Attribution Theory in Organizational Sciences: A Meta-Analysis

This report presents a meta-analysis of attribution theory in the organizational sciences, examining its predictive power on workplace outcomes. The study, published in The Academy of Management Perspectives, analyzes existing research to determine the significance of attributions in shaping employee behavior, leadership, and motivation. The authors focus on attributional dimensions like locus of causality, stability, and controllability, linking them to various organizational phenomena. The research highlights the underutilization of attribution theory in organizational studies compared to social psychology and aims to address the debate on its efficacy. Findings suggest that attributions have consistent effect sizes, comparable to other predictor variables, emphasizing their importance in understanding and managing organizational outcomes. The report concludes with suggestions for expanding attributional research to optimize its contribution to organizational sciences, emphasizing the need for further exploration of its role in various contexts.

S Y M P O S I U M
ATTRIBUTION THEORY IN THE ORGANIZATIONAL
SCIENCES: THE ROAD TRAVELED AND THE PATH AHEAD
PAUL HARVEY
University of New Hampshire
KRISTEN MADISON
Mississippi State University
MARK MARTINKO
University of Queensland
T. RUSSELL CROOK
TAMARA A. CROOK
University of Tennessee
Individuals make attributions when they infer causes aboutparticular outcomes.
Several narrative reviews of attributional research have concluded that attributions
matter in the workplace,but note that attribution theory has been underutilized in
organizational research. To examine the predictive power of attributions in organiza-
tional contexts, we present a meta-analysis of existing attribution theory research. Our
findings suggest that attributions have consistently demonstrated effect sizes that are
comparable to more commonly utilized predictor variables of workplace outcomes.
Expanding on these findings,we argue that attributions are an integral part of indi-
viduals’ cognitive processes that are associated with critical organizational outcomes.
We conclude with suggestions to help expand and optimize the contribution of attri-
butional research to understanding and managing organizational outcomes.
In his seminal work on attribution theory, Heider
(1958)characterized people as naive psychologists
with an innate interest in understanding the causes of
successesand failures. Causal explanations,as
Heider asserted, enable individuals to make sense of
their world and control their environments.Re-
searching how people think about causation helps us
to better understand how and why people engage in
both productive and counterproductiveorganiza-
tional behaviors (Luthans & Church, 2002). For exam-
ple, attributionalprocesses can help explain what
causes an employee’s aggression (e.g., Douglas & Mar-
tinko, 2001; Spector, 2011), an applicant’s interview
success (e.g., Ashkanasy, 1989; Silvester, Anderson-
Gough, Anderson, & Mohamed, 2002), a leader’s be-
havior (e.g., Campbell & Swift, 2006; Green & Liden,
1980),and many otherorganizationalphenomena
(see Martinko, Douglas, & Harvey, 2006, for a review).
The topic of attributions emerged with Heider’s
(1958) work and evolved to become a dominant the-
ory in social psychology in subsequent decades, par-
ticularly after the works of Kelley (1967, 1971, 1973)
and Weiner (1985, 1986, 1995, 2004). The breadth of
phenomena to which attribution theory can be ap-
plied, including learned helplessness, depression, re-
ward/punishmentdecisions, and motivation, has
driven its widespread adoption by social psycholo-
gists (Reisenzein & Rudolph, 2008). By contrast, it has
been asserted that attribution theory has been unde-
rutilized and underappreciated in the organizational
sciences (Martinko, Harvey, & Dasborough, 2011).
We thank the entire editorial team for their help and
guidance throughout the process.We also thank the re-
viewers for the developmentalnature of their reviews
and the help they offered.
娀 The Academy of Management Perspectives
2014, Vol. 28, No. 2, 128–146.
http://dx.doi.org/10.5465/amp.2012.0175
128
Copyright of the Academy of Management, all rights reserved. Contents may not be copied, emailed, posted to a listserv, or otherwise transmitted without the copyright holder’s ex
written permission. Users may print, download, or email articles for individual use only.

In support of this argument,Martinko and col-
leagues (2011) compared the two fields and found
nearly six times as many attributional articles pub-
lished in psychology journals as in management
journals.Additionally, as of this writing,a search
of Google Scholar shows more than 7,000 citations
of Kelley’s and Weiner’s works, but only 9% of
those are found in business-related disciplines.
Even accounting for the difference in the number of
journals between the two fields,the magnitude of
this disconnect is surprising,especially consider-
ing the similarity in interpersonal phenomena
studied in both fields.
The recognition of a disconnect between the use
of attribution theory in the organizational sciences
versus the social psychology literature is not new.
As far back as 1995,Martinko (p. 11) commented
that in an examination of organizational behavior
texts “attribution theory as a theoreticalbody of
work was frequently ignored.” Nevertheless,after
viewing the body of research thathad been con-
ducted in the organizational sciences, Weiner
(1995,p. 5) was optimistic with regard to the po-
tential of attribution theory and commented that
“putting attribution theory in the context of organ-
izational behavior should be of great benefit.”
To make the potentialcontributions of attribu-
tion theory to the organizational sciences more sa-
lient, two comprehensive conceptual reviews were
conducted.A review by Martinko and colleagues
(2006) provided an overview of research in
industrial/organizationalpsychology.The review
concluded that the research “unequivocally docu-
ments thatattributions play a significantrole in
behaviors associated with the topics that are central
to [industrial/organizationalpsychology]such as
individual differences,counterproductive behav-
ior, leader/member interactions,impression man-
agement, conflict resolution, training, selection in-
terviewing, and performance appraisal” (p. 174). A
second review by Martinko,Harvey, and Douglas
(2007) arrived at a similar conclusion based on an
examination of attributional studies in the realm of
leadership research. Given this conclusion, the au-
thors noted that they were puzzled as to why attri-
bution theory had not emerged as a major theory of
workplace outcomes.
