The Structure of Affect: Reconsidering the Relationship Between Negative and Positive Affectivity

Abstract

During the past decade organizational scientists have devoted considerable research attention to the topic of workplace affect. Despite important advances, continued progress depends on a better understanding of the structure of affective experience. The goal of this paper is to review progress to date. In particular, we review evidence pertaining to four constructs that have been widely used to organize research on affect: positive affectivity, negative affectivity, hedonic tone, and affect intensity. We review various structural models pertaining to these four constructs, devoting special attention to integrative frameworks and future research needs.

Full text

Journal of Management 2003 29(6) 831–857
The Structure of Affect: Reconsidering the
Relationship Between Negative and
Positive Affectivity
Russell Cropanzano∗
Department of Management and Policy, Eller College of Business and Public Administration,
The University of Arizona, McClelland Hall, Room 405U, P.O. Box 210108, Tucson, AZ, 85721-0108, USA
Howard M. Weiss
Purdue University, Lafayette, IN, USA
Jeff M.S. Hale
The University of Arizona, Tucson, AZ, USA
Jochen Reb
The University of Arizona, Tucson, AZ, USA
Received 28 February 2003; received in revised form 19 May 2003; accepted 21 May 2003
During the past decade organizational scientists have devoted considerable research attention
to the topic of workplace affect. Despite important advances, continued progress depends on
a better understanding of the structure of affective experience. The goal of this paper is to
review progress to date. In particular, we review evidence pertaining to four constructs that
have been widely used to organize research on affect: positive affectivity, negative affectivity,
hedonic tone, and affect intensity. We review various structural models pertaining to these four
constructs, devoting special attention to integrative frameworks and future research needs.
© 2003 Elsevier Inc. All rights reserved.
In their influential 1995 paper, Ashford and Humphrey (1995) commented that “research
has generally neglected the impact of everyday emotions on organizational life” (p. 97). In
1995 such statements could be made with little fear of contradiction (for historical reviews,
seeAshkanasy,Härtel&Zerbe,2000;Weiss & Brief, 2001; Wright & Doherty,1998).How-
ever, in recent years, research on affect has resurfaced in many traditional topics of manage-
ment inquiry. For example, contemporary research suggests that job satisfaction is partially
∗Corresponding author. Fax: +1-520-621-4171.
E-mail address: Russell@eller.arizona.edu (R. Cropanzano).
0149-2063/$ – see front matter © 2003 Elsevier Inc. All rights reserved.
doi:10.1016/S0149-2063(03)00081-3

832 R. Cropanzano et al./Journal of Management 2003 29(6) 831–857
influenced by affective dispositions (Judge & Larsen, 2001) and states (Weiss, Nicholas &
Daus, 1999), emotional exhaustion lowers job performance (Cropanzano, Rupp & Byrne,
2003;Wright & Cropanzano, 1998), and transformational leadership often involves the
management of subordinate emotions (Ashkanasy & Tse, 2000).
In addition, affect has stimulated new areas of inquiry. We have learned that emotional la-
borcanproducealienation(Ashford&Humphrey,1993;Grandey,2000; Morris & Feldman,
1996), emotional intelligence promotes effective leadership (Goleman, 1998;Goleman,
Boyatzis & McKee, 2002) and affective displays increase sales effectiveness (Tsai, 2001;
Tsai & Huang, 2002). Carried along with this tide have been new research methods, such
as ecological momentary assessment (Beal & Weiss, 2002), qualitative inquiry (Rafaeli &
Sutton, 1990; Sutton, 1991; Sutton & Rafaeli, 1988), and cross-level analysis (George &
Bettenhausen, 1990). These are but modest samples of a young but exciting new literature.
Nonetheless, the problem of how to conceptualize the structure of human affective expe-
rience awaits resolution. Unless we fashion a workable solution the progress seen in the last
few years could be impeded by our lack of a shared taxonomy. The centerpiece of this need
is a theoretical portrait of affective structure. This rendering would portray the intervening
variables that occupy the conceptual space between the antecedents that cause our feelings
and the work outcomes that our feelings are said to engender. We need a way to “visualize”
the affective mediators that drive many work behaviors.
Understanding the Problem
To better appreciate the problem of how to organize human affective experience it might
be helpful to briefly review the two structures that have been commonly used (more detailed
summaries can be found in Cropanzano & Wright, 2001; George, 1992; Judge & Larsen,
2001; Larsen & Diener, 1992). The first structure organizes affective states and traits around
two broad bipolar constructs. These are known by various names, but for now we shall term
them hedonic tone and affect intensity. Hedonic tone is anchored at the high (or positive)
pole by pleasant feelings (e.g., happiness) and at the low (or negative) pole by unpleasant
feelings (e.g., sadness). Affect intensity (AI) refers to the power or force with which a
feeling is experienced. At the high pole affect is a strong and profound experience; at the
low pole it is mild and weak.
The second structure is probably more familiar to organizational scientists. This structure
also includes two constructs, positive affectivity and negative affectivity (George, 1996;
Judge, 1992). Positive affectivity (PA) refers to the tendency to experience intense pleasant
feelings. At the high pole enthusiasm and excitement anchor the dimension. At the low
pole these feelings tend to be absent, but negative affect is not necessarily present. Negative
affectivity (NA) refers to the tendency to experience intense unpleasant feelings. At the high
pole such feelings as anxiety and anger tend to be present. At the low pole these negative
feelings tend to be absent, but positive affect is not necessarily present.
Taken together, these two structures for affective experience yield four constructs—
hedonic tone, affect intensity, positive affectivity, and negative affectivity. As this paper
progresses,weshall have much tosayabouttheresearch generated in support of each model.
However, in broad outline these four constructs provide a map—actually two maps—of the

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 833
affective terrain under exploration by organizational scientists. Herein lies the problem.
Since some researchers use the hedonic tone/affect intensity structure, while others use the
PA/NA structure, then the findings from one research tradition are not directly analyzable
in terms of the other.
Our inability to translate between the two structures poses an impediment to our knowl-
edge,sincethemeaningof any givensetoffindingsislikelyto be obfuscated by the potential
inabilitytomovebetween structures. Wewill demonstrate the magnitude of the predicament
by briefly reviewing evidence pertaining to two important areas of study—creativity and
job performance.
Creativity: When Some Constructs are Missing
As organizations struggle to stay innovative, creativity becomes an important tool for
maintaining competitiveness. Given the changing economy, it should not surprise us that
affective states have been examined as possible mechanisms for promoting creativity. Neg-
ative affect and creativity do not seem to be related in a direct, simple and consistent
fashion (Amabile, 1996;James, Clark & Cropanzano, 1999). On the other hand, some re-
search suggests that positive affect enhances creative problem solving (Isen, 1999; Isen &
Baron, 1991), though others have called even this association into question (e.g., Vosburg
& Kaufmann, 1999). Affect seems to matter, but older models might be insufficiently rich
to account for all of these effects. To address these concerns, George and Zhou (2002) stud-
ied the antecedents of creative performance among 67 workers employed by a helicopter
manufacturer. George and Zhou found unpleasant mood states were positively associated
with creativity when (a) creative endeavors were recognized and rewarded and (b) workers
had clarity regarding their feelings. Pleasant mood states were negatively associated with
creativity under these same conditions.
These findings are practically important for organizations and, more to the present point,
represent an area of inquiry that has made cumulative progress. Despite these strengths, a
consideration of a different structural model could shed new light on the phenomenon. As
we shall see, this is not simply a matter of replacing one construct (say, PA) with another
(such as hedonic tone). Rather, using an integration of both models could allow us to better
specify the theoretical mechanism underlying creative problem solving.
Let us consider how using hedonic tone where only PA and NA were used might change
our thinking. If affect is treated as a unidimensional bipolar construct ranging from very
negative to very positive, then creative performance will change as one slides up or down
on the scale. The key point is that movement toward pleasant moods automatically entails
movementaway(thatis, adecline)in unpleasant moods.Thereverseisalso true. Ontheother
hand, if NA and PA are independent, then an increase in one need not necessitate a decrease
in the other. If we accept that affect is organized as NA and PA, then scholars and managers
needtostay cognizant of two dimensions. Forinstance,whenrecognition/rewardsand mood
clarity are high, then creativity is maximized when NA is high and PA is simultaneously
low. Consequently, achieving an optimal fit of mood to the environment will likely require
two processes—an enhancement mechanism for increasing the appropriate affect and an
accompanying suppression mechanism for decreasing the inappropriate affect. Failure in
either process could foil creative performance.

