
.34
.45
.37
G- -S C- -A C- -V
Self-reported
intention
Measure of
help giving
7.47
7.69
7.34
7.07
7.42
.01
S- -V
.77
.85
.76
.52
.58
.46
.37
A- -V
7.52
7.73
7.22
7.06
7.43
7.04
S- -A
Note: For (publication) status: 1 = published; 2 = unpublished or submitted. C = controllability; responsibility; S = sympathy; A = anger; V = violence/aggression.
First order correlation coefficients are reported.
Rodrigues & Lloyd (1988)
Author(s)
18.
No.
TABLE 2
(Continued)
HELP GIVING AND AGGRESSION
831
above, only zero order relations were used in the analyses. This approach is
consistent with Becker and Schramm (1994), who suggest that incorporating
measures of partial relations in a meta-analysis provides information about
different partial relations and thus only zero order relations should be used.
Study-level correlation coefficients were then weighted according to Shadish
and Haddock (1994; Formula 18±10).2 Because a majority of the primary studies
used single-item indices to operationalise the variables of interest or because
reliability information was not provided when scales were used, these weighted
correlations were not disattenuated for unreliability.
After the weighting procedure, correlation coefficients were aggregated
according to a fixed effects model and the homogeneity of the correlation
coefficients was assessed with the Q-statistic (Hedges & Olkin, 1985). A
fixed effects model was chosen for two reasons. First, we had no a priori
reason to believe that the effect sizes (correlation coefficients) from our primary studies would differ from the true underlying population effect size
by more than sampling error. Second, if this assumption proved to be incorrect, we proceeded to explain this heterogeneity using study-level variables
(moderators) rather than incorporating this heterogeneity using a random
effects model.3
The statistical methodology for testing mediational models with meta-analytic data is in its infancy (see Becker & Schramm, 1994; Shadish, 1996, for
reviews). To date, there is no agreed upon approach to assess models with
mediating mechanisms. Shadish (1996) has outlined the advantages and disadvantages associated with the existing methods, but concludes that these limitations should not impede researchers from employing one of the existing
methods. The methodological approach taken in the current analysis resembles
that used by Premack and Hunter (1988) and Shadish and Sweeney (1991), and
was largely a function of the data available to the authors. The method is
described below.
2
The potential problems of aggregating correlations at the study level in mediational metaanalyses (e.g., it ignores within-study dependencies) have been well chronicled (see Shadish, 1996).
However, approaches that overcome this limitation, such as Becker's Generalised Least Squares
method, have limitations of their own, such as imputation of missing data and/or data structures that
are not readily conformable to existing hierarchical linear modelling software packages (e.g., HLM).
3
As noted by Shadish and Haddock (1994), the decision to aggregate data according to either a
fixed or a random effects model is rarely indisputable. While these authors suggest that the presence
of a statistically significant variance component might indicate the presence of random effects, they
further state that this is not definitive evidence, and that fixed effects covariates (moderators) might
be sufficient to explain this variance component. We attempted to do so in this study. When the data
of the current meta-analysis were aggregated according to a random effects model, the effect sizes
were virtually identical to those effect sizes aggregated according to a fixed effects model. However,
the range of the confidence intervals for these effect sizes were slightly larger for data aggregated
according to a random effects model.
832
RUDOLPH ET AL.
For each path-analytic model, a weighted correlation matrix was submitted to
Structural Equation Modelling (EQS). Results reported in this review are based
on the maximum likelihood estimation procedure. Each correlation within a
correlation matrix was, however, based on a different number of subjects (see
Table 3), which can potentially bias parameter estimates. To run the analyses in
EQS, the sample size was fixed to reflect the correlation coefficient based on the
smallest number of subjects. This is a pragmatic but conservative approach
because it results in slightly inflated standard errors for each parameter estimate
in comparison to using the average sample size as the functional sample size. To
test the fit of each model, the w2 likelihood ratio test was used. However,
because this test is dependent on sample size, and we were dealing with
extremely large sample sizes, the comparative fit index (CFI; Bentler, 1990) was
used as an index of descriptive fit. As Bentler (1990) pointed out, CFI values of
greater than .93 are indicative of well-fitting models.
Four primary path models were used to examine the relations specified in the
judgments of responsibility models (predicting Help Giving and Aggression,
respectively; see Figure 2). Models 1 and 2 were selected because they capture
the hypothesised ordering of a motivated episode, but test whether thought, as
well as emotion, can be considered a proximal determinant of behaviour. This is
a central issue when considering the relations between thinking, feeling, and
acting. In addition, many of the primary studies did not report the correlation
between Anger and Sympathy, which precluded us from testing Models 1 and 2
based on the maximum number of subjects available from the pool of primary
studies. For this reason, we developed two additional models for testing in which
single affective states (rather than Sympathy and Anger within the same model)
were used.
The weighted correlation coefficients from the univariate analyses were used
as the unit of analysis to test these models. In Model 1, direct paths from
Controllability/Responsibility (hereafter referred to as Control) to Sympathy,
Anger, and Help Giving (Aggression) were specified. In addition, direct paths
from Sympathy and Anger to Help Giving (Aggression) were hypothesised.
Model 2 was identical to Model 1 except that this model did not specify a direct
path from Control to Help Giving (Aggression). The plausibility of this path was
evaluated using the w2 difference test (Dw2), comparing the fit of Model 1 to the
fit of Model 2. Because Model 2 is nested within Model 1, a nonsignificant Dw2
suggests that the models fit equally well and that the more parsimonious model
(Model 2 in this case) is the ``best-fitting'' model (see Bollen, 1989). If a
significant difference results, the model with the lowest w2 value is deemed the
``best-fitting'' model. Each of these models tested components of Models 1 and
2. In Model 3, direct paths from Control to Sympathy as well as to Help Giving
(Aggression) were specified. In addition, a direct path from Sympathy to Help
Giving (Aggression) was hypothesised. In Model 4, direct paths from Control to
Anger as well as to Help Giving (Aggression) were specified. In addition, a
HELP GIVING AND AGGRESSION
833
Figure 2. Theoretical models of helping behaviour analysed by means of path analyses.