Both reviews attributed this gap in theory at least
in part to earlier criticisms of attribution theory by
Mitchell (1982) and Lord (1995).Mitchell argued
that attributions accounted for only a small propor-
tion of the variance in causal explanations. In short,
he asserted that attributions might matter, but prob-
ably not very much.Lord (1995) asserted that the
rational information-processing modelsuggested
by attribution theory was unrealistic and that peo-
ple generally rely on more efficient cognitive schema
and implicit assumptions when forming causal per-
ceptions. Martinko and colleagues (2006) countered
these criticisms, asserting that they had been overgen-
eralized and misinterpreted by subsequent research-
ers. Indeed, it has long been recognized that routine
or unimportantoutcomes are less likely to invoke
detailed causal search processes than events that are
unexpected and personally relevant (e.g., Lord, 1995;
Weiner, 1986).
More recent process models,such as that pro-
posed by Douglas, Kiewitz, Martinko, Harvey, Kim,
and Chun (2008), suggest that the degree of cogni-
tive elaboration devoted to attributional analyses of
outcomes varies depending on the nature ofthe
trigger event as well as individual difference factors
(e.g., need for cognition). Thus, empirical studies of
attributions,including those analyzed here,typi-
cally consider attributions in response to trigger
events that are of sufficient personal relevance to
invoke attributional searches. Although it is some-
what speculative to suggest that the underutiliza-
tion of attribution theory in the organizational sci-
ences as compared to social psychology has been
driven primarily by these critiques and their sub-
sequent interpretations by other scholars,it is no-
table thatthe use of attributionalperspectives in
organizational research appeared to taper off after
their publication.
Given the debate about whether attributions mat-
ter, and whether attribution theory has been unde-
rutilized in the organizationalsciences,there are
lingering questions about the efficacy of attribution
theory. Our objective in this study, therefore, is to
shed light on this debate by exploring whether and
to what extent attributions are related to important
workplace outcomes.We do so through a meta-
analysis.Meta-analysis allows researchers to esti-
mate the size of relationships found in a body of
research (Hunter & Schmidt, 2004). If relationships
exist, meta-analysis furnishes an estimate ofthe
magnitudes (i.e.,effectsizes) while ensuring that
the estimates account for important study artifacts
such as sampling and measurement error (Hunter &
Schmidt, 2004).
Applying meta-analysis to key attribution–work-
place outcome relationships appears to be an ap-
propriate and timely way to inform the debate re-
garding the utility and contribution of attribution
theory to understanding the causes and conse-
2014 129Harvey, Madison, Martinko, Russell Crook, and Crook

quences of organizational behaviors. Thus, a goal of
our meta-analytic assessmentof key attributional
relationships is to estimate not only whether attri-
butions matter,but also the extent to which they
matter. Appendix A reports the papers included in
our analyses;Appendix B describes our method-
ological approach in more detail.
If we find that attribution processes do not mat-
ter, we can conclude that researchers have rightly
ignored the role of attributions in the workplace.
However, if we find that attributions do matter, and
matter substantially (i.e.,have strong effects),we
can conclude that researchers have erred in ignor-
ing the role of attributions in the workplace and
hopefully describe new contexts and research ar-
eas—such as macro-level inquiry—that present
new terrain for the theory.
ATTRIBUTIONAL DIMENSIONS AND
WORKPLACE OUTCOMES
Before discussing our meta-analysis, we first re-
view the attributional dimensions that are linked to
workplace outcomes. As Weiner (1986) explained,
attributions generally vary along severalunderly-
ing dimensions. Although several dimensions have
been identified,our search of the extant literature
suggested thatwithin the organizationalsciences
the most commonly studied are the locus of cau-
sality, stability, and controllability dimensions
identified by Weiner (1985).We therefore focus
on studies of these dimensions in the present
meta-analysis.
Locus of Causality
The most commonly studied attributional di-
mension by far is locus of causality, which refers to
whether the perceived cause of an outcome is in-
ternal or external. In the case of attributions made
for one’s own outcomes (i.e.,self-attributions),an
internal attribution occurs when the cause is per-
ceived to reflect some characteristic of the person
such as effort or ability. For example, an employee
who misses a deadline and believes that this out-
come is due to his or her own lack of effort or
ability is making an internal attribution. An exter-
nal attribution for the same outcome might take the
form of blaming coworkers or a supervisor for the
missed deadline. Similarly, when observing others’
outcomes (i.e., social attributions), internal attribu-
tions refer to dispositional or behavioral character-
istics (e.g., effort or ability) of the person being
observed, whereas external attributions refer to sit-
uational factors that are often beyond the observed
individual’s control.
Weiner (1985) observed that the locus dimension
was particularly relevant to the emotional re-
sponses individuals form in response to trigger
events.His widely cited achievementmotivation
model (Weiner, 1985) suggeststhat individuals
form attributions in response to trigger events,
which, as noted above,typically take the form of
outcomes thatare negative,surprising, or unex-
pected (e.g., an unexpectedly poor evaluation).
When personally relevantnegative outcomes are
attributed to internalfactors (e.g.,insufficient ef-
fort), Weiner’s framework indicates that self-fo-
cused emotions such as guilt and shame are likely.
Conversely, when these outcomes are attributed to
external causes (e.g., a biased supervisor), emotions
such as anger and frustration typically follow.
Weiner’s modelalso accommodates positive out-
comes and suggests thatinternal attributions for
these events promote feelings ofpride, whereas
external attributions are associated with gratitude
and other externally focused positive emotions.
Understanding the relationship between attribu-
tions and emotions is important because it is the
affective response to attributions thatis thought
to shape behavioral reactions to trigger events
(Weiner, 1985).For example,research has shown
that the locus of an attribution for negative work-
place outcomes can influence the choice between
passive and aggressive behavioral responses (e.g.,
Douglas & Martinko, 2001).
The locus dimension is also relevant to leader–
member relationships and interactions. Conceptual
and empirical studies indicate that when supervi-
sors and employees conflict in their attributions for
negative outcomes,interpersonalconflict and di-
minished evaluations of relationship quality occur
(e.g., Ashkanasy, 1989, 1995; Green & Mitchell,
1979; Martinko, Moss, Douglas, & Borkowski, 2007).
This effect is particularly likely when the attribu-
tional conflict leads each party to blame the other for
negative outcomes orwhen each claims personal
credit for desirable outcomes.