834 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
It is also instructive to consider the implications of affect intensity for creativity. George
and Zhou’s (2002) study suggests that both PA and NA matter, albeit that their effects vary
with the circumstances. Interestingly, the high poles on each of these dimensions denote
high levels of affective intensity (cf. Thayer, 1989; Wright & Staw, 1999). This might
suggest that only powerful feeling states engender more creative problem solving. Milder
states could have weaker effects. A failure to consider affective intensity, therefore, might
have produced some of the inconsistent findings mentioned earlier.
Job Performance: When NA and PA Don’t Predict
When compared to creativity research, a somewhat different situation emerges from
findingson job performance. Generallyspeaking,research suggests that abipolarwell-being
dimension consistently predicts job performance (Staw & Barsade, 1993; Wright & Bonett,
1997a; Wright & Cropanzano, 2000) and this is even true in longitudinal (Cropanzano &
Wright, 1999;Wright, Bonett & Sweeney, 1993;Wright & Staw, 1999) and experimental
(Staw, Sutton & Pelled, 1994) studies. However, attempts to predict job performance from
measuresof PA and NAhavehad conspicuously less luck(Cropanzano, James & Konovsky,
1993;Wright, Cropanzano, Denney & Moline, 2002) and this seems true regardless of
whether these constructs are measured as states or traits (Wright & Staw, 1999, Study 2).
While it is useful to identify another predictor of job performance, it is also important
that empirical findings inform our theoretical understanding. Assuming the aforementioned
resultswithstandthe test of replication, what can be made of the observation that one bipolar
affective construct (i.e., hedonic tone or psychological well-being) predicts performance
more consistently than do NA and PA? Based on the creativity literature, it would seem
imprudent to dismiss the PA/NA structure as “incorrect” (a point we will emphasize at
length later in this paper). However, until we know more about the relationship between the
two structures, we are in a weak position for answering this important question.
Overview of the Present Paper
In the pages that follow we offer a “status report” on the structure of affective experience.
We shall review different structures that have been proposed for affect. In so doing, we
will devote considerable time to integrative research, as well as to future research needs.
To accomplish these goals, the present paper will have three principle sections. In the first,
we will discuss the basic (and at times confusing) terminology used in this literature. In
particular,we will distinguish betweentworelated constructs—moods and emotions. While
each of these involves affective experiences, there also have some important differences.
After this overview, we will then discuss the basic challenges facing scholars. In this portion
of the paper we will describe two factor structures for affective experience—an unrotated
solution consisting of hedonic tone and affect intensity and a rotated solution consisting
of positive affectivity and negative affectivity. As we shall review in detail, there is solid
evidence in favor of both of these models. In the third section of the paper, we will discuss
three integrative models for these different structures: (1) each factor solution holds for
certain contexts, (2) the two structures are hierarchically related, and (3) the two structures

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 835
are parts of a common circumplex. Finally, we close the paper with a discussion of future
research directions on the nature of affective experience in organizations.
Some Terminology: Distinguishing Moods and Discrete Emotions
The basic and applied research on affect is vast. As a consequence, discussions of this re-
search that have not begun with a clarification of constructs have sometimes ended in confu-
sion. For this reason, we think it useful to provide a brief overview of how we use such terms
as mood and emotion. Weiss (2002) has suggested that the well known Linnaean Classifi-
cation system can provide a useful metaphor for developing a taxonomic framework. In this
framework, affective states are viewed as a family of constructs. Subsumed within this fam-
ily name are the various types of affect constructs, including moods and discrete emotions.
While these “genera” have distinguishing features that we will explore in a moment, they
share certain common features of the family. First, they can all be viewed as states and
describe transient psychological experiences. Second, they share a common subjective (or
feeling) component. Third, all of these states appear to be accompanied by physiological
reactions, although there is disagreement about the precise elements of the physiology of
each state. Finally, these states all have an evaluative component to them. People prefer
some of the states to others.
Next, let’s distinguish between moods and emotions. As affective states, moods and
emotions are closely related (see Weiss, 2002 for a longer discussion). Often moods and
emotions are distinguished by their duration and and/or intensity, with moods being con-
sidered milder affective states of longer duration. However, emotions can last quite a long
time as people often, and dysfunctionally, keep their emotional states active by rumination.
Moods, on the other hand, can be quite short lived.
Rather than intensity or duration, most contemporary affect researchers would argue that
“diffuseness” is the central distinction between moods and emotions. Emotions have an
object of interest or a defining event that is part of the subjective experience itself. This
is to say, emotions are always about something or someone. You’re angry at your boss.
You’re proud of your performance. You’re ashamed of your behavior. Moods, on the other
hand, lack such an object or defining event. In this narrow sense, then, moods are affective
experiences disconnected from their proximate causes.
This diffuseness may explain a defining hallmark of mood research that is not typically
present in the study of emotions. Those who explore moods have tended to search for a
few underlying dimensions to describe affective experience. Usually, the results of these
searchers boil down to one of two structures: hedonic tone/affect intensity or PA/NA. While
emotion researchers recognize the need for parsimony, they have focused less on founda-
tional dimensions are more on the delineation of “basic emotions.” The distinct structural
approaches taken within mood research, as opposed to emotion research, make sense when
one recognizes that a key experiential element in emotions is that they are directed toward
some object. Description in terms of underlying dimensions could strip the emotion of
this object. Therefore, such dimensions might seem less useful for emotional description.
This problem does not exist for moods and therefore our subsequent discussions of affect
structure are primarily discussions from the mood literature.

836 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
Unrotated and Rotated Structures for Affect
Generally speaking, scholars have used two types of dimensional analyses to uncover
the structure of affect (Feldman, 1995a). Probably the most common approach is factor
analysis. Usually, the raw data that goes into these analyses are self-ratings of mood. An-
other popular approach involves the use of multi-dimensional scaling techniques. In this
method participants are given a set of mood-relevant adjectives (e.g., “hostile,” “jubilant”)
and asked to rate their similarity. The resulting distances can be taken as an indication of
semantic meaning. Though these two methods yield generally similar results, there are a
few noteworthy points of departure. Later, we will discuss these differences in more detail.
For now, though, let us emphasize the common findings.
The Unrotated Solution: Hedonic Tone and Affect Intensity
In both unrotated factor analyses of self-reported mood, as well as multi-dimensional
scaling of similarity ratings, the first dimension to appear corresponds to hedonic tone.
It has also been termed pleasantness–unpleasantness (Larsen & Diener, 1992; Weiss &
Cropanzano, 1996; Wundt, 1897) and mood valence (Russell & Carroll, 1999a). “Happy”
and “pleased” would be good anchors for the positive end of this axis, while “sad” and
“despondent” would be suitable for the negative side. This hedonic tone factor is similar
to psychological or subjective well-being (PWB; Cropanzano & Wright, 2001). However,
strictly speaking the two are not the same. Hedonic tone refers narrowly to one’s typical
affect,while PWBissometime viewedasamore general constructsubsumingboth cognitive
and affective dimensions (Campbell, Converse & Rodgers, 1976;Diener, Suh, Lucas &
Smith, 1999). Given these considerations, it seems best to treat hedonic tone as a specific
dimension of PWB. In particular, it refers to one’s affect or emotional well-being (Diener
& Larsen, 1993).
After the hedonic tone or pleasantness factor is extracted, the unrotated solution usually
providesa seconddimension.This secondaxishasgenerally beenunderstoodas the intensity
with which one experiences both negative and positive affective states (Wessman & Ricks,
1966). People with high scores on this dimension might describe themselves as “restless”
and “changeable.” Low scorers might use such terms as “peaceful” and “controlled” (Weiss
&Cropanzano, 1996). Thissecondfactor typicallyaccountsfor only abouthalfas muchvari-
ance as the first factor. Furthermore, the engagement/activation axis is somewhat stronger
for studies that use similarity ratings and somewhat weaker for studies that use self-ratings
of mood, though this can vary among individuals (Feldman, 1995b).
This second axis is known by a variety of names, including excitement–calm (Wundt,
1897),affectintensity (AI; Larsen&Diener,1987;Simonsson-Sarnecki, Lundh&Törestad,
2000), activation (Larsen & Diener, 1992; Russell & Feldman Barrett, 1999), engagement
(Watson, Wiese, Vaidya & Tellegen, 1999;Watson & Tellegen, 1985), and arousal (Faith
& Thayer, 2001; Feldman, 1995a, 1995b). When using the designation “arousal,” it should
be understood to indicate “perceived” and not actual physical arousal. Generally speaking,
peoplehaveat best mixed success recognizing their physical symptoms (Pennebaker,1982).
As a consequence, the relationship between perceived and actual arousal is not a simple
one (Frijda, Ortony, Sonnemans & Clore, 1992).