834
RUDOLPH ET AL.
direct path from Anger to Help Giving (Aggression) was hypothesised.4 The
path coefficients for Models 3 and 4 were similar to those for Models 1 and 2.
For this reason, our description of the results focuses primarily on Models 1 and
2. However, for a complete description of the results, we present the path
coefficients for Models 3 and 4 in Table 4.
We also tested three additional nonpredicted paths that start with emotions to
test alternative explanations of the data. That is, models with direct paths from
emotion to cognition to behaviour for both help giving and aggression were
tested. These models turned out to produce nonsignificant results and poor fit
(CFI always < .50).
After the initial path analyses, the four models were tested at levels of the
moderator variables described earlier (see Study categorisation subsection). A
formal multigroup analysis was not performed for each moderator variable
because of the inappropriateness of this technique when a correlation (rather
than a covariance) matrix is used as the unit of analysis. Thus, relative differences in the magnitude of path coefficients between levels of the moderator
variables are discussed in terms of practical effect size differences rather than
statistical effect size differences. Path differences of .300, which correspond to
medium effect sizes (Cohen, 1988), were considered as practical differences.
RESULTS
Correlation and path coefficients for the help giving
models
Weighted correlation coefficients. Overall weighted correlation coefficients
between Control, Sympathy, Anger, and Help Giving as well as the number of
individuals that each correlation is based on are presented in the top half of
Table 3. All correlations were significant at p < .001 and were medium to large
effect sizes. Control was negatively related to Sympathy and Help Giving, but
was positively related to Anger. Sympathy was positively related, while Anger
was negatively related, to Help Giving. Sympathy and Anger were negatively
related. We shall refer to this configuration as the ``Predicted Data Pattern''. In
sum, more Help Giving was associated with less Control, more Sympathy and
less Anger. The Q statistic was significant (p < .001) for each weighted
correlation coefficient, suggesting that the variance associated with these
correlations was heterogeneous, and that moderators for each correlation
potentially exist.
4
Models 3 and 4 are just-identified models. Therefore, each of these models will perfectly
reproduce the observed correlations and fit perfectly.
835
7.35
7.45
7.32/7.39
7.43/7.46
CI
2509
7416
N
.61
7.31
.52
7.39
r
.59/.63
7.28/7.35
.50/.54
7.37/7.42
CI
Anger
4448
1976
7140
5484
N
.49
7.44
.56
7.25
.42
7.24
r
.47/.52
7.41/7.47
.54/.58
7.23/7.27
.40/.44
7.22/7.26
CI
Behaviour
3719
2377
3458
6840
7382
6800
N
Note: High scores on the variables reflect more controllability, sympathy, anger, and behaviour (help giving, aggression). All correlations are significant at
p < .001.
Aggression
Controllability
Sympathy
Anger
Help giving
Controllability
Sympathy
Anger
r
Sympathy
TABLE 3
Overall weighted correlations among controllability, sympathy, anger, and behaviour (Help Giving, Aggression)
836
RUDOLPH ET AL.
Overall judgements of responsibility models. Using the weighted correlations from the top of Table 3, the fit of Models 1 and 2 and the paths of all four
models were evaluated. Both Model 1, w2 (1, N = 5484) = 234.26, p < .05, CFI =
.95, and Model 2, w2 (2, N = 5484) = 234.49, p < .05, CFI = .94, fit well
descriptively. The path coefficients for these models are presented in the top half
of Table 4. The Control-Sympathy path coefficient was significant and negative
and the Control-Anger path coefficient was significant and positive in the
relevant models; more Control was associated with more Anger and less
Sympathy. The Sympathy-Help Giving path coefficient was significant and
positive and the Anger-Help Giving path coefficient was significant and
negative in all relevant models; more Sympathy and less Anger was associated
with more Help Giving. Finally, the Control-Help Giving path coefficient was
significant and negative. The Control-Help Giving path coefficient, however,
was smaller in magnitude than Sympathy-Help Giving and Anger-Help Giving
path coefficients in all relevant models. To test whether or not dropping the
Control-Help Giving path would result in a decrement in overall model fit, a Dw2
was used to compare Models 1 and 2. Eliminating the Control-Help Giving path
did not result in a significant decrement in model fit, Dw2 (1) = 0.23, p > .05;
Model 2 therefore was the most parsimonious model and considered the best fit
to the data.
We also evaluated the fit of three other nonpredicted models. First, a model
with direct paths from Sympathy and Anger to Control as well as a direct path
from Control to Help Giving was tested (i.e., a feeling-thought-action sequence).
This model did not fit well, w2 (3, N = 5484) = 1669.90, p < .05, CFI = .61.
Second, a model identical to the one just described with the addition of direct
TABLE 4
Path coefficients for Help Giving and Aggression models
Model 1
Model 2
Model 3
Model 4
Help Giving models
Control-Sympathy
Control-Anger
Control-Help Giving
Sympathy-Help Giving
Anger-Help Giving
7.45*
.52*
7.05*
.37*
7.07*
7.45*
.52*
±
.39*
7.09*
7.45*
±
7.08*
.39*
±
±
.52*
7.15*
±
7.17*
Aggression models
Control-Sympathy
Control-Anger
Control-Aggression
Sympathy-Aggression
Anger-Aggression
7.35*
.61*
.17*
7.27*
.38*
7.35*
.61*
±
7.30*
.48*
7.35*
±
.38*
7.31*
±
±
.61*
.24*
±
.42*
Note: ± is not part of target model. * p < .05.