The locus of an attribution can also influence
reward and punishmentdecisions. As one might
expect, there is evidence that employees are more
likely to be rewarded for high levels of performance
when the supervisor attributes the performance to
internal characteristics,such as the employees’
ability, rather than to external factors (Johnson,
Erez, Kiker, & Motowidlo, 2002). Similarly, punish-
130 MayThe Academy of Management Perspectives

ments are more likely when undesirable outcomes
are attributed to internal employee characteristics
or behaviors,such as insufficient effort (Wood &
Mitchell, 1981).
Stability
The stability dimension of attributions refers to
the perceived variability or permanence of a causal
factor. To illustrate, a person’s intelligence is typi-
cally viewed as a relatively stable factor,whereas
effort level is more variable (Weiner, Frieze, Kukla,
Reed, Rest, & Rosenbaum,1971).Unlike locus of
causality, the stability dimension is rarely studied
separately from other dimensions. More com-
monly, researchers examine the locus and stability
dimensions in tandem. This design is logical in that
the stability of a cause can attenuate or exacerbate
the emotional and behavioral responsesdriven
by the locus of the attribution.More specifically,
perceptions ofcausal stability help shape a per-
son’s expectations for future outcomes,and these
expectations can soften or amplify the emotional
response to a trigger event (Weiner, 1985). For ex-
ample, if an employee attributes a poor evaluation
to a lack of ability (an internal and relatively stable
cause) the individual is likely to experience shame,
given that a lack of ability is likely to cause similar
problems in the future.This shame, in turn, can
promote withdrawal behaviors (Hall, Hladkyi,
Perry, & Ruthig, 2004).If the evaluation is attrib-
uted to insufficient effort (an internal and unstable
cause),the employee is more likely to experience
guilt and a motivation to exert more effort in the
future.
The perceived stability of a cause is also likely to
help shape leader–memberrelationship quality
and reward/punishmentdecisions. As described
above, the negative impact of attributions that tar-
get blame for undesirable outcomes on one’s super-
visor or subordinate is often weaker if the attribu-
tion is unstable rather than stable in nature. This is
because unstable attributions reduce the expecta-
tion that similar outcomes will occur in the future.
This attenuating effect has been shown to reduce
the frequency and severity of punishments for un-
desirable workplace outcomes (Wood & Mitch-
ell, 1981).
Controllability
Of the three attributional dimensions included in
this analysis, controllability has received the small-
est amount of research attention.Controllability
refers to the extent to which an observer perceives
the cause of an outcome to be under someone’s
volition (Weiner, 1985).Factors such as luck and
task difficulty are generally perceived to be uncon-
trollable,whereas effort and,to a much lesser ex-
tent, ability are viewed as controllable factors.
As these examples suggest, there is some overlap
between the controllability dimension and the lo-
cus and stability dimensions.For instance,effort,
which is usually viewed as internal and unstable, is
most often seen as controllable, whereas task diffi-
culty, which is seen as external and stable, is most
often viewed as uncontrollable. As such, controlla-
bility has been linked to a number of the same
affective and leadership outcomes thatthe other
dimensions predict,albeit in a smaller number of
studies. This overlap might explain why controlla-
bility has received less attention than the other two
dimensions. Nonetheless,there was a sufficient
amountof research on controllability as a stand-
alone dimension to warrant inclusion in our
analysis.
A META-ANALYTIC REVIEW OF
ATTRIBUTIONAL RESEARCH IN THE
ORGANIZATIONAL SCIENCES
The reviews and observations discussed atthe
outset of this article argue that attribution theory
has been underutilized by scholars in the organiza-
tional sciences.Although conceptual reviews and
articles have attempted to address these issues, we
believe that a meta-analytic assessment summariz-
ing the predictive power of attributions in organi-
zational research is beneficialin showcasing the
utility of attribution theory.In particular, our as-
sessment is intended to help ease concerns stem-
ming from various interpretationsof Mitchell’s
(1982) aforementioned criticism regarding effect
sizes that may have caused some scholars to ignore
the predictive potential of attribution theory.Ad-
ditionally, a meta-analytic study allows us to pres-
ent an overview of the nomological network of
attribution theory, which can help scholars identify
gaps in the existing empirical literature.
To this end, we meta-analyze relationships be-
tween the attributional dimensions of locus, stabil-
ity, and controllability and four categories of work-
place outcome variables.These include affective
outcomes (e.g., discrete emotions such as anger and
sympathy,attitudes such as job satisfaction,emo-
tional evaluations such as self-esteem),perfor-
2014 131Harvey, Madison, Martinko, Russell Crook, and Crook

mance outcomes (e.g., task performance, effort lev-
els), leader–member relationship quality outcomes
(e.g., leader-memberexchange [LMX]ratings,per-
ceived managerial trustworthiness, conflict between
employees and supervisors), and reward/punishment
decisions (e.g., award allocation decisions, intent to
punish subordinates).As discussed in the previous
section, attributions have been linked to each of these
outcomes in organizational research.Further,these
categories allow a logical categorization of most stud-
ies identified for inclusion in the analysis.Within
each of these outcome categories we examine the
predictive power ofattributions.Although attribu-
tional variables have also been conceptualized as out-
come variables (e.g.,attributions ofresponsibility;
Weiner, 1995), this body of research has not yet ma-
tured to the point where enough studies are available
for a meta-analytic assessment.
RESULTS
In this section we present our meta-analytic re-
sults organized by workplace outcomes within
each of the three described attributionaldimen-
sions. These results are summarized in Table 1.
Locus of Causality
Affect. Across our sample of studies, the locus of
attributions significantly influenced emotional reac-
tions to undesirable 娀 r៮c ⫺⫽ .16 娀 and desirable out-
comes 娀 r៮c ⫽ .13 娀 . These observations suggest that ex-
ternal attributions for personally relevant unfavorable
outcomes (e.g., attributing a missed deadline to a neg-
ligent coworker) were associated with less negative
emotional and attitudinal reactions than were inter-
nal attributions (e.g., attributing a missed deadline to
one’s own lack of effort) for these outcomes.Simi-
larly, internal attributionsfor favorable outcomes
(e.g., attributing a bonus to one’s ability) were associ-
ated with stronger positive affective reactions than
externalattributions (e.g.,attributing a bonus to a
supervisor’sgenerosity).The findings concerning
both negative and positive trigger events are consis-
tent with research on self-serving attributional biases,
which suggests thatpeople often feelbetter when
they attribute blame for problematic outcomes to ex-
ternal factors and take credit for favorable outcomes
(Huff & Schwenk, 1990; Zuckerman, 1979).