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 837
Barring some caveats we will discuss later, scholars have located the hedonic tone and (to
a lesser extent) the affect intensity dimensions through the use of factor analysis (Russell &
Carroll, 1999a, 1999b). While organizational scientists recognize the importance of these
two dimensions (e.g., Wright & Doherty, 1998; Weiss et al., 1999), it is the rotated factors
that seem to have generated the most research within the management sciences (for relevant
reviews see Brief & Weiss, 2002; George, 1992, 1996; Judge, 1992; Judge & Larsen, 2001).
For this reason we shall now turn our attention to the rotated solution.
The Rotated Solution: Positive Affectivity and Negative Affectivity
When scholars rotated mood ratings by 45◦they discovered that negative and positive
emotions tended to divide into separate factors (for seminal studies, see Bradburn, 1969;
Bradburn & Caplovitz, 1965;Warr, Barter & Brownbridge, 1983;Watson, 1988a, 1988b;
Zevon & Tellegen, 1982). For the most part, this finding holds true for both mood states and
mood traits (e.g., Watson, 2000; Watson & Clark, 1984; Watson & Tellegen, 1985). The
cluster of pleasant emotions has been dubbed positive affectivity (PA) while the cluster of
negativeemotionshasbeen dubbed negativeaffectivity(NA). ThePA/NAfactorsolutionhas
been obtained in both self and peer ratings and among Americans and Japanese respondents
(Watson, Clark & Tellegen, 1984). In fact, the PA/NA factor structure has been widely
replicated (e.g., Brief & Roberson, 1989;Burke, Brief, George, Roberson & Webster, 1989;
Watson, Clark & Tellegen, 1988).
Those who are high on PA are likely to report experiencing positive emotions that in-
volvehigh levels of activationor engagement. For instance, high PAsare apt to endorsesuch
items as “enthusiastic” and “excited.” Things get interesting, and perhaps a bit counterin-
tuitive, when one considers the negative pole. Those who are low on PA do not necessarily
experience negative emotion. Rather, their moods are characterized by feeling states that
denote both the absence of positive emotions as well as a dearth of energy. An item such
as “drowsy,” “dull,” or “sluggish” could fit here (Judge, 1992; Watson & Clark, 1992;
Watson & Tellegen, 1985; Wright & Staw, 1999). NA works in a parallel fashion. Those
who are high on NA are likely to report experiencing emotions that are unpleasant and
involve high activation or engagement. For instance, high NAs are apt to endorse such
terms as “irritable,” “angry,” or “hostile.” However, low NAs are not especially likely to
feel positive emotions. Rather, the defining attribute of low NA is a tendency to endorse
low-energyfeeling states. These connotethe absence of negativeemotion.Forexample,low
NA might indicate a tendency to be “calm,” “placid,” or “relaxed” (Brief & Weiss, 2002;
George, 1992, 1996). It is important to recognize, then, that positive and negative affect are
distinct, though sometimes correlated (Watson & Tellegen, 1999).
Some Thoughts on the Two Structures
What could be more different than these two structures of mood? One incorporates
positive and negative emotions onto a single axis (hedonic tone), the other separates them
into NA and PA. One model treats affect intensity as an independent dimension, the other
separates intensity into its positive and negative components. At first glance, there seems
to be little common ground, and our task would be easier if we could prove one structure

838 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
“right” and the other “wrong.” Our argument, however, is not with the structures but with
the data. The available evidence suggests that both models have merit.
To get a sense of this, let’s consider a study by Diener and Iran-Nejad (1986, Study 1).
These researchers had 72 undergraduate participants read two vignettes—one that was
designed to produce positive mood (a rescue) and another designed to produce negative
mood (a murder). Notice here that each subject experienced both the positive and also the
negative mood induction. After reading each story, individuals rated their mood. When the
two sets of mood ratings were aggregated across the two vignettes, Diener and Iran-Nejad
found evidence for an affect intensity dimension. Consistent with evidence we reviewed
earlier (Larsen, 1987;Larsen, Diener & Cropanzano, 1987), positive and negative mood
exhibited appreciable negative associations. This suggests that people who feel unhappy at
the sight of a murder will be especially likely to feel happy at the sight of a rescue. Put
differently, people who respond profoundly to one type of affective event are apt to do the
same with another. Thus, when the data are aggregated across the two types of events, there
is solid support for an AI dimension.
However, Diener and Iran-Nejad’s (1986) findings are also consistent with the separation
ofPAandNA.When each condition was considered separately,positive and negativemoods
were virtually uncorrelated. That is, the murder affected NA but had less impact on PA.
Conversely, the rescue had a strong impact on PA but less on NA. The evidence for this
effect is now quite strong. More generally, those high in NA are especially responsive to
negativemood inductions, while those high in PAare especially responsive to positivemood
inductions.
In the final analysis, we cannot easily dismiss either framework. We are left with two
structures and four mood-relevant dimensions. There is an unrotated solution characterized
by hedonic tone and intensity, and a competing rotated solution characterized by positive
and negative affectivity. Both models work in the general sense that (a) it is possible to
obtain the predicted factor structure, (b) there is theoretical evidence consistent with each,
and (c) each framework yields testable propositions. Nevertheless, the proponents of both
models view them only as approximations of a more complex reality (Green, Goldman &
Salovey, 1993;Green, Salovey & Truax, 1999;Tellegen, Watson & Clark, 1999a).
Where Do We Go From Here?: Three Approaches to Integration
Our review so far places us in an interesting quandary. Scholars in each camp (if “camp”
is the correct term) can muster considerable support for their preferred factor solution.
Consequently, the reader is left with a wealth of evidence in favor of both NA/PA as well as
hedonic tone/affect intensity. The evidence has not yet allowed us to falsify either structure.
Given this situation, research has moved away from testing the two factor solutions against
each other and toward some theoretical synthesis that accommodates the existence of both
frameworks.
Research integrating these seemingly competing structures is ongoing and for the most
part recent. Nevertheless, it is at least apparent that no single means of synthesizing NA/PA
with hedonic tone/affect intensity has emerged. Instead, three possibilities have been prof-
fered for the distinct structures: (1) they appear in different research contexts (Zautra, Potter

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 839
& Reich, 1997), (2) they can be integrated into a hierarchical structure (Diener et al., 1995),
and (3) they can be integrated into a circumplex structure (Larsen & Diener, 1992). Most
of the remainder of this report will be spent reviewing each of these possibilities. We shall
begin with context effects, proceed through the hierarchical model, and conclude with a
detailed discussion of the affect circumplex.
The First Possibility: Context Affects the Structure of Affective Experience
One very simple way to resolve the controversy is simply to declare that one model holds
in some situations whereas the other model holds in others. In effect, the situational context
serves as a moderator of affect structure. Though one could posit an almost endless array
of contextual moderators, in practice four have received the most attention (for reviews see
Diener,1994; Zautra et al., 1997). These include: (a) the temporal frame under investigation,
(b) the intensity of the mood experience, (c) the response format, and (d) the possibility of
measurement error in the mood scales.
Context effect #1: temporal frame. Diener (1994) suggested one straightforward pos-
sibility. Diener argued that while PA and NA seem to be roughly orthogonal over time, it
is unlikely that individuals can feel positive and negative emotion in the same instance. In
other words, over a short time either positive or negative feelings dominate, but over the
long run the two classes of mood states tend to disentangle (Russell & Carroll, 1999a).
Diener and Emmons (1985) tested this possibility in several studies. In one study, for
instance, participants were required to rate their positive and negative feelings for two
time frames: over “the past year” and over “the past month.” Positive and negative affect
exhibitedonly a lowcorrelation when ratedoverthe pastyear(r=−.18).Thisisthefamiliar
and often observed NA/PA factor structure discussed by Burke et al. (1989),Watson and
Clark (1984, 1992),Zevon and Tellegen (1982),George (1992, 1996), and others. Different
results came when feelings were rated over the past month. When this shorter temporal
frame was used there was a much higher association (r=−.63). These findings supported
Diener and Emmons contention that immediate mood is experienced somewhat differently
than long-term mood. The shorter, one month format produced a structure that roughly
correspondstotheone-factorhedonictoneor emotional well-being dimension (as described
by Myers, 1992; Russell, 1978, 1979, 1980). In subsequent studies Diener and Emmons
(1985) obtained even more evidence consistent with their thesis that time moderates the
structure of affect (see also Diener & Iran-Nejad, 1986;Staats, Partlo & Adam, 1989).
These findings have not gone unchallenged (cf. Watson & Tellegen, 1985). In one investi-
gation,Watson(1988a, Study 1) compared one popular scale for the measurementofPA and
NA, the Positive and Negative Affect Schedule (PANAS), to the mood inventories used by
Diener and Emmons (1985). He did so over six time intervals—“moment,” “today,” “past
few days,” “past few weeks,” “year,” and “general.” Watson made two critical observations.
First, the PANAS produced generally lower PA/NA associations than did the scales utilized
by Diener and Emmons. Second, and more relevant here, the time interval had at best a
small effect of the correlation between the measures of negative and positive mood. Based
on these findings, Watson (p. 132) maintained “that NA and PA are largely independent of
one another, even when subjects rate how they are feeling at a given moment.” Other work