HELP GIVING AND AGGRESSION
837
paths from Sympathy and Anger to Help Giving was tested. Likewise, this
model did not fit well, w2 (1, N = 5484) = 904.70, p < .05, CFI = .79. And third, a
model with direct paths from Sympathy, Anger, and Controllability to Help
Giving was specified. This model also did not fit well, w2 (3, N = 5484) =
3237.38, p < .05, CFI = .25.
In sum, as predicted by theory, the results suggest that the emotions of Anger
and Sympathy, rather than Control, were more proximate determinants of Help
Giving. Control was directly related to both Anger and Sympathy, and thus was
indirectly associated with Help Giving through the two emotion mediators.
Moderator analysis. The four target models were then tested at each level
of the moderator variables (i.e., type of culture, type of investigation, type of
sample, year the study was conducted, publication status, and group in need of
help). No practical differences were found between levels of culture, type of
investigation, type of sample, year of study, publication status, and group in
need of help. The path coefficients for these models were similar to those found
in the overall models presented above.
As the simulational character of many studies in the attribution domain has
sometimes been criticised, the moderator analysis for type of investigation is
especially instructive. As can be seen from Table 5 (top half), no practical
differences were obtained between the path coefficients for the type of inves-
TABLE 5
Path coefficients for Help Giving and Aggression (Model 1) for
moderator type of investigation
Simulated level
Real event level
Help giving
Control-Sympathy
Control-Anger
Control-Help Giving
Sympathy-Help Giving
Anger-Help Giving
7.47*
.54*
7.07*
.40*
7.04*
7.37*
.45*
.01
.28*
7.12*
Aggression
Control-Sympathy
Control-Anger
Control-Aggression
Sympathy-Aggression
Anger-Aggression
7.40*
.63*
.19*
7.31*
.34*
7.28*
.56*
.18*
7.13*
.42*
Note: * p < .05. For help giving (Model 1), there were 25 studies including
simulational data and 6 studies including real event data. For aggression
(Model 1), there were 15 studies including simulational data and 6 studies
including real event data.
838
RUDOLPH ET AL.
tigation moderator (simulation vs. recounting or participation in a real event).
Additionally, for both simulational and real event data, the size of the coefficients and their signs are in accord with Weiner's theory of social conduct.
Therefore, the present results are not limited to simulational data or ``thought
experiments'', but apply as well to the analysis of real events.
Summary
In sum, Model 2 provided a very good fit with the data. In this model, perceptions of control influence anger and sympathy, which in turn determine help
giving. Hence, thoughts are distal, but not proximal, determinants of behaviour
through their influence on emotions. No practical differences were found for the
six moderator variables. In addition, models specifying a feeling-thinking-action
sequence, a feeling-action sequence, and a feeling-and-thinking-action sequence
were all disconfirmed. The entire pattern of data is very supportive of the
attributional position as espoused in the introduction.
Correlation and path coefficients for the aggression
models
Weighted correlation coefficients. Overall weighted correlation coefficients
between Control, Sympathy, Anger, and Aggression as well as the number of
individuals that each correlation is based on are presented in the bottom half of
Table 3. All correlations were significant at p < .001, with medium to large
effect sizes, and conforming entirely to the Predicted Data Pattern. In sum, more
Aggression was associated with less Sympathy, more Control, and more Anger.
The Q-statistic was significant (p < .001) for each weighted correlation
coefficient, suggesting that the variance associated with these correlations was
heterogeneous, and that moderators for each correlation potentially exist.
Overall judgements of responsibility models. Using the weighted correlations from the bottom half of Table 3, the fit of Models 1 and 2 and the paths of
all four models were tested. Model 1, which had a path from Control to
Aggression, w2 (1, N = 1976) = 33.67, p < .05, CFI = .99, fit well descriptively,
whereas Model 2, w2 (2, N = 1976) = 89.74, p < .05, CFI = .96, also fit well
descriptively. The path coefficients for these models are presented in the bottom
half of Table 4. To test whether or not dropping the Control-Aggression path
would result in a decrement in overall model fit, a Dw2 was used comparing
Models 1 and 2. Eliminating the Control-Aggression path resulted in a
significant decrement in model fit, Dw2 (1) = 56.07, p < .05; Model 1 therefore
was determined to be the best fit to the data.
As in the help giving meta-analysis, we also evaluated other models that
could explain the relations among Control, Sympathy, Anger, and Behaviour
(Aggression). First, a model with direct paths from Sympathy and Anger to
HELP GIVING AND AGGRESSION
839
Control as well as a direct path from Control to Aggression was tested (i.e., a
feeling-thinking-action sequence). This model did not fit well, w2 (3, N = 1976)
= 696.44, p < .05, CFI = .69. Second, a model identical to the one just described
with the addition of direct paths from Sympathy and Anger to Aggression was
tested. This model also did not fit well, w2 (1, N = 1976) = 199.55, p < .05, CFI =
.91. And finally a model with direct paths from Sympathy, Anger, and Control to
Aggression was specified. Similarly, this model did not fit well, w2 (3, N = 1976)
= 1210.95, p < .05, CFI = .46.
The results suggest that the emotions of Anger and Sympathy as well as
Control were all proximate determinants of Aggression (see Model 1). In
addition, Control was directly related to Anger and Sympathy, and thus also was
indirectly associated with Aggression through the two emotion variables.
Moderator analysis. The four target models were tested at each level of the
moderator variables (i.e., year a study was published, publication status, culture,
type of sample, and type of investigation). No practical differences were found.
The path coefficients for these models were similar to those found in the overall
models presented above. Practical differences were found for one moderator
variable: Publication status (published, unpublished). Using the practical
significance level of .30, the Sympathy-Aggression path coefficient, while
significant and negative in models from unpublished manuscripts (path
coefficient = 7.09, p < .05), was practically different from the path coefficient
in models to be found in published manuscripts (path coefficient = 7.45, p <
.05). In addition, the Anger-Aggression path coefficient, while significant and
positive in models from unpublished data (path coefficient = .49, p < .05), was
practically different from the same path based on models tested with data from
published manuscripts (path coefficient = .08, p < .05).