Performance. The locus of attributions for unfa-
vorable outcomessignificantly impacted perfor-
mance 娀 r៮c ⫽ .11 娀 , as did the locus of attributions for
TABLE 1
Attribution and Workplace Outcome Results
Attributional relationship N K Corrected effect 娀 r៮c娀
Locus of causality: undesirable affect 1672 13 ⫺.16
Locus of causality: desirable affect 2077 13 .13
Locus of causality: unfavorable performance 2,800 14 .10
Locus of causality: favorable performance 2042 14 .21
Locus of causality: unfavorable LMX 1385 5 ⫺.12
Locus of causality: favorable LMX 923 7 .07 (ns)
Locus of causality: punishment intentions 1413 5 ⫺.15
Locus of causality: reward intentions 1342 7 ⫺.01 (ns)
Stability: undesirable affect 192 4 .13 (ns)
Stability: desirable affect 776 7 ⫺.13
Stability: favorable performance 276 4 ⫺.35
Stability: unfavorable LMX 2090 7 .10
Stability: favorable LMX 267 4 ⫺.17 (ns)
Controllability: undesirable affect 1569 5 ⫺.35
Controllability: desirable affect 98 2 .16 (ns)
Controllability: unfavorable performance 406 2 .03 (ns)
Controllability: unfavorable LMX 1,143 4 ⫺.09
Notes: We ran a meta-analysis for each attributional relationship. High scores for locus of causality and for controllability suggest mor
external and less controllable. High scores for stability suggest more stability (and thus, less ability to change situation). N depicts aggre
number of primary study observations;K is the number of primary studies effects were drawn from.The corrected effects 娀 r៮c娀 were
computed by taking the sample-size weighted effects across the primary studies and then correcting those effects for measurement erro
Significance tests take into account N, K, and variance (Arthur, Bennett, & Huffcutt, 2001). When the confidence interval does not contain
a zero, there is evidence suggesting a significantpositive or negative relationship (Hunter & Schmidt,2004).All relationships are
significant at p ⬍ .05 unless noted (ns).
132 MayThe Academy of Management Perspectives

favorable outcomes 娀 r៮c ⫽ .21 娀 .This suggests that
external attributions for personally relevant unfa-
vorable events were associated with increases in
negative self- and leader-reported performance rat-
ings, whereas internal attributions for positive out-
comes were positively related to performance rat-
ings. The latter finding is consistent with research
indicating that internal attributions for desirable
outcomes,such as attributing a job offer to one’s
own abilities and experience,promote self-confi-
dence and efficacy (Silver, Mitchell, & Gist, 1995)—
which, in turn, can lead to improved performance
levels (Stajkovic & Luthans, 1998). Consistent with
this perspective, the .21 corrected r also suggests that
external attributions for desirable outcomes, such as
attributing a job offer to luck,were detrimental (or,
more accurately,less beneficial)to performance.
More peculiar is the finding that the apparent recip-
rocal of this tendency, externalizing rather than inter-
nalizing blame for negative outcomes, was also detri-
mental to performance.
A possible reconciliation of these findings might
be found in Harvey and Martinko’s (2009) argu-
ment that erroneously attributing undesirable out-
comes to external factors could prevent employees
from accurately evaluating and remedying their
performance weaknesses.The observation that in-
dividuals tend to be more objective when forming
attributions for positive outcomes is also relevant
(Huff & Schwenk,1990).Thus, it may be that the
internal attributions for desirable outcomesare
more accurate than the external attributions for un-
desirable events.In this case we would expect
the self-efficacy gains from internalizing positive
events to provide a legitimate source of confidence
that can fuel performance increases,whereas the
self-efficacy protection provided by external attri-
butions for negative events might attenuate motiva-
tion for performance improvements.
Leader–member relationship valuations. Exter-
nal attributions for unfavorable events (e.g.,em-
ployees attributing a poor raise to a biased super-
visor or to economic factors) were associated with a
significant improvementin leader–memberrela-
tionship quality rating 娀 r៮c ⫺⫽ .12 娀 , indicating a
decrease in negative evaluationsof relationship
quality) across our sample of studies. This is a
somewhat surprising finding, given that the super-
visor is often thought to be a common targetof
employees’external attributions (Harvey & Mar-
tinko, 2009). By contrast, if employees attribute
negative outcomes to externalfactors otherthan
their supervisors (e.g., economic factors), the find-
ing appears more intuitive.This relationship sug-
gests thatthese non-supervisor–targeted external
attributions by subjects may be more common than
we had expected.
Employees’ internal attributions for favorable
events were more strongly associated with im-
provements in leader–member relationship quality娀 r៮c ⫽ .07 娀 than were external attributions. Although
we might have expected externalattributions for
these outcomesto improve relationship quality
more strongly than internal attributions if the
leader was the targetof the externalattributions,
the aforementioned findings concerning negative
outcomes again suggestthat employees consider
several other external factors when forming these
attributions.
Reward/punishment decisions. Attributional stud-
ies on reward and punishment decisions generally
look at leaders’ attributions for subordinate perfor-
mance.These attributions differ from the self-fo-
cused attributions studied in the context of affec-
tive responses,performance,and leader–member
relationships in that they are social attributions
that consider observers’(usually leaders’) attribu-
tions for other people’s (usually subordinates’) out-
comes.In these studies,internal attributions refer
to causal factors that are internal to the employee
being observed (e.g., ability, effort), whereas exter-
nal attributions refer to factors outside the employ-
ee’s control (e.g.,task difficulty, bad luck). Corre-
lations from these studies were coded such that
higher scores denote external attributions. Predict-
ably, results suggested that external attributions for
employees’negative outcomes(generally opera-
tionalized as poor performance)were associated
with lower punishment intentions 娀 r៮c ⫺⫽ .15 娀 . A
significant relationship between locus attributions
for favorable outcomes and reward decisions
was not observed.