840 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
has reached similar conclusions (Larsen, McGraw & Cacioppo, 2001;Mayer & Gaschke,
1988;Watson et al., 1988). These findings call into question the possibility that the NA/PA
factor structure can be completely eliminated by measuring the constructs close in time.
Leaving aside these conflicting results, it is interesting to speculate as to why the factor
structure of affective experience would be impacted by temporal frame in the first place. A
cognitive explanation was recently offered by Robinson and Clore (2002). These authors
have introduced a theory of affective memory and self reports that may shed light on
the temporal issues of affect structure. They argue that when individuals report affective
experiencesin the neartermtheirreports are based uponepisodicmemory. Episodic memory
stores specific affective states and experiences. On the other hand, when reporting on their
more distal experiences or their aggregated “general” affective states their reports are based
uponsemanticmemory. That is, these reports are based notuponactualexperiencesbut upon
“beliefs about affective experiences.” Since research on affect structure is driven by self
reports it is possible that temporal differences in structure may reflect differences between
actual experiences and beliefs about such experiences.
Context effect #2: intensity of mood experience. A related possibility is that strong
positive moods eliminate negative moods, while strong negative moods eliminate their
positive counterparts. In other words, positive and negative moods are most negatively
related in a situation that tends to provoke strong feelings. Evidence consistent with this
possibilityis reported by DienerandEmmons (1985, Study 4).Theyfound that intensemood
terms were more negatively related to feelings of the opposite mood than were mood terms
labeled as less intense. Similar findings were reported by Diener and Iran-Nejad (1986).
These authors found that positive and negative moods were unlikely to occur together
when intensity was high. However, at moderate and lower levels of intensity, the two mood
conditions were less likely to be (negatively) correlated.
A related line of evidence comes from Baron’s (1976) incompatible response theory.
Baron maintains that when two opposing feelings come together the stronger feeling state
willsupplantandeliminate the weaker.In one experimental test of these ideas, Baron (1984)
hadresearch participants negotiatewith an experimental accomplice. For our purposes here,
the most important conditions were those where subjects received a gift or viewed three
humorous cartoons. Relative to a control group, individuals in these latter two conditions
reported better moods, more favorable impressions of the accomplice, and a greater will-
ingness to use constructive conflict resolution procedures.
Context effect #3: response format. A third possibility is that each factor structure is
an artifact of the type of scale used to measure affect (Brenner, 1975). Warr et al. (1983)
directly tested this idea. Warr and his colleagues compared a count of positive experiences
to a percentage measure. That is, respondents indicated the proportion of time that they
experienced positive and negative mood states. Generally speaking, the event count yielded
separate factors for positive and negative events, while the proportion measures were highly
and negatively correlated. These findings were replicated by Watson (1988, Study 3).
The impact of response scales on the structure of mood might have a substantive interpre-
tation. Recall our earlier comments suggesting that mood states might be more negatively
associated when they are particularly strong. In view of these findings, it seems likely that

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 841
Warr et al.’s (1983) proportion measures could have functioned as indices of intensity. If
this were so, then it should not surprise us that the association between positive and negative
mood changes. Different formats could be assessing somewhat different constructs.
Context effect #4: measurement error. One of the more controversial theories of late is
the argument that the independence of PA and NA partially results from measurement error
(Russell & Carroll, 1999a). It is well known that errors in measurement can attenuate the
correlation among otherwise related constructs. For instance, Lorr, Quing She and Youniss
(1989) found that controlling for response acquiescence increased the correlation between
positiveand negativeactivity.Ifthe latent variablecorrelationbetweennegativeand positive
affectivity is high enough, then one could argue that they are not separate factors at all.
Thus, the removal of measurement error could lead to one bipolar construct corresponding
to hedonic tone.
Green et al. (1993) explored just this possibility. These scholars used four different
scale formats to measure mood. The relationship among these various formats allowed
the researchers to estimate the magnitude of measurement errors. When these were taken
into account, the relationship between positive and negative mood jumped to −.58. In a
follow-up study, Green et al. (1999) obtained similar findings, suggesting that hedonic tone
might provide a simpler and more parsimonious structure for affect.
Tellegen et al. (1999a) took issue with these findings. They estimated response errors
though a different technique. Tellegen and his colleagues used three acquiescence scales.
The items in each scale formed two pairs. Each pair consisted of related, though opposite
items. For instance, in one example, “rested” and “tired” were paired with “friendly” and
“grouchy” (p. 299). When controlling for measurement error in this fashion, Tellegen et al.
found that positive and negative affectivity formed separate factors, though the factors were
correlated at −.43. Tellegen and his colleagues (1999b) acknowledged that NA and PA are
not “strictly orthogonal” (p. 307), though they did not share Green and Salovey’s (1999)
opinion that mood may form “a quasi-unidimensional structure” (p. 307).
It is noteworthy that most researchers seem to agree that NA and PA are correlated,
so long as measurement error is taken into account (Green et al., 1999; McConville &
Cooper, 1992). However, there is still some debate regarding whether NA and PA should
be collapsed into a single unidimensional bipolar factor (like hedonic tone), or whether
they should retain their two-dimensional form. In the final analysis, this debate over mood
structure rests on one more question: How high a correlation is necessary for us to conclude
that NA and PA comprise a single construct? For a variety of technical reasons—including
suchconsiderations as item selection, response format, and so on—theobservedassociation
betweenNA andPA could easilybeaslow as−.40and still bemeaningful,even ifweassume
thatnegativeand positiveaffectare opposite ends of the same continuum (Russell & Carroll,
1999a). The upshot of this debate is not simple and we shall return to it later.
Summary of the four context effects. It is clear from the research that context effects do
indeed impact the relationship between NA and PA (Zautra et al., 1997). When emotions
are intense, measured as a percentage of time experienced, or when measurement error is
controlled, the correlation between these two dimensions is substantially higher than when
these conditions are not met. It is less clear whether temporal framing has an effect, with