For the moderator type of investigation (simulation vs. recounting or participation in a real event), the path coefficients are presented in the bottom half of
Table 5. All coefficients are very similar in size and sign; no practical differences were obtained. Therefore, as for the helping studies, our findings are not
limited to imagined courses of action. Rather, we also found strong support for
Weiner's theory of social conduct using ``real'' behaviour as a dependent
variable.
Summary
In sum, Model 1 provided the best fit to the data. In this model, perceptions of
control as well as emotions are proximal determinants of aggression, and control
beliefs also influence emotional experiences. Only one of the five moderator
variables, publication status, yielded any differences, and this unexplainable
difference was confined to the Sympathy-Aggression path. In addition, alter-
840
RUDOLPH ET AL.
native models specifying a feeling-thinking-action sequence, a feeling-action
sequence and a thinking-and-feeling-action sequence were all disconfirmed.
COMPARING THE HELPING AND AGGRESSION
MODELS
In the introduction, it was contended that both pro- and antisocial behaviour can
be explained with the same conceptual framework. To what extent do our data
support this assertion? The answer to this is: It depends. It depends on how
molar versus molecular an analysis is considered, or if the focus is on the most
general law as opposed to modified specific instances of the general law(s).
At the most molar level, it can be concluded from our data that behaviour is a
function of cognition and affect. Perhaps this provides one (among many)
alternatives to the equally broad statement that behaviour is a function of the
person and the environment. A somewhat more precise formulation of this
general law is that thinking is a distal determinant of motivated or goal-seeking
behaviour, whereas emotion is proximal. This principle was partially supportedÐemotions were indeed proximal in accounting for both help giving and
aggression, but thoughts (i.e., perceptions of causal control and responsibility
inferences) were distal determinants in help giving, whereas they also proximally affected aggression. Another way of stating these findings is that
aggression has more immediate attributional determinants than does help giving,
with hostile responding being an immediate product of both the head and the
heart, whereas helping ultimately involves only the heart. Based on these
findings, one might attempt to collect charity funds by eliciting sympathy (e.g.,
from ``bleeding hearts''), but stop aggression by appealing to reason and/or
emotion (see Lerner, Goldberg, & Tetlock, 1998).
Why might this be the case? Here one can offer only guarded speculations. It
may be that most help giving or social support (with the exception of extreme
personal sacrifice) has relatively minor personal consequences. On the other
hand, even minor aggressive retaliation may come with a great cost. Thus,
having thoughts as proximal determinants in hostile contexts may be quite
functional in terms of personal survival. As intimated above, however, one must
be weary of such conclusions. The type of aggression reviewed in the metaanalysis primarily was often laboratory-induced or involved role-playing, hence
minimising the amount of aroused anger and thus perhaps rendering thinking
more important in influencing action. Further, the help giving studies typically
did not involve great sacrifice: Cognitive work may have been minimised the
limited set of data available. Speaking against this potential criticism, however,
is the fact that the recollections of autobiographical helping and aggression
incidents (e.g., George, 1997; Rudolph & Greitemeyer, 2001) included severe
forms of aggression and forms of helping that required great costs, and the same
data pattern was found for these studies as well.
HELP GIVING AND AGGRESSION
841
Another issue of importance concerns the magnitude of the relations that
were exhibited in the two motivational domains. Given an attributional (causal)
perspective, what is more predictable, help giving or aggression? That is, which
of these areas is ``better'' suited for an attributional explanation in terms of the
amount of explained variance? Examination of the right-hand columns of Table
1 and Table 2 clearly reveals that perceptions of responsibility and emotions
(particularly anger) are more predictive of aggression than they are of helping.
Therefore, in a final exploratory analysis, the overall model fit for Models 1
and 2 from the help giving meta-analysis was compared to the overall model fit
for Model 2 from the aggression meta-analysis. While these models cannot be
directly compared statistically because they are based on different datasets, we
used the w2-values, CFI values, and the expected cross validation index (ECVI;
Browne & Cudeck, 1989, 1993) to make descriptive comparisons only. The
ECVI is useful when choosing among several a priori models, and is typically
used when a series of models are tested on the same dataset (see Roesch, 1999).
Lower ECVI values are indicative of a model that is more likely to be reproduced in a validation sample, and by implication, to be the ``better-fitting''
model. It should be noted that we are using these indices in a strictly exploratory
fashion, and that our interpretations should be considered with great caution.
The indices of overall model fit for Model 1 in the aggression meta-analysis
were all indicative of better fit, w2(1, N = 1976) = 33.67, p < .05; CFI = .99;
ECVI = .0211, than the indices of overall model fit for both Model 1, w2(1, N =
5484) = 234.26, p < .05; CFI = .95; ECVI = .0445, and Model 2, w2(2, N = 5484)
= 234.49, p < .05; CFI = .94; ECVI = .0443, in the help giving meta-analysis.
This suggests that Model 1 in the aggression meta-analysis better represents the
aggression data than either Model 1 and Model 2 represent the help giving data.
However, all of these models do fit extremely well and the ECVI additionally
suggests that these models would be reproduced in a validation sample for both
help giving and aggression.
It can be reasoned, therefore, that help giving may have more determinants
outside the attribution-emotion framework than does aggression. Stated somewhat redundantly, a harmful intent, along with the anger this generates, apparently is a more salient or encompassing influence of aggressive retaliation than
are the corresponding cognitive and emotional determinants of helping.