Stability
Affect.As noted above, the locus dimension,
rather than the stability dimension, is generally
thought to impact affective outcomes. Not surpris-
ingly then, the results did not indicate that stable
attributions for unfavorable eventswere signifi-
cantly associated with increases ordecreases in
affective outcomes.A significant relationship be-
tween stable attributions for positive events and
affective responses was observed,indicating that
stable attributions were associated with lower lev-
2014 133Harvey, Madison, Martinko, Russell Crook, and Crook

els of unfavorable affective reactions than unstable
attributions 娀 r៮c ⫺⫽ .13 娀 .
Performance.Across the sample of studies,we
observed a significant positive association between
stable attributions for favorable outcomes and per-
formance 娀 r៮c ⫺⫽ .35; note that the sign is negative
because a high score indicates poor performance).
Only one study in our sample investigated the re-
lationship between stable attributions for undesir-
able outcomes and poor performance, so this cate-
gory was not included in the analysis.
Leader–member relationship evaluations.Our
findings concerning leader–memberrelationship
outcomes and stability indicated that stable attri-
butions for unfavorable events were associated
with an increase in reports of poor relationship
quality 娀 r៮c ⫽ .10 娀 . This is logical because if a cause
is deemed to be stable, it follows that the employee
views the leader as incapable or unwilling to rem-
edy the situation.The stability of attributions for
favorable outcomes was unrelated to leader–mem-
ber relationship quality.
Reward/punishmentdecisions.An insufficient
number of published studies on the relationship
between stability attributions and reward/punish-
ment decisions were identified to perform a meta-
analysis on this relationship.
Controllability
Affect.The attribution of negative outcomes to
personally controllable causes was associated with
reduced levels of negative emotions and attitudes娀 r៮c ⫺⫽ .35 娀 . This is consistent with the notion that
controllable causes can frequently be avoided in
the future and are therefore less likely to provoke
negative affective reactions (Aquino,Douglas, &
Martinko, 2004).Although our sample of observa-
tions for this relationship was small, no significant
association between the controllability dimension
and affective outcomes was observed in the context
of positive trigger events.
Performance.Very few studies have investi-
gated the relationship between controllability attri-
butions and performance. Only two studies in our
sample investigated this relationship,both in the
context of negative trigger events, and did not sug-
gest any significant associations.
Leader–member relationship evaluations.Per-
ceived controllability was associated with a slight
decrease in reports of unfavorable leader–member
relationship quality 娀 r៮c ⫺⫽ .09 娀 . Not surprisingly,
this suggeststhat leader–memberrelationships
did not suffer when employees perceived that they
personally controlled the causes of an undesirable
outcome. Conversely, the correlation suggests that
a slight worsening of leader–member relationship
quality occurred when the cause was seen as be-
yond the employees’control. Only one study ex-
amined this relationship in the context of desirable
outcomes, so meta-analysis was not possible.
Reward/punishmentdecisions.An insufficient
number of published studies on the relationship
between controllability attributionsand reward/
punishment decisions were identified to perform a
meta-analysis on this relationship.
DISCUSSION
The purpose of this meta-analysis was to provide
an empirical investigation into the importance and
impact of attributions on organizational outcomes.
We begin this section with a brief comparison of
our findings against those of other predictor vari-
ables that are frequently used in organizational re-
search. We then discuss implications of our analy-
sis for the future of attributional research.
Locus of Causality
Across our sample of studies we found that the
most commonly studied attributionaldimension,
locus of causality,consistently showed a signifi-
cant influence on each of the outcome variables
included in the analysis. The average magnitude of
the effects was .14. Aguinis, Gottfredson, and
Wright (2011) asserted that perhaps the best way to
interpret the magnitude of effect sizes is to put
them into the context of other relationships as-
sessed via meta-analysis. Our review suggests that
this predictive power is on par with that shown by
other predictors of similar outcome variables.
As an illustration, we use the job performance
outcome studied here (locus r៮c ⫽ .16). Cohen-
Charesh and Spector’s (2001) meta-analysis of jus-
tice perceptions indicated average weighted corre-
lation coefficients of .13, .45, and .16 between
distributive, procedural,and interactionaljustice
perceptions, respectively, and work performance in
field studies. In laboratory studies,their analysis
indicated a mean weighted correlation of .05 (non-
significant) between distributive justice percep-
tions and performance and .11 between procedural
justice perceptions and performance. Similarly,
Barrick and Mount (1991) observed corrected cor-
relations of .10, ⫺.04, ⫺.05, .17, and .00 when
134 MayThe Academy of Management Perspectives

examining the impactof Big Five personality di-
mensions of extroversion, emotional stability,
agreeableness,conscientiousness,and openness,
respectively,to experience on performance mea-
sures.In their analysis of the impact of core self-
evaluation components on performance, Judge and
Bono (2001) observed corrected correlations of .26,
.29, .22, and .19 for self-esteem, self-efficacy, locus
of control, and emotional stability, respectively,
and performance. A meta-analysis by Gernster and
Day (1997) showed a corrected correlation coeffi-
cient of .11 between LMX and objective perfor-
mance ratings.
Based on this comparison, the effect sizes of the
locus of causality dimension fall within what ap-
pears to be a range that is acceptable to most schol-
ars, as evidenced by the frequency with which jus-
tice perceptions, core self-evaluation, the Big Five
personality traits, and LMX are studied as predictor
variables in the organizational sciences. A compar-
ison against predictors of the other outcome vari-
ables studied here indicated a similar pattern:
Some predictors show larger effects and some show
smaller effects, but in most cases attributional pre-
dictors demonstrated effectsizes comparable to
those of variables thatare utilized far more fre-
quently than attributionalvariables by organiza-
tional researchers.
Stability
Although the number and scope of studies exam-
ining the impact of the stability dimension were
limited compared to the locus of causality studies,
we observed significant associations between this
dimension and affect, performance,and leader–
member relationship evaluations. The average cor-
rected correlation coefficient of these relationships
was .18. Using the comparisons reported in the
previous section,this effect size again appears to
compare favorably with more frequently studied
predictor variables. It is also worth reiterating that
the locus and stability dimensions are often studied
together,a combination that could explain still
more variance than the two dimensions explain
individually.