842 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
some arguing that it does and others expressing skepticism (see for instance, Diener, 1984
vs. Watson, 1988a). Regardless, it does not seem to us that measurement issues can address
themostfundamental question: Can NAandPA bereasonablytreatedasdistinct, though not
fully independent, dimensions? Though bruised by these analyses the NA and PA structure
remains viable (Diener et al., 1995; Watson & Clark, 1997; Russell & Carroll, 1999a).
To learn more we must now turn our attention toward the two remaining possibilities.
These approaches attempt to combine the best of the PA/NA and hedonic tone/intensity
structures. Unlike research on contextual effects, these forthcoming models maintain that
the seemingly distinct factor structures can be operational at the same time, with the same
items, and even when measurement error is controlled. To illustrate these points, we will
begin with the hierarchical model.
The Second Possibility: The Hierarchical Model
In science, it often happens that one debate leads to a serendipitous discovery in another.
This may have occurred with respect to the structure of mood. Earlier in this chapter, we
reviewed the debate between Tellegen et al. (1999a, 1999b) and Green and Salovey (1999).
To some degree, this discussion hinged on the size of the PA/NA correlation. Green et al.
(1993) found a stronger relationship of −.58. From this they argued for the tenability of a
bipolarlatentvariable.Tellegenand his colleagues uncoveredaweaker relationship of −.42,
thereby supporting their contention that NA and PA are best treated as distinct constructs.
While associations of this size do not necessarily argue for the separation of these two
affective dimensions (Russell & Carroll, 1999a), in the course of framing their arguments
Tellegen and associates may have presented a more powerful resolution to this debate (see
also Watson & Tellegen, 1999).
In formulating their rebuttal Tellegen et al. (1999b) assembled 29 affect items. When
these were subjected to a first-order factor analysis, nine factors emerged (see their Table 1,
p. 308). Tellegen and his colleagues do not label them. However, the clusters are recogniz-
able to emotion researchers. For instance, the second factor is composed of the adjectives
“delighted,”“excited,”“happy,”and“joyful,”whilethesixthfactoris composed of “angry,”
“contempt,” and “distaste.” In other words, the nine first-order factors roughly overlap with
Shaver, Schwartz, Kirson and O’Conner’s (1987) six dimensions of emotions, developed
by cluster analyses of word meanings.
Tellegen and his collaborators subjected these nine dimensions to a second-order factor
analyses. Two factors were reported. These were labeled positive activation (second-order
PA) and negative activation (second-order NA). Though many of the 29 items cross-loaded,
there were some interpretable patterns. The surprise-related items cross-loaded on both
second-order factors. The fear, shame, and sadness-related items loaded on second-order
NA. The happiness and mental alertness items tend to load on second-order PA. The reader
should recognize that these were the NA and PA dimensions we discussed above. As indi-
cated,they werecorrelatedat r=−.42. This relationshipwasdue tothecross-loadingitems.
As a final step, Tellegen et al. (1999a, 1999b) subjected their positive activation and
negative activation factors to a third-order factor analysis. At this time, a single hedonic
tone dimension emerged. This dimension was interpreted as “happiness vs. unhappiness.”
Thesedataholdoutanintriguing possibility.That is, it could be that basic emotions, NA/PA,

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 843
and psychological well-being are all hierarchically related. Emotions cluster into two mood
dimensions,andthesedimensionsclusterinto overallhappiness (Watson& Tellegen,1999).
Earlier we had made the distinction between moods and discrete emotions. These findings
allowus to elaborateonthis distinction for the purpose of identifying simplifying structures.
Aswe said, emotions comprise discrete affectivestates withthe object of the state being part
of the experience. When we say that we are, say, afraid, it implies being afraid of someone
or something. Moods are affective states divorced from their antecedents. The subjective
affective experience contains no object or event. Hence, as one moves up the hierarchy the
object of emotion is lost. What remains is the undirected affective experience. This may be
either positive or negative. In this way if not in others, Tellegen et al. (1999b) may move us
toward integration of divergent viewpoints about structure.
Of course, the fact that emotional experiences can cluster into general dimensions should
not be taken to mean that these dimensions are identical to the emotional experiences.
Instead positive and negative affect may simply organize those experiences in ways that
have taxonomic utility. The experience of emotions is still in discrete terms and includes a
target for one’s feelings. Thus, the choice of level of organization to study depends entirely
upon the scientific purposes of the investigator.
A similar investigation was conducted by Diener, Smith and Fujita (1995). These authors
took psychometric error into account through the use of different measurement formats.
They also had participants respond to five four-item scales corresponding to six different
emotions: love, joy, fear, anger, shame, and sadness. When these six emotion scales were
factor analyzed, love and joy loaded on a PA factor. Shame, fear, sadness, and anger loaded
on a NA factor. The corrected correlation between PA and NA was −.44, almost identical
to the r=−.42 obtained by Tellegen et al. (1999a). Also in support of Tellegen and his
colleagues, latent variable modeling showed PA and NA to be separate dimensions. Though
NA and PA are correlated, Diener and his colleagues do not view them as the same construct
(for additional development of these ideas, see Diener, 1994).
Unlike Tellegen et al. (1999b),Diener and his colleagues (1995) did not investigate
whether NA and PA clustered at a still higher level of abstraction, though this seems likely
inviewof Tellegenetal.’sfindingsandinviewofthe−.44relationship.Ifweconsiderthese
two structures together, we obtain the three-level hierarchy depicted in Figure 1. At the base
of the hierarchy are Diener et al.’s six emotions: love, joy, shame, fear, sadness, and anger.
At the second level, love and joy cluster into a positive affectivity factor, while the others
clusterintonegativeaffectivity.Finally,consistent with Watsonand Tellegen(1999), PAand
NA cluster into a broad dimension of emotional well-being. We strongly emphasize that our
analysis is largely speculative. Much more research is needed to validate this hierarchical
model. Nevertheless, the evidence to date appears both promising and supportive.
The Third Possibility: The Affective Circumplex
Origins of the affective circumplex. A circumplex model is a structure in which the key
constructs can be organized into a rough ring (for general discussions see Fabrigar, Visser
& Browne, 1997;Wiggins, 1980;Wiggins, Phillips & Trapnell, 1989;Wiggins, Steiger
& Gaelick, 1981;Yik, Russell & Feldman Barrett, 1999). Like an expectant parent, the
circumplex was present at the very birth of scientific psychology. Wundt (1897) organized

844 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
Figure 1. The three-level hierarchy of affect. Adapted from Diener et al. (1995).

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 845
affect terms into a rough sphere around three dimensions: tension release, pleasant–un-
pleasantness, excitement–calm (Wertheimer, 1987). The use of tension release as an or-
ganizing dimension has waned over the years, but interest in pleasant–unpleasantness and
excitement–calm has endured.
Even in these contemporary times, Feldman (1995a: 806) remarked that “many re-
searchers agree that the structure of mood is well represented as a circumplex.” Remington,
Fabrigar and Visser (2000: 286) contend that “The circumplex model of affect has been
among the most widely studied representations of affect.” In making these statements,
these authors are not referring to just any circumplex model. Rather, they are indicating a
particular type of circumplex that corresponds to the one implied by Wundt and others. As
we have already seen, this mood circumplex has two dimensions (or, as we shall soon con-
sider, a rotational equivalent): hedonic tone and intensity/activation (for additional reviews
see Cropanzano & Wright, 2001; Judge & Larsen, 2001; Larsen & Diener, 1992; Larsen
et al., 2002).
The basic model. An idealized version of the mood circumplex is presented in Figure 2,
panel 1. The reader should recognize the vertical line as belonging to the dimension la-
beled affect intensity (Larsen & Diener, 1987), activation (Carroll, Yik, Russell & Feldman
Barrett,1999;Larsen & Diener, 1992; Russell & Feldman Barrett, 1999),arousal (Feldman,
1995a, 1995b; Weiss & Cropanzano, 1996), or engagement (Watson et al., 1999; Watson &
Tellegen, 1985). The horizontal dimension corresponds to hedonic tone (Larsen & Diener,
1985), pleasure–displeasure (Judge & Larsen, 2001), pleasantness–unpleasantness (Larsen
& Diener, 1992; Wundt, 1897), and mood valence (Russell & Carroll, 1999a).
Something like this circumplex has been derived from two different types of analyses.
Much of the original work used multi-dimensional scaling techniques (e.g., Russell, 1978,
1979, 1980;Russell, Weiss & Mendelsohn, 1989). In these studies, respondents classified
affect adjectives in terms of their semantic similarity. For example, “content” and “happy”
wereclose(i.e.,similar)toeachother,but both were distant (i.e., dissimilar) from “scornful”
(Watson & Tellegen, 1985). Other work has tended to use factor analytic techniques. With a
few exceptions the results of this work have been generally consistent with the findings for
similarity ratings (e.g., Feldman, 1995a, 1995b; Mayer & Gaschke, 1988; Meyer & Shack,
1989; Remington et al., 2000).
Integrating hedonic tone/activation with PA/NA. As considered thus far, the circumplex
modeldoeslittletointegratethedifferentaffectivestructures.However, this matter becomes
clearer when one considers how the principle axes are derived. If the “real” underlying
structure is circular, and if the adjective ratings are factor analyzed, the unrotated solution is
similar to the aforementioned dimensions—hedonic tone and intensity (Wright & Bonett,
1997a, 1997b). These factors are displayed in Figure 2, panel 1.
However, when these axes are rotated by 45◦two other factors appear (Cropanzano &
Wright, 2001; Judge & Larsen, 2001). These are the negative and positive affectivity axes
describedbyWatson(1988a, 1988b). The two diagonal axes in Figure 2, panel 2, show these
rotated dimensions. According to Watson et al. (1988), the high PA dimension is anchored
with such items as “enthusiastic” and “active,” while the high NA dimension is anchored
with such items as “distressed” and “irritable.”