Again, why might this be the case? One answer that readily comes to mind
(but is not very exciting) is that in an attributional context one is not concerned
with all aggression, but rather only with aggressive retaliation. This greatly
limits the number of determinants of hostile behaviour. When one is attacked, it
is quite likely that attributional thoughts are elicited in that this behaviour is
negative and possibly unexpected and important (see Weiner, 1985b). Subsequent behaviour will be very much guided by the answers to questions such
``Why did he hit me?'' or ``Was it on purpose?''. On the other hand, it may be
that when another is approached for help, a number of nonattributional questions
842
RUDOLPH ET AL.
also immediately come to mind, including ``Is this person related to me?'', ``Do
I have the extra money to help?'', or ``How great is the need?''. That is,
nonattributional factors appear to be more likely to play a role in help giving
than they are in aggressive retaliation. Of course, any number of other speculative explanations can also be proposed; answers obviously await more
research and theoretical insight.
GENERAL DISCUSSION
It has already been well documented that perceptions of causal controllability
and, by implication, beliefs about personal responsibility, elicit the emotions of
anger and sympathy (for a review, see Weiner 1995). These feelings, with or
without causal thoughts, give rise to pro- or antisocial behaviour. These statements explain only some instances of help giving and some instances of
aggression. As stated in the introduction, our goal was neither to propose a
complete theory of behaviour across all motivational areas nor to explain all
actions within a content domain. Rather, our intent was to offer a theory at an
intermediate level of generality, and to document that the same theoretical
scaffold can account for a restricted set of phenomena within more than one
motivational area. The present data suggest that this goal has been reached. It is
also evident that thoughts are proximal determinants of reactive aggression, but
only are distal determinants of help giving, which is a previously undiscovered
empirical generalisation made possible by the present comparisons between the
meta-analyses. Hence, thoughts are (indirectly) influential on helping behavior
because of their emotional consequences. Thus, there are different theoretical
sequences, albeit subtle, that characterise an attributional approach in these two
domains.
It may be that other documented determinants of helping and aggression also
exert their influence through perceptions of controllability and/or the emotion of
anger. However, one must be modest about such claims and it is not our contention that all (or even much of) the variance in help giving and aggression
accounted for by these other variables can be linked to the concepts advanced by
attribution theorists. For example, considering help giving, it may be that
relatives are held less responsible for an untoward state or event than are nonrelatives, which explains the greater help giving exhibited towards relatives in
need (see Greitemeyer, Rudolph, & Weiner, 2003). Further, perhaps helping is
increased when in a good mood because positive mood states colour perceptions
of responsibility in a positive direction, such that when feeling good, others are
perceived as less responsible for their plights. In a similar manner and now in the
domain of aggression, it could be contended that on hot days individuals tend to
overestimate negative intents of others and/or are more prone to experience
anger, hence accounting for the temperature-aggression relation. One could go
about this reinterpretation enterprise for many (but surely not all) of the helping
HELP GIVING AND AGGRESSION
843
and aggression literatures. However, as already indicated, this is suggested with
caution and without the belief that these other variables will be rendered merely
distal antecedents within an attribution-affect-action sequence. Nonetheless,
such integrative research is surely called for.
Perhaps the main issue remaining to be discussed is whether the theory
advocated here can be extended to other domains of motivated behaviour as
well. The most obvious place to search for theoretical generality is in the
achievement domain, where attribution theory has played a major role in
understanding reactions to success and failure (see Weiner, 1986). Here, it is
useful to distinguish how others are evaluated following an achievement outcome (which falls under the rubric of interpersonal motivation) from how one
performs following that outcome (which is subsumed within intrapersonal
motivation).
It has been well documented that the two most common perceived causes of
success and failure in achievement settings are ability and effort. Success tends
to be ascribed to high ability and/or hard work, while failure is perceived as
caused by the absence of ability and/or insufficient effort. There is unequivocal
evidence that failure attributed to lack of effort is evaluated more negatively as
compared to failure ascribed to lack of ability (see Weiner, 1986). Furthermore,
this pattern of appraisal is considered ``just'' or ``fair'' (see Farwell & Weiner,
1996).
Ability and effort differ in their personal controllability, with ability judged
as similar to aptitude and construed as an uncontrollable cause, whereas effort is
subject to volitional alteration and thus is seen to be personally controllable.
Hence, one is responsible for failure due to not trying, but not for failure due to
the absence of ability. It is suggested that lack of effort and responsibility
inferences, in turn, give rise to anger and negative evaluation, whereas lack of
ability ascriptions and the absence of responsibility give rise to sympathy and
little negative sanction.
Unfortunately, unlike the voluminous data in the fields of help giving and
aggression, there is not an empirical literature that includes all stages in this
sequence. There are numerous studies capturing one or two of the linkagesÐfor
example, lack of effort is perceived as more controllable than lack of ability;
lack of effort elicits more anger and less sympathy as a cause of failure than does
lack of ability; and lack of effort is evaluated more negatively when causing
failure than is the absence of ability (see Weiner, 1985a, 1986). Thus, although
studies that include attributions, affect, and evaluative reactions have not been
conducted, there is reason to believe that this research will produce the same
pattern of relations as shown in Tables 3 and 4 (the thinking-behaviour link may
or may not be hypothesised).
An explanation of achievement performance, rather than achievement evaluation, requires a leap to another level of theoretical generality. It has been well
documented that attribution of failure to the absence of ability (aptitude) is
844
RUDOLPH ET AL.
particularly debilitating, whereas ascriptions to lack of sufficient effort serves to
facilitate subsequent performance. The question is, what are the mediators of
these attribution-performance relations?
It has been proposed elsewhere in detail thatÐin addition to disparities in
personal controllabilityÐgiven a desired goal, attribution to lack of effort for
failure gives rise to feelings of guilt. Guilt, in turn, serves as a goad to future
action. On the other hand, ascriptions of failure to low ability generate feelings
of humiliation, embarrassment, and shame. These emotions promote withdrawal
and are motivational inhibitors (see Weiner, 1986).