Controllability
The relatively small number of studies using the
controllability dimension limited the scope of our
analysis but indicated that it was a relevant predic-
tor of affective outcomes and leader–member eval-
uations.Again, the magnitude of the observed ef-
fects 娀 r៮c ⫺⫽ .35 and ⫺⫽ .07, respectively) appears
comparable to other predictors thatare far more
commonly used by organizational scholars.
THE PATH AHEAD
The underlying theme of our findings, consistent
with the conclusions of earlier conceptual reviews,
is that attribution theory has significant predictive
power that is similar or equal to other theories that
attempt to predict and explain workplace phenom-
ena. Thus attribution theory adds an important
piece of predictive power to the arsenal of frame-
works and constructs available to organizational
researchers.Nevertheless,it appears thatattribu-
tion theory has been underutilized compared to
other theories,suggesting thatthere is ample op-
portunity for scholars to add to our understanding
of workplace behaviorby using an attributional
perspective.
The four categoriesof outcome variables in-
cluded in our analysis have been reasonably well
examined through an attributional lens,but there
are numerous other avenues for new outcomes to
be explored.In fact, as social psychologists have
observed, causal perceptions arguably influence al-
most every aspectof a person’s behavior.Thus,
there are probably very few workplace behaviors
and outcomes that cannot be investigated from an
attributional perspective. Although space consid-
erations limit our ability to discuss all the poten-
tial areas where attribution theory can be ex-
tended, we will highlight those that appear
particularly promising. These potentialareas in-
clude examining additional attributional dimensions,
expanding attribution theory across new topical ar-
eas,investigating temporal effects,and introducing
attribution theory in the macro domain.
Additional Attributional Dimensions
In the areas of organizational science where at-
tribution theory has been utilized,we observed a
narrow focus on the locus of causality and,to a
lesser extent,the stability and controllability di-
mensions. While our analysis suggests that this fo-
cus is not unwarranted, we also note that there are
other dimensions such as intentionality and global-
ity that likely have additional explanatory power.
While these dimensions have received some empir-
ical attention, the number of studies was too small
to include in our analysis.
2014 135Harvey, Madison, Martinko, Russell Crook, and Crook

Intentionality. In an attributional context, intent
describes the extent to which an outcome is attrib-
uted to deliberate as opposed to unintentional ac-
tion. As the name suggests,attributions ofinten-
tionality occur when a perceiver believes that an
outcome was caused by the deliberate and in-
formed actions of another. While behaviors that are
internal and controllable are often assumed to be
intentional, studies that make this assumption
overlook causes such as misinformed decisions
that are well-intentioned buthave unanticipated
consequences.Betancourtand Blair (1992) noted
that this distinction is particularly important in
studies of violence and aggression, but our analysis
found that most studies on these subjects relied on
only the locus of causality, stability, and controlla-
bility factors.
Globality.The perceived globality of a cause
(i.e., the extent to which a causal factor is some-
thing that can impact outcomes across situations
such as personality versus something specific such
as task-specific ability) can also have important
emotional and behavioral implications beyond its
closest corollary, the stability dimension. In an at-
tributional sense,many stable causes,such as in-
telligence level,are also relatively globalin that
they can impact outcomes across a wide range of
situations (e.g.,grades in school, performance at
work, success in business transactions).However,
intelligence can also be perceived as stable and
specific, as in the case of a lack of ability with
regard to a specific task such as a computer appli-
cation.By contrast,unstable factors,such as tem-
porary illness or injury, can also have a global
causal effect during a finite span of time. We there-
fore suggest that globality attributions be assessed
in addition to stability attributions in studies of
cross-situationaloutcomes,such as work–family
conflict and self-esteem (e.g., Seligman, 1998).
Consistency,consensus,and distinctiveness.
Kelley’s (1967) attributional dimensions of consis-
tency, consensus,and distinctiveness mightalso
warrant more attention in organizational research.
Commonly used in the psychology literature, these
dimensions describe how observations help shape
attributions.The consensus dimension of causal
information measures the extentto which an ob-
served behavior is common among multiple indi-
viduals in a given situation. High-consensus behav-
iors (e.g., every employee arriving late to work on a
given day) suggest external/situational casual fac-
tors (e.g., a traffic delay); low-consensus behaviors
promote the formation of internal/dispositional at-
tributions (Martinko & Thomson, 1998).
The consistency dimension looks at variations in
within-person behaviors in the context of a single
situation (e.g.,work). High-consistency behaviors
are those thatan observed individual engages in
frequently (e.g., always arriving to work late); low-
consistency behaviors occur infrequently (e.g., only
rarely arriving to work late). The former suggest
stable causes for the behavior,whereas the latter
suggest unstable causes.
Kelley’s (1967) distinctiveness dimension is also
based on within-person behaviors but concerns
variations between situations.Highly distinct be-
haviors are those that an individual exhibits only in
a particular situation (e.g., late for work but punc-
tual for most other events);nondistinct behaviors
are those thatare observed to be common for an
individual across situations (e.g., late for most en-
gagements).Martinko and Thomson (1998) ex-
plained that observers typically attribute nondis-
tinct behaviors to global causes and distinct
behaviors to specific causes. Thus, Kelley’s dimen-
sions address the issue of how people turn ob-
served variations in behaviors into attributions,
while Weiner’s (1985,2004) dimensions focus on
the affective and behavioral outcomes of those
attributions.
Kelley’s (1967) perspective on the interpretation
of causal information is particularly relevant to the
development of attributions in the workplace given
the role that observation of others’ behaviors plays
in performance evaluations and other management
activities. While organizational research has fo-
cused on understanding the consequences of work-
place attributions, the process by which individu-
als form these attributions has received much less
attention.Although the role of consensus,consis-
tency,and distinctiveness observations in the de-
velopment of attributions may seem fairly clear in
theory,in practice the interpretation of causal in-
formation is subject to the same biases and distor-
tions that affect all perceptions.Managers often
have limited capacity to observe employees (e.g.,
Zalesny & Graen,1987),for example.They may,
therefore,have incomplete information about em-
ployees’behaviors relative to their peers (consen-
sus information),over time (consistency informa-
tion), and in different contexts (distinctiveness
information).