846 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
Figure 2. The circumplex model of affect. The unrotated solution appears in panel 1, the rotated solution is in panel 2.

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 847
There are few clean markers of low PA and low NA (Weiss & Cropanzano, 1996),
prompting Zevon and Tellegen (1982: 112) to describe the dimensions as “descriptively
bipolar but affectively unipolar.” This problem not withstanding (at least for now), the
circumplex offers an elegant integration of all four affective dimensions. Essentially, they
mark different slices of a circle and, commensurate with this, are different factor solutions.
The unrotated solution recovers hedonic tone and intensity; the rotated solution nicely
recovers NA and PA. In theory, either solution can fit (Green & Salovey, 1999; Meyer &
Shack, 1989), as long as one samples adjectives evenly across the circumplex (Gotlib &
Meyer, 1986; Larsen & Diener, 1992).
Interpreting the circumplex. For many, the circumplex model is difficult to grasp intu-
itively. Trait models, such as Watson and Tellegen’s (1999) hierarchy, make more sense.
The reason for this might be the epistemology of our discipline. In common parlance, we
tend to think of factors as clusters of related things, such as items on a test. These clusters
serve as indicators for an underlying construct, which—even if not directly observable—is
often presumed to have an actual or objective quality about it.
This view of factors as clusters of indicators has considerable merit, but it is not the only
perspective that one could take. An alternative, though not necessarily inconsistent, view is
to conceptualize factors as points on a compass. From this perspective, a dimension is not
necessary a “real” thing, rather it is a directional coordinate that helps you locate the items
that interest you. Within circumplex models, the dimensions or axes function as guideposts.
They help us organize mood experiences, much as four primary directions—north, south,
east, and west—help us organize our physical environments.
Instead of mapping the surface of the earth, mood theorists are mapping the landscape of
human feelings. They need directions. So researchers use factor analysis to identify the key
axes that allow them to locate different types of affect. These axes involve some choices
by researchers (though these are constrained by the data), in the loose sense that more than
one set of dimensions could fit the data (Green & Salovey, 1999).
The circumplex and discrete emotions. Given that, as we have already observed, emo-
tions and moods are distinguishable constructs (e.g., Kelly & Basade, 2001; Larsen et al.,
2002), it is useful to consider the relationship between circumplex models and discrete
emotions. This issue was taken up by Feldman Barrett (1998),Feldman Barrett and Russell
(1998),andRussell and Feldman Barrett (1999). The “bottom line” result from this program
is displayed in Figure 3.
Here, disgust, anger, and fear are all unpleasant/high activation emotions. Sadness is
unpleasant/low(er) activation, while happiness is pleasant, high activation. Surprise is char-
acterized by an activated state, but no hedonic tone is implied. For this reason, surprise is
percheddirectly on the activationaxis. It is debatable whether surprise should be considered
an emotion at all.
Interestingly, Russell and Feldman Barrett (1999) present no basic emotions that char-
acterize the pleasant/low activation state. One possibility is that the absence of emotion is
a non-event. Thus, it is not well represented in our language. A more intriguing prospect
is that researchers may have undersampled the domain of emotion. One possible candi-
date may be the feeling of long-term, companionate love. According to Sternberg (1987),

848 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
Figure 3. Integration of basic emotions with the affect circumplex. Adapted from Russell and Feldman Barrett
(1999).
companionate love lacks “hot” passion (Feldman Barrett, 1998, would refer to this as “low
activation”). However, it remains a pleasant affective state characterized by a good deal of
intimacy. It could be that broadening the circumplex to include different kinds of love might
allow scholars to fill in the missing quadrants. Even if a slice of the mood circumplex were
devotedto love, the empty space in Figure 3wouldstill leavealastingimpression. Emotions
comprise well under 50% of the circumplex. This suggests that our mood experiences are
not always emotional in nature. In other words, all emotions seem to involve some aspect
of mood, but all moods do not seem to involve some aspect of emotion.
Critiques of the circumplex: problems with activation/intensity axis. As means of in-
tegrating the NA/PA structure with the hedonic tone/activation model, the circumplex has
an egalitarian quality about it. Both models are “correct;” it’s all a matter of how you look
at things. However, when real data are used the results can be a messy departure from the
ideal presented in Figure 2.
Most notably, various studies have had difficulty identifying adjectives that correspond
to low hedonic tone with low activation (Mayer & Gaschke, 1988; Meyer & Shack, 1989).
In other words, the affect intensity dimension may lack the bipolarity that is present in the
hedonic tone dimension. As a result, one axes of the circumplex often fails to provide a
clean fit to the data (Watson & Clark, 1997; Watson & Tellegen, 1999; Watson et al., 1999).
Such a conclusion has a conceptual tidiness about it. However, the matter is complicated
by some additional findings. Feldman (1995a, 1995b) found that there is a tendency for
individuals to attend less to arousal and more to hedonic tone, but this tendency varies
across individuals. Feldman (1995b) reports that at least some respondents will attend to the
arousal/intensity dimensions when reporting daily mood experiences. For these individuals,

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 849
Figure 4. Portions of the affect circumplex sampled by the Positive and Negative Affectivity Schedule (PANAS).
at least, the affective circumplex would presumably hold. However, the dialogue on this
issue continues and only future research will help us resolve this matter.
The activation/intensity axis and its implications for NA and PA. With so few low ac-
tivation adjectives, measures sometimes lack items indicative of low NA and low PA. For
example, the widely used PANAS instrument (Watson, 1988a, 1988b) tends to emphasize
high positive affectivity and high negative affectivity (Larsen & Diener, 1992; Russell &
Carroll, 1999a). Or, to state the matter in terms of the unrotated solution, PA roughly
corresponds to high hedonic tone/high arousal, while NA roughly corresponds to low he-
donic tone/high arousal (see Figure 3). There are no markers for low activation (Wright &
Doherty, 1998). For this reason, Thayer (1989) terms these dimensions “energetic arousal”
and “tense arousal” for PA and NA, respectively. These missing items have implications
for the structure of affect. Indeed, for the PANAS the low correlation between NA and PA
might have been anticipated based on this concern.
We illustrate this issue with Figure 4.Figure 4 displays three items used by the PANAS
to assess NA: upset, irritable, and nervous. There are also three items used by the PANAS to
assess PA: interested, excited, and enthusiastic. Regarding the segments of the circumplex
delineated by these six adjectives, there are two patterns of note. First, since the low activa-

850 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
tion adjectives are missing, neither axis can bisect the circumplex. As a consequence, most
of the structure is excluded. We are left with one slice for NA and another for PA. Second,
and more critical here, the NA and PA segments are not directly opposite one another. In a
circumplex model, high negative correlations are to be expected for adjectives that occupy
opposite sides of the circle. In our model, “happy” should be highly negatively correlated
with “sad.” However, without low activation adjectives, the negative pole of each dimension
lies unspecified. To the extent that one accepts the circumplex as a valid representation, then
PANAS–NA and PANAS–PA cannot be opposites. As illustrated in Figure 4, they are al-
most,though not quite, adjacent. Given their location, one would expect a modest (negative)
correlation between PANAS–NA and PANAS–PA (Russell & Carroll, 1999a, 1999b). In
fact, this is what the data show (Tellegen et al., 1999a, 1999b; Watson et al., 1999).
This discussion suggests an intriguing possibility. NA and PA might appear distinct be-
causeproponentsof the NA/PAstructurehave undersampledtheadjectivesthatrepresentthe
bulkof the circumplex (Larsen & Diener,1992). When a more complete sampling of items is
included,thecorrelation between NA andPAappearstoincrease (Russell & Carroll, 1999a).
As a consequence of these issues, some scholars are beginning to argue that affect is not,
strictly speaking, comprised of two orthogonal dimensions. Rather, positive and negative
moods are negatively associated (Russell & Carroll, 1999a; Tellegen et al., 1999a, 1999b).
Diener and his colleagues (1995: 130) opined that “pleasant and unpleasant affect are not
strictly orthogonal, but they are separable.” Of course, some issues remain. As we have
seen, Tellegen and his colleagues prefer a hierarchical view of affect, while Russell and
his colleagues seem to prefer a circumplex view. Regardless, it is important to emphasize
that neither of these models rules out the importance of NA or PA. Rather, they make sense
of these two constructs within the context of different taxonomic systems. Which of these
systems proves to be the most useful can only be determined by additional inquiry.
Conclusion
As new questions are asked and as new data are collected, some of the ideas and conclu-
sions expressed herein will no doubt be sharpened, revised, and in some cases disproved.
We see the contributions of this paper as two fold. First, we seek to alert scholars to the
importance of taxonomic issues in the development and testing of conceptual models. Sec-
ond, we attempt to provide a thorough review of scientific progress to date. In order to
bring these two contributions home, we will consider three final speculations on the place
of structure in the study of affect: the problem of the whole, the problem of the parts, and
managerial prescriptions.
The Problem of the Whole
At a high level of abstraction there exists what we term the problem of the whole. Our
concern here is that affect researchers could divide themselves into groups, some selecting
NA and PA, while others are content with hedonic tone and affect intensity. While worth-
while research can be conducted within both traditions, a full understanding requires that
they be integrated. This integration, in turn, implies a meta-structure that can subsume both