Note that the emotions of anger and sympathy that were a key component of
the interpersonal theory of motivation have been supplanted with the affects of
guilt and shame in the intrapersonal domain. Thus, although the specific content
of the theory has been altered, the most general rule, that a motivational
sequence progresses from attributional thinking to its elicited emotion and then
to behaviour, has been retained (again with uncertainty regarding the responsibility-behaviour link).
Unfortunately, a literature does not exist to evaluate these motivational
sequences in the achievement field. Thus far, only two studies have included
attributions, emotions, and performance (Covington & Omelich, 1984; Van
Overwalle, Mervielde, & DeSchuyter, 1995). These investigations have yielded
mixed results, but with sufficient positive findings to view the cup as half full
rather than half empty. Specifically, Covington and Omelich (1984) report that
personal attribution of failure to lack of ability correlates positively with shame
and humiliation, and that humiliation (but not shame) is negatively related to test
performance. However, contrary to predictions, ascriptions to low ability also
correlated positively with reported guilt. Furthermore, self-attributions for
school failure to lack of effort correlated positively with guilt, as predicted,
which augmented test performance. However, again inconsistent with the theory, effort ascriptions related positively with reported humiliation and shame.
Thus, affective discriminations were only weakly displayed. In the investigations by Van Overwalle et al. (1995) a different pattern of results were obtained,
with attributions linked as expected to emotions, but school performance was
unrelated to emotional experience.
Again, however, a number of studies examining one link in the sequence
(e.g., lack of effort is perceived as more controllable than lack of ability; controllable causality gives rise to guilt, uncontrollable causality to shame; lack of
effort ascriptions are associated with better future performance than attributions
to low ability) have yielded the anticipated results (see Weiner, 1986). Thus,
there is reason for optimism regarding the hypothesised motivational system.
It now can be tentatively concludedÐor at least not unreasonably suggestedÐthat aspects of helping, aggression, performance evaluation, and
achievement strivings can be subsumed within the same theoretical framework.
We believe that no other motivational system can make this claim. Perhaps other
HELP GIVING AND AGGRESSION
845
areas as well can be reinterpreted from this point of view, and perhaps the
general theoretical scaffold can be recast so that thoughts other than causal
controllability and responsibility, affects other than anger, pity, guilt, and shame,
and relations other than sequential, can be incorporated. In this manner, an even
more general motivational formulation will have been reached.
Manuscript received 15 February 2002
Revised manuscript received 14 March 2003
REFERENCES
* Allred, K. G., Mallozzi, J. S., Matsui, F., & Raia, C. P. (1997). The influence of anger and
compassion on negotiation performance. Organizational Behavior and Human Decision Processes, 70, 175±187.
Anderson, C., & deNeve, K. M. (1992). Temperature, aggression, and the negative affect escape
model. Psychological Bulletin, 111, 347±351.
Averill, J. R. (1982). Anger and aggression: An essay on emotion. New York: Springer.
Baron, R. M., & Kenny, D. A. (1986). The moderator-mediator variable distinction in social psychological research: Conceptual, strategic, and statistical consideration. Journal of Personality
and Social Psychology, 51, 1173±1182.
Batson, C. D. (1998). Altruism and prosocial behavior. In D. T. Gilbert, S. T. Fiske, & G. Linzey
(Eds.), Handbook of social psychology (4th ed., Vol. 2, pp. 282±316). New York: McGraw-Hill/
Oxford University Press.
Burnstein, E., Crandall, C., & Kitayama, S. (1994). Some neo-Darwinian rules for altruism:
Weighing cues for inclusive fitness as a function of the biological importance of the decision.
Journal of Personality and Social Psychology, 67, 773±789.
Becker, B. J., & Schramm, C. M. (1994). Examining explanatory models through research synthesis.
In T. D. Cook, H. M. Cooper, D. S. Corday, H. Hartmann, L. V. Hedges, R. J. Light, T. A. Louis,
& F. Mosteller (Eds.), The handbook of research synthesis (pp. 357±381). New York: Russell
Sage.
Bentler, P. M. (1990). Comparative fit indices in structural models. Psychological Bulletin, 107,
238±246.
* Betancourt, H. (1990). An attribution-empathy model of helping behavior. Personality and Social
Psychology Bulletin, 16, 573±591.
* Betancourt, H., & Blair, I. (1992). A cognition (attribution)-emotion model of violence in conflict
situations. Personality and Social Psychology Bulletin, 18, 343±350.
* Betancourt, H., Hardin, C., & Manzi, J. (1995). Beliefs, value orientation, and culture in attribution
processes and helping behavior. Journal of Cross-Cultural Psychology, 23, 179±195.
Bollen, K. A. (1989). Structural equations with latent variables. New York: Wiley.
Browne, M. W., & Cudeck, R. (1989). Single-sample cross validation indices for covariance
structures. Multivariate Behavioral Research, 24, 445±455.
Browne, M. W., & Cudeck, R. (1993). Alternative ways of assessing model fit. In K.A. Bollen & J. S.
Long (Eds.), Testing structural equation models (pp. 136±162). Newbury Park, CA: Sage.
* Byrne, C. A., & Arias, I. (1997). Marital satisfaction and marital violence: Moderating effects of
attributional processes. Journal of Family Psychology, 11, 188±185.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ:
Erlbaum.
Cook, T. D., Cooper, H. M., Cordray, T. S., Hartmann, H., Hedges, L. V., Light, R. J., Louis, T. A.,
& Mosteller, F. (Eds.). (1992). Meta-analysis for explanation: A casebook. New York: Russell
Sage.
846
RUDOLPH ET AL.
Covington, M. V., & Omelich, C. L. (1984). The trouble with pitfalls: A reply to Weiner's critique of
attribution research. Journal of Educational Psychology, 76, 1199±1213.