This incomplete information can,in turn, pro-
mote inaccurate attributions and counterproduc-
tive behaviors toward employees.For example,a
136 MayThe Academy of Management Perspectives

manager with many subordinates might observe a
single instance ofemployee lateness and errone-
ously assume it is a common occurrence. The man-
ager might then inaccurately attribute the lateness
to internal, stable, and global factors (e.g., this em-
ployee is an irresponsible person), form a negative
emotionalreaction,and punish the employee ac-
cordingly. Research on factors and tactics that can
improve accuracy in the observationalphase of
attribution development might therefore be in-
strumental in helping managers avoid counter-
productive decisions and behaviors regarding
their employees.
The area of abusive supervision in particular
might benefit from a consideration of the types of
information subordinates use to form perceptions
(i.e., attributions) of abuse.Currently most of the
research on abusive supervision examines the con-
sequences of perceptions of abuse and supervisory
characteristicsthat are associated with subordi-
nates’ perceptions of abuse, while generally ignor-
ing how and why subordinates form these impres-
sions (Martinko, Harvey,Brees,& Mackey, 2013).
We expect that consensus information thatcom-
pares a particular supervisor’s behavior with that of
other supervisors, consistency information that de-
termines if a behavior is displayed consistently or
sporadically,and distinctiveness information that
considers both typical and unusual circumstances
might all play a role in determining whether or not
subordinates see a particular supervisor’s behavior
as abusive.
Reconsideringcontrollability.While we see
value in devoting more attention to the aforemen-
tioned dimensions,our analysis and review also
call into question the use of the controllability di-
mension. There is a fair amount of overlap between
this dimension and the locus of causality dimen-
sion, which may help explain the similarity in out-
comes linked to both types of attributions (e.g.,the
affective and leadership outcomes discussed above).
It also suggests that much of the variance explained
by controllability can be captured by measures of
locus of causality. More specifically, external
causal factors are generally beyond a person’s con-
trol, whereas internal causes are often controllable.
The latter is not always true, however, in that some
causal factors are internal but relatively uncontrol-
lable, such as illness or general intelligence level.
We therefore suggestthat scholars consider the
controllability dimension when designing studies
where such factors are relevant.In studies of per-
formance levels,for example,internal and uncon-
trollable factors such as innate intelligence and in-
ternally controllable factors such as effort are likely
to be relevant. In these cases, assessing both locus
of causality and controllability would be necessary
to optimize prediction.
Perceptions of Justice and Mistreatment
Because attributions are perceptualconstructs,
they are relevant to the formation of other percep-
tions such as justice.More specifically, Martinko
and colleagues (2003) argued that the attributions
employees make aboutthe distribution of work-
place rewards such as promotions shape employ-
ees’ perceptions ofjustice and their beliefs as to
whether or not people who have been promoted
have “paid their dues.” Although their article ex-
plicitly links specific attributions to three different
forms of justice, we are not aware of any empirical
test of the model they presented.Thus the link
between attributions and perceptions of justice is
likely to be a fruitful avenue for exploring how
justice perceptions are formed and the develop-
ment of interventions designed to encourage justice
perceptions.
We also suggest that attributional studies could
also provide insight into an employee’s decision to
engage in behaviors that others might deem unjust
or unethical. Several studies have shown that em-
ployees’ willingness to engage in deviant work-
place behaviors increases when the behavior can be
justified as a response to perceived mistreatment
(Greenberg, 1990; Mars, 1973, 1974). For example,
Greenberg (1990) found that employees admitted to
higher theft levels during a period when their pay
rate was cut by 15%, presumably to adjusttheir
compensation in response to the pay cut.Weiner
(1995) noted that perceptions of responsibility for
this type of undesirable outcome are associated
with external and controllable attributions. Putting
these pieces of information together,we suggest
that factors impacting attributions for negative
workplace outcomes can also impact the likelihood
of a deviant response. While this line of reasoning
might be difficult to test in a field setting, con-
trolled laboratory experiments in which deviant
behaviors can be discretely monitored (see Camp-
bell, Bonacci, Shelton, Exline, & Bushman,2004,
Study 5, for an example) could provide an alterna-
tive means for examining the link between attribu-
tions and deviant or unethical behavior.
Related research has suggested that perceptions
of abusive supervision and entitlement are also in-
2014 137Harvey, Madison, Martinko, Russell Crook, and Crook

fluenced by attributional biases. Martinko, Harvey,
Sikora, and Douglas (2011) found that employees
biased toward external and stable attributions for
negative outcomes rated their supervisors as more
abusive than other employees. More broadly, Har-
vey and Martinko (2009) observed that self-serving
attributionalbiases were associated with psycho-
logical entitlement,a construct that involves in-
flated self-perceptions and unfulfilled reward ex-
pectations that can promote perceptions of unjust
treatment.Taken together,this research suggests
that several workplace perceptions can be more
fully understood by considering them through an
attributional lens.
Ethics
The relative lack of attributionalresearch con-
cerning unethical behavior is surprising. The study
of ethics has long recognized the subjective nature
of ethical perceptions (e.g., Barnett & Karson,
1987),suggesting thata person’s beliefregarding
the cause of an undesirable outcome is a key factor
in shaping ethical judgments.Although ethics re-
searchers recognize that cognitive processes are re-
lated to ethical judgments(e.g., Bommer et al.,
1987; Trevino & Youngblood, 1990), there is almost
no research on the role of attributions in the forma-
tion of ethicality perceptions. Thus it appears that
scholars studying workplace ethics could benefit
by incorporating attributional processes into their
decision models.
Group and Relational Attributions
While there has been some research in the area of
collective attributions—the phenomenon in which
members of groups form similar or identical attri-
butions regarding outcomes that are relevant to the
group (e.g., Islam & Hewstone, 1993)—the great
majority of studies we identified examined attribu-
tions at the individual or dyadic levels of analysis.
Given the consistency with which attributions ap-
pear to impact outcome variables at these levels of
analysis, it seems worthwhile to explore the impact
of group attributions on group-level outcomes such
as performance and cohesion.