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 851
factor solutions. Lacking this, it is difficult to understand how the results could converge
together to enhance our knowledge. Of course, a broader model that integrates the two
factors structures—such as the affect hierarchy and the circumplex—goes far in address-
ing this concern. While many questions remain to be asked and answered, we believe that
considerable progress has been made toward resolving the problem of the whole.
The Problem of the Parts
At a lower level of abstraction, lack of understanding about the structure of affect also has
a different kind of pernicious effect. We term this the problem of the parts. If researchers do
not have a nuanced understanding of affective structure, then they are more likely to exclude
potentially important variables from their investigations. As a consequence of losing these
parts affective experience will be less well understood than it could be. For example, organi-
zational scientists have spent little time investigating the role of affect intensity in organiza-
tions (but see Weiss et al., 1999, for an exception). This state of affairs may have transpired
because the popular PA/NA structure does not contain an explicit role for intensity. Many
aspects of affect will require more research to determine their importance for organizational
behavior. A firm grounding in the structure of affect will facilitate this advancement.
Managerial Prescriptions
Researchers operating from different conceptual understandings of the structure of af-
fect might offer different managerial prescriptions. Consider the following example. We
know that organizations benefit from employees’ PA through increased pro-social behavior
(George, 1991; George & Bettenhausen, 1990) and reduced absenteeism (George, 1989).
Thus, in many cases, managers might be encouraged to increase PA. Yet depending upon
how the structure of affect is understood, managers could take different actions to achieve
this goal. If one assumes that NA and PA are opposite ends of a single bipolar dimension,
then reducing NA would be expected to increase PA, since movement from one pole neces-
sarily implies movement toward the other. However, in NA and PA are separate dimensions,
then reducing NA might have little or no effect on PA.
We can illustrate this conundrum with a simple question: Will reducing the threat of
punishment raise positive affect and, by extension, promote more effective organizational
behavior? Perhaps and perhaps not. Larsen and Ketelaar (1989, 1991) have shown that the
threat of punishment increases negative affect (Larsen & Ketelaar, 1989, 1991). Positive
affect, on the other hand, is more closely related to the promise of reward. If Larsen and
Ketelaar are correct, then increasing positive affect will entail more than simply reducing
threat. This is because a decrease in NA does not necessarily result in an increase in PA.
(Of course, there are lots of other good reasons for taking fear out of the workplace!)
Concluding Thoughts
This chapter has taken a long walk over controversial ground. It is somewhat akin to
reviewing a film before the show is concluded. We await the rolling credits only to discover
that the climax is still in store. In the final analysis, we are more like correspondents and

852 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
less like historians. Is there anything we can conclude for now? We believe there are some
tantalizing suggestions from the available literature.
First,while contexteffectsare importantintheir ownright,it does notseemlikelythat they
will explain away the NA/PA factor structure. While it seems to be the case that accounting
for intensity and removing measurement error (and perhaps using a short time frame) will
increase the relationship between NA and PA, the overall structure remains viable. In fact,
it was only the increase in this association that made the higher-order, hedonic tone factor
possible (Tellegen et al., 1999a, 1999b). Thus, future debate is likely to center on the
three-level hierarchy vs. the mood circumplex.
Second, the principle discrepancy between the hierarchy and the circumplex seems to
regard the axis variously labeled “activation,” “arousal,” “intensity,” or “engagement.”
Within the circumplex model, intensity is (supposed to be) a vigorous axis, on par with
hedonic tone. Within the hierarchy model, NA and PA subsume intensity. Clearly, future
research needs to be brought to bear on this issue. At the moment, the activation dimension
is extremely variable, showing up more strongly in similarity ratings than in self-reports
(Feldman,1995a) and more strongly for some individualsthan forothers (Feldman, 1995b).
The debate between the mood circumplex and mood hierarchy may well be decided on this
issue. If the activation dimension fails to hold up, then we will be seeing a good deal more
of the hierarchy model and a good deal less of the circumplex model.
Third, there seems to be emerging consensus regarding the hedonic tone factor. Schol-
ars seem to agree that the structure of affect is at least partially captured by a bipolar
dimension describe by such terms as “happy-sad” and “good-bad.” Though this dimension
goes by different names, such as pleasant–unpleasant (Larsen & Diener, 1992), it serves
an important role in both the hierarchical and the circumplex model of affect. This is good
news for researchers in that it allows scholars to safely consider the important construct of
psychological well-being in understanding organizational behavior (Warr, 1999).
Fourth, while NA and PA remain useful dimensions, they need to be understood as part
of a larger taxonomic structure. In particular, NA and PA could be either lower order factors
in Watson and Tellegen’s (1999) hierarchyor segments of the affect circumplex (in Russell
& Carroll’s, 1999, view). In either case, we would have richer insight into both variables,
since our knowledge would be more fully integrated with related constructs.
Research on affective structure is like a tapestry that is not yet complete. As scholars sew
and weave, the outlines of our narrative have begun to appear, but the details (and indeed,
some issues larger than details) remain for the future.
References
Amabile, T. M. 1996. Creativity in context. Boulder, CO: Westview Press.
Ashford, B., & Humphrey, R. H. 1993. Emotional labor in service roles: The influence of identity. Academy of
Management Review, 18: 88–115.
Ashford, B. E., & Humphrey, R. H. 1995. Emotion in the workplace: A reappraisal. Human Relations, 48: 97–125.
Ashkanasy, N. M., Härtel, C. E. J., & Zerbe, W. J. 2000. Emotions in the workplace: Research, theory, and practice.
In N. M. Ashkanasy, C. E. J. Härtel, & W. J. Zerbe (Eds.), Emotions in the workplace: Research, theory, and
practice: 1–18. Westport, CT: Quorum Books.