* Dagnan, D., Trower, P., & Smith, R. (1998). Care staff responses to people with learning disabilities and challenging behavior: A cognitive-emotional analysis. British Journal of Clinical
Psychology, 37, 59±68.
Dijker, A. J., & Koomen, W. (2003). Extending Weiner's attribution-emotion model of stigmatization of ill persons. Basic & Applied Social Psychology, 25, 51±68.
Dovidio, J. F., Piliavin, J. A., Gaertner, S. L., Schroeder, D. A., & Clark, R. D., III (1991). The
Arousal-Cost-Reward Model and the process of intervention: A review of the evidence. In M. S.
Clark (Ed.), Review of personality and social psychology (Vol. 12, pp. 86±118). Newbury Park,
CA: Sage.
Enzle, M. E., & Shopflocher, D. (1978). Instigation of attribution processes by attribution questions.
Personality and Social Psychology Bulletin, 4, 595±599.
Farwell, L., & Weiner, B. (1996). Self-perceptions of fairness in individual and group contexts.
Personality and Social Psychology Bulletin, 22, 867±881.
* Feather, N. T. (1996). Reactions to penalties for an offense in relation to authoritarianism, values,
perceived responsibility, perceived seriousness, and deservingness. Journal of Personality and
Social Psychology, 71, 571±587.
* George, D. (1997). An attribution-affect-efficacy model of helping behavior. Unpublished manuscript.
* George, D., Harris, S., & Price, I. (1998). Determinants of helping behavior: An attributional
perspective. Manuscript submitted for publication.
* Graham, S., & Hoehn, S. (1995). Children's understanding of aggression and withdrawal as social
stigmas: An attributional analysis. Child Development, 66, 1143±1161.
* Graham, S., Hudley, C., & Williams, E. (1992). Attributional and emotional determinants of
aggression among African-American and Latino young adolescents. Developmental Psychology,
28, 731±740.
Graham, S., & Weiner, B. (1991). Testing judgments about attribution-emotion-action linkages: A
lifespan approach. Social Cognition, 9, 254±276.
* Graham, S., Weiner, B., & Zucker, G. S. (1997). An attributional analysis of punishment goals and
public reactions to O. J. Simpson. Personality and Social Psychology Bulletin, 23, 331±346.
* Greitemeyer, T., & Rudolph, U. (2003). Help giving and aggression from an attributional perspective:
Why and when we help or retaliate. Journal of Applied Social Psychology, 33, 1069±1087.
* Greitemeyer, T., Rudolph, U., & Weiner, B. (2003). Whom would you rather help: An acquaintance
not responsible for her plight or a responsible sibling? Journal of Social Psychology, 143, 331±340.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. Orlando, FL: Academic
Press.
Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley.
* Higgins, N. C., & Watson, C. E. (1995). Negative life experiences predict attributional and
emotional determinants of aggression. Paper presented at the Annual Meeting of the American
Association for the Advancement of Science, Pacific Division. University of British Columbia.
* Ho, R., & Venus, M. (1995). Reactions to a battered woman who kills her abusive spouse: An
attributional analysis. Australian Journal of Psychology, 47, 153±159.
Huesmann, L. R., & Eron, L. D. (Eds.). (1986). Television and the aggressive child: A cross-national
comparison. Hillsdale, NJ: Erlbaum.
* Johnson, T. E., & Rule, B. G. (1986). Mitigating circumstance information, censure, and aggression. Journal of Personality and Social Psychology, 50, 537±542.
* Karasawa, T. (1991). The effects of onset and offset responsibility on affects and helping judgments. Journal of Applied Social Psychology, 21, 482±499.
Kelley, H. H. (1992). Common-sense psychology and scientific psychology. Annual Review of
Psychology, 43, 1±23.
HELP GIVING AND AGGRESSION
847
Kelley, H. H., & Michela, J. (1980). Attribution theory and research. Annual Review of Psychology,
31, 457±501.
* Kojima, M. (1992). An analysis of attributional processes in helping behavior. Bulletin of the
Tamagawa Guken Junior College for Women, 17, 57±83.
LataneÂ, B., & Darley, J. M. (1970). The unresponsive bystander: Why doesn't he help? New York:
Appleton-Century-Croft.
Lazarus, R. S. (1991). Emotion and adaptation. New York: Oxford University Press.
Lerner, J. S., Goldberg, J. H., & Tetlock, P. E. (1998). Sober second thoughts: The effects of
accountability, anger, and authoritarianism on attributions of responsibility. Personality and
Social Psychology Bulletin, 24, 563±574.
* Matsui, T., & Matsuda, Y. (1992). Testing for the robustness of Weiner's attribution-affect model of
helping judgments for exogenous impact. Unpublished manuscript, Rikkyo University, Tokyo,
Japan.
* Menec, V. H., & Perry, R. P. (1995). Reactions to stigmas: The effects of target's age and
controllability of stigmas. Journal of Aging and Health, 7, 365±383.
* Menec, V. H., & Perry, R. P. (1998). Reactions to stigmas: Testing an attribution-affect-help
judgment model with Canadian students. Journal of Social Psychology, 138, 443±453.
* Meyer, J. P., & Mulherin, A. (1980). From attribution to helping: An analysis of the mediating effects of affect and expectancy. Journal of Personality and Social Psychology, 39,
201±210.
Premack, S. L., & Hunter, J. E. (1988). Individual unionization decisions. Psychological Bulletin,
103, 223±234.
* Reisenzein, R. (1986). A structural equation analysis of Weiner's attribution-affect model of
helping behavior. Journal of Personality and Social Psychology, 50, 1123±1133.
* Rodrigues, A., & Lloyd, K. L. (1998). Reexamining bases of power from an attributional perspective. Journal of Applied Social Psychology, 28, 973±997.