This approach seems particularly relevant for or-
ganizationsthat employ group- and team-based
structures in which collective attributions may oc-
cur frequently. Research by Islam and Hewstone
(1993)showed thatgroup members often exhibit
self-serving attributional biases that favor the eth-
nic groups to which they belong.It seems logical
that similar tendencies might bias the attributions
of employees who identify strongly with their de-
partments or other subgroups, potentially creating
intergroup conflict within organizations. More om-
inously, it also suggests that hiring decisions may
be unfairly, if unknowingly, shaped by biased out-
group attributions formed when evaluating candi-
dates (e.g., individuals currently employed at rival
companies, those who graduated from rival
schools). To our knowledge, this is a facet of group/
team research that has not been investigated.
Similarly, it is possible that attributional dis-
agreements could arise within groups.Although
scholars have investigated attributional conflict at
the individual level (e.g., Baron, 1985), it could be
informative to investigate group dynamics that can
promote or impede intragroup attributional conflict
and the extent to which such conflict is desirable
(e.g., in the avoidance of groupthink).
Identifying and learning how to manage group-
level attributional tendencies could therefore have
significant value for practitioners as well as schol-
ars, but research in this area has failed to gain
traction. We suspect this is due, in part, to the
logistical difficulty of measuring and aggregating
the attributions of large numbers of people in re-
sponse to a specific outcome, as was done in Islam
and Hewstone’s (1993) study.We are sympathetic
to this concern, but argue that valuable knowledge
could also be gleaned by studying smaller groups,
such as the workplace teams and departments men-
tioned above. We also argue that attribution theory,
in its present state,is well-suited to the study of
group attributions.Our meta-analysis has shown
consistent effects involving the locus of causality
and stability dimensions on individual-levelout-
comes that either aggregate or are strongly related
to group-level outcomes (e.g., individual LMX and
group-level LMX, individual performance and
group performance). Much of the conceptual logic
and empirical measures used to study these rela-
tionships could likely be applied to the study of
social attributions.
Recent research on relationalattributions also
suggests that there may be new ways to conceptu-
alize well-established attributionaldimensions in
interpersonal contexts. Eberly, Holley, Johnson,
and Mitchell (2011) suggested thatin organiza-
tional contexts the locus of causality dimension
should be thought of in terms of relational attribu-
tions as well as internal and external attributions.
As the name implies, relational attributions cite the
138 MayThe Academy of Management Perspectives

nature of the relationship between two or more
people as the cause of an outcome. As an example,
Eberly et al.noted that employees might attribute
negative workplace outcomes to the fact that they
do not communicate wellwith their supervisors.
Such an attribution focuses on the interpersonal
dynamic between individuals and may provide a
more realistic perspective than assuming that em-
ployees will attribute outcomes primarily to them-
selves (internal) or others (external). This perspec-
tive could be particularly useful in examining the
group and team dynamics common in workplace
settings.
Interactive Effects of Leader–Member
Attribution Styles
Another area that warrants attention is the inter-
action of leaders’and members’attribution styles.
Attribution styles are stable tendencies to form at-
tributions that reflect specific combinations of the
attributional dimensions included in this meta-
analysis. For example, a pessimistic attribution
style denotes a tendency to form internal and stable
attributions for negative outcomes, whereas a hos-
tile attribution style favors external, stable, and
controllable attributionsfor such outcomes. Be-
cause these are consistent perceptual patterns they
can impact a person’s attributions across a range of
outcomes and situations (e.g.,over time,at multi-
ple places of employment).
Building on the seminal article by Green and
Mitchell (1979),a later article by Martinko and
Gardner (1986) proposed that the interactions be-
tween leaders’ and members’ attribution styles
could predispose them to conflict in their attribu-
tions for failure, potentially compromising their
relationship quality.As far as we know only one
study (Martinko et al., 2007) has explored that
proposition and found that the worst leader–mem-
ber relations existed in conditions where leaders
had pessimistic styles and subordinates had opti-
mistic styles.
Conceptually,however, Martinko (2002) noted
that 16 different combinations of social attribution
styles involving the locus and stability dimensions
could exist between a manager and a subordinate.
A number of these combinations could potentially
promote attributional conflict between the two par-
ties. For example, a manager who is biased toward
attributing subordinate success and failure to inter-
nal and stable causes (e.g., low employee ability) is
likely to conflict with subordinates who have ex-
ternal attribution styles and tend to attribute both
success and failures to environmental causes. Lead-
ers with this type of style may be more likely to fire
poorly performing subordinates as opposed to of-
fering training and development and are likely to
blame subordinates for problems thatare beyond
their control.
Kelley’s (1967) aforementioned dimensions also
appear to be particularly relevant in this regard. As
discussed, these dimensions are germane to work-
place settings that require managers to observe the
consensus, consistency, and distinctiveness of em-
ployees’ behaviors relative to other employees.
These observations help shape the locus, stability,
and controllability of attributions (Martinko &
Thomson, 1998), and it seems logical that individ-
uals might show stable tendencies and biases in
forming these attributional observations that influ-
ence their responses to employee behaviors and
outcomes.Thus, a manager mightdemonstrate a
tendency to assume that undesirable employee be-
haviors are high in consistency and low in consen-
sus and distinctiveness.At the individual level,
this could take the form of observing one instance
of employee tardiness and assuming that the behav-
ior is typical of this one employee.At the group
level, a manager might observe similar behavior in
one member of another department and assume it is
the norm in that department.Such a tendency
could contribute to the well-documented actor-ob-
server bias (Jones & Nisbett,1971) that causes ob-
servers (e.g., managers)to discount situational
causes when forming attributions forthe perfor-
mance of others (e.g., employees) at both levels of
analysis.
Many other forms of interpersonalattribution
style conflict and bias might exist,but up to this
point there has been only limited empirical inves-
tigation of how differences in leaders’ and subordi-
nates’ attribution styles interact and affect leader–
member relations and subordinate productivity.
This is an area that is ripe for investigation.
Temporal Effects
The vast majority of the studies in our sample
examined attributions at a single point. Given that
the impact of attributions on emotional and behav-
ioral reactions to trigger events is generally thought
to occur quickly, if not instantaneously (Douglas et
al., 2008),these cross-sectionalstudies undoubt-
edly provide useful information. Nevertheless, we
must also recognize that attributions are frequently
2014 139Harvey, Madison, Martinko, Russell Crook, and Crook