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 853
Ashkanasy, N. M., & Tse, B. 2000. Transformational leadership as management of emotions. In N. M. Ashkanasy,
C. E. J. Härtel, & W. J. Zerbe (Eds.), Emotions in the workplace: Research, theory, and practice: 221–235.
Westport, CT: Quorum Books.
Baron, R. A. 1976. The reduction of human aggression: A field study of the influence of incompatible responses.
Journal of Applied Social Psychology, 6: 260–274.
Baron, R. A. 1984. Reducing organizational conflict: An incompatible response approach. Journal of Applied
Psychology, 69: 272–279.
Beal, D. J., & Weiss, H. M. 2002. Methods of ecological momentary assessment in organizational research.
Unpublished manuscript.
Bradburn, N. M. 1969. The structure of psychological well-being. Chicago: Aldine.
Bradburn, N. M., & Caplovitz, D. 1965. Reports on happiness: A pilot study. Chicago: Aldine.
Brenner, B. 1975. Quality of affect and self-evaluated happiness. Social Indicators Research, 2: 315–331.
Brief, A. P., & Roberson, L. 1989. Job attitude organization: An exploratory study. Journal of Applied Social
Psychology, 19: 717–727.
Brief, A. P., & Weiss, H. M. 2002. Organizational behavior: Affect in the workplace. In S. T. Fiske, D. L. Schacter,
& C. Zahn-Waxler (Eds.), Annual review of psychology: Vol. 53, 279–307. Palo Alto, CA: Annual Reviews,
Inc.
Burke, M. J., Brief, A. P., George, J. M., Roberson, L., & Webster, J. 1989. Measuring affect at work: Confirmatory
analyses of competing mood structures with conceptual linkages to cortical regulatory systems. Journal of
Personality and Social Psychology, 57: 1091–1102.
Campbell, A., Converse, P. E., & Rodgers, W. L. 1976. The quality of American life. New York: Russell Sage
Foundation.
Carroll, J. M., Yik, M. S. M., Russell, J. A., & Feldmand Barrett, L. F. 1999. On the psychometric principles of
affect. Review of General Psychology: Vol. 3, 14–22. Palo Alto, CA: Annual Reviews, Inc.
Cropanzano, R., James, K., & Konovsky, M. A. 1993. Dispositional affectivity as a predictor of work attitudes
and job performance. Journal of Organizational Behavior, 14: 595–606.
Cropanzano, R., Rupp, D. E., & Byrne, Z. S. 2003. The relationship of emotional exhaustion to work attitudes,
job performance, and organizational citizenship behaviors. Journal of Applied Psychology, 88: 160–169.
Cropanzano, R., & Wright, T. A. 1999. A 5-year study of change in the relationship between well-being and job
performance. Consulting Psychology Journal, 51: 252–265.
Cropanzano, R., & Wright, T. A. 2001. When a “happy” worker is really a “productive” worker: A review and
further refinements of the happy—productive worker thesis. Consulting Psychology Journal, 53: 182–199.
Diener, E. 1984. Subjective well-being. Psychological Bulletin, 95: 542–575.
Diener, E. 1994. Assessing subjective well-being: Progress and opportunities. Social Indicators Research, 31:
103–157.
Diener, E., & Emmons, R. A. 1985. The independence of positive and negative affect. Journal of Personality and
Social Psychology, 47: 1105–1117.
Diener, E., & Iran-Nejad, A. 1986. The relationship in experience between various types of affect. Journal of
Personality and Social Psychology, 50: 1031–1038.
Diener, E., & Larsen, R. J. 1993. The experience of emotional well-being. In M. Lewis & J. M. Haviland (Eds.),
Handbook of emotions: 450–415. New York: The Guilford Press.
Diener, E., Smith, H., & Fujita, F. 1995. The personality structure of affect. Journal of Personality and Social
Psychology, 69: 130–141.
Diener, E., Suh, E. M., Lucas, R. E., & Smith, H. L. 1999. Subjective well-being: Three decades of progress.
Psychological Bulletin, 125: 276–302.
Fabrigar, L. R., Visser, P. S., & Browne, M. W. 1997. Conceptual and methodological issues in testing circumplex
data in personality and social psychology. Personality and Social Psychology Review, 1: 184–203.
Faith, M., & Thayer, J. F. 2001. A dynamical systems interpretation of a dimensional model of emotion.
Scandinavian Journal of Psychology, 42: 121–133.
Feldman, L. A. 1995a. Variations in the circumplex structure of mood. Personality and Social Psychology Bulletin,
21: 806–817.
Feldman, L. A. 1995b. Valence-focus and arousal-focus: Individual differences in the structure of affective
experience. Journal of Personality and Social Psychology, 69: 153–166.

854 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
Feldman Barrett, L. A. 1998. Discrete emotions or dimensions? The role of valence focus and arousal focus.
Cognition and Emotion, 12: 576–599.
Feldman Barrett, L. A., & Russell, J. A. 1998. Independence and bipolarity in the structure of affect. Journal of
Personality and Social Psychology, 74: 967–984.
Frijda, N. H., Ortony, A., Sonnemans, J., & Clore, G. L. 1992. The complexity of intensity: Issues concerning the
structure of emotion intensity. In M. S. Clark (Ed.), Review of personality and social psychology: Emotion:
Vol. 13, 60–89. Newbury Park, CA: Sage.
George, J. M. 1989. Mood and absence. Journal of Applied Psychology, 74: 317–324.
George, J. M. 1991. State or trait: Effects of positive mood on prosocial behaviors at work. Journal of Applied
Psychology, 76: 299–307.
George, J. M. 1992. The role of personality in organizational life: Issues and evidence. Journal of Management,
18: 185–210.
George, J. M. 1996. Trait and state affect. In K. M. Murphy (Ed.), Individual differences in behavior in
organizations: 145–171. San Francisco, CA: Jossey-Bass.
George, J. M., & Bettenhausen, K. 1990. Understanding prosocial behavior, sales performance, and turnover: A
group level analysis in a service context. Journal of Applied Psychology, 75: 698–709.
George, J. M., & Zhou, J. 2002. Understanding when bad moods foster creativity and good ones don’t: The role
of context and clarity of feelings. Journal of Applied Psychology, 87: 687–697.
Goleman, D. 1998. Working with emotional intelligence. New York: Bantam Books.
Goleman, D., Boyatzis, R., & McKee, A. 2002. Primal leadership: Realizing the power of emotional intelligence.
Boston, MA: Harvard Business School Press.
Gotlib,I. H., & Meyer,J. P.1986. Factor analysisof the multiple affect adjectivecheck list: A separation of positive
and negative affect. Journal of Personality and Social Psychology, 50: 1161–1165.
Grandey, A. A. 2000. Emotion regulation in the workplace: A new way to conceptualize emotional labor. Journal
of Occupational Health Psychology, 5: 95–110.
Green, D. P., Goldman, S. L., & Salovey, P. 1993. Measurement error masks bipolarity in affect ratings. Journal
of Personality and Social Psychology, 64: 1029–1041.
Green, D. P., & Salovey, P. 1999. In what sense are positive and negative affect independent?: A reply to Tellegen,
Watson, and Clark. Psychological Science, 10: 304–306.
Green, D. P., Salovey, P., & Truax, K. M. 1999. Static, dynamic, and causative bipolarity of affect. Journal of
Personality and Social Psychology, 76: 865–867.
Isen, A. M. 1999. On the relationship between affect and creative problem solving. In S. W. Russ (Ed.), Affect,
creative experience, and psychological adjustment: 3–17. Philadelphia, PA: Taylor & Francis.
Isen, A. M., & Baron, R. A. 1991. Positive affect as a factor in organizational behavior. In B. M. Staw & L. L.
Cummings (Eds.), Research in organizational behavior: Vol. 13, 1–54. Greenwich, CT: JAI Press.
James, K., Clark, K., & Cropanzano, R. 1999. Positive and negative creativity in groups, institutions, and
organizations: A model and theoretical extension. Creativity Research Journal, 12: 211–226.
Judge, T. A. 1992. The dispositional perspective in human resource research. In G. R. Ferris & L. M. Rowland
(Eds.), Research in personnel and human resources management: Vol. 11, 31–72. Greenwich, CT: JAI Press.
Judge, T. A., & Larsen, R. J. 2001. Dispositional affect and job satisfaction: A review and theoretical extension.
Organizational Behavior and Human Decision Processes, 86: 67–98.
Kelly, J. R., & Basade, S. G. 2001. Mood and emotions in small groups and work teams. Organizational Behavior
and Human Decision Processes, 86: 99–130.
Larsen, R. J. 1987. The stability of mood variability: A spectral analytic approach to daily mood assessments.
Journal of Personality and Social Psychology, 52: 1195–1204.
Larsen, R. J., & Diener, E. 1985. A multitrait–multimethod examination of affect structure: Hedonic level and
emotional intensity. Personality and Individual Differences, 6: 631–636.
Larsen, R. J., & Diener, E. 1987. Affect intensity as an individual difference characteristic: A review. Journal of
Research in Personality, 21: 1–39.
Larsen, R. J., & Diener, E. 1992. Promises and problems with the circumplex model of emotion. In M. S. Clark
(Eds.), Review of personality and social psychology: Emotion: Vol. 13, 25–59. Newbury Park, CA: Sage.
Larsen, R. J., Diener, E., & Cropanzano, R. S. 1987. Cognitive operations associated with individual differences
in affect intensity. Journal of Personality and Social Psychology, 53: 767–774.

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 855
Larsen, R. J., Diener, E., & Lucas, R. E. 2002. Emotion: Models, measures, and individual differences. In R. G.
Lord, R. J. Klimoski, & R. Kanfer (Eds.), Emotions in the workplace: Understanding the structure and role of
emotions in organizational behavior: 64–106. New York: Jossey-Bass.
Larsen, R. J., & Ketelaar, T. 1989. Extraversion. Personality and Individual Differences, 10: 1221–1228.
Larsen, R. J., & Ketelaar, T. 1991. Personality