Roesch, S. C. (1999). Modeling stress: A methodological review. Journal of Behavioral Medicine,
22, 249±269.
* Rudolph, U., & Greitemeyer, T. (2001). Autobiographical recollections of pro- and anti-social
behavior: Evidence for Weiner's theory of responsibility. Manuscript in preparation, Technical
University Chemnitz.
* Schmidt, G., & Weiner, B. (1988). An attribution-affect-action theory of behavior: Replications of
judgments of help giving. Personality and Social Psychology Bulletin, 14, 610±621.
Shadish, W. R. (1996). Meta-analysis and the exploration of causal mediating processes: A primer of
examples, methods, and issues. Psychological Methods, 1, 47±65.
Shadish, W. R., & Haddock, C. K. (1994). Combining estimates of effect size. In T. D. Cook, H. M.
Cooper, D. S. Corday, H. Hartmann, L. V. Hedges, R. J. Light, T. A. Louis, & F. Mosteller
(Eds.), The handbook of research synthesis (pp. 261±281). New York: Russell Sage.
Shadish, W. R., & Sweeney, R. B. (1991). Mediators and moderators in meta-analysis: There's a
reason we don't let dodo birds tell us which psychotherapies should have prizes. Journal of
Consulting and Clinical Psychology, 39, 883±893.
* Sharrock, R., Day, A., Qazi, F., & Brewin, C. (1990). Explanation by professional care staff,
optimism and helping behavior An application of attribution theory. Psychological Medicine, 20,
849±855.
Smith, C. A., & Ellsworth, P. C. (1985). Patterns of cognitive appraisal in emotion. Journal of
Personality and Social Psychology, 48, 813±848.
* Steins, G., & Weiner, B. (1999). The influence of perceived responsibility and personality characteristics on the emotional and behavioral reactions to persons with AIDS. Journal of Applied
Social Psychology, 139, 487±495.
* Stiensmeyer-Pelster, J. (1995). Aggressive behavior among children and adolescents. Paper presented at the Social Psychology Colloquium at the University of Konstanz, Germany.
848
RUDOLPH ET AL.
* Stiensmeyer-Pelster, J., & Gerlach, H. (1997). Aggressive behavior among children and adolescents from an attribution theoretical point of view [Aggressives Verhalten bei Kindern und
Jugendlichen aus attributionstheoretischer Sicht]. German Journal of Educational Psychology,
11, 203±209.
* Sunmola, A. (1994). Perceived controllability, affective reactions of sympathy and anger as
determinants of subjects' tendency to offer help to government. IFE-Psychologia An International Journal, 2, 113±122.
* Thompson, S. C., Medvene, L. J., & Freedman, D. (1995). Care-giving in the close relationships of
cardiac patients: Exchange, power, and attributional perspectives on caregiver resentment.
Personal Relationships, 2, 125±142.
* Vala, J., Monteiro, M., & Leyens, J. P. (1988). Perception of observer's ideology and actor's group
membership. British Journal of Social Psychology, 27, 231±237.
Van Overwalle, F., Mervielde, I., & DeSchuyter, J. (1995). Structural modeling of the relationships
between attributional dimension, emotions, and performance of college freshman. Cognition and
Emotion, 9, 59±85.
Vollmer, G. (1984). Mesocosm and objective knowledge. In F. M. Wuketits (Ed.), Concepts and
approaches to evolutionary epistemology towards an evolutionary theory of knowledge (pp.
69±121). Dordrecht: D. Reidel.
* Watson, J. E., & Higgins, N. C. (1999). A test of Weiner's (1995) responsibility judgment model:
Does the judgment target matter? Manuscript submitted for publication.
* Weiner, B. (1980a). A cognitive (attribution)-emotion-action model of motivated behavior: An
analysis of judgments of help giving. Journal of Personality and Social Psychology, 39, 186±200.
* Weiner, B. (1980b). May I borrow your class-notes? An attributional analysis of judgments of help
giving. Journal of Educational Psychology, 72, 676±681.
Weiner, B. (1985a). An attributional theory of achievement-related emotion and motivation. Psychological Review, 29, 548±573.
Weiner, B. (1985b). ``Spontaneous'' causal thinking. Psychological Bulletin, 97, 74±84.
Weiner, B. (1986). An attributional theory of motivation and emotion. New York: Springer.
Weiner, B. (1995). Judgments of responsibility: A foundation for a theory of social conduct. New
York: Guilford Press.
Weiner, B. (2000). Intrapersonal and interpersonal theories of motivation from an attributional
perspective. Educational Psychology Review, 12, 1±14.
* Weiner, B., & Graham, S. (1989). Understanding the motivational role of affect: Life-span research
from an attributional perspective. Cognition and Emotion, 3, 401±419.
* Weiner, B., Perry, R. P., & Magnusson, J. (1988). An attributional analysis of reactions to stigmas.
Journal of Personality and Social Psychology, 55, 738±748.
* Wingrove, J., & Bond, A. J. (1998). Angry reactions to failure on a cooperative computer game:
The effect of hostility, behavioral inhibition, and behavioral activation. Aggressive Behavior, 24,
27±36.
* Yamauchi, H., & Lee, K. (1999). An attribution-emotion model of helping behavior. Psychological
Reports, 84, 1073±1074.
Zucker, G. S. (1999). Attributional and symbolic predictors of abortion attitudes. Journal of Applied
Social Psychology, 29, 1218±1245.
* Zucker, G. S., & Weiner, B. (1993). Conservatism and perception of poverty: An attributional
analysis. Journal of Applied Social Psychology, 23, 925±943.
* Zumkley, H. (1981). The influence of different kinds of intentionality attributions on aggressive
behavior and activation [Der Einfluû unterschiedlicher Absichtsattributionen auf das Aggressionsverhalten und die Aktivierung]. Psychologische BeitraÈge, 23, 115±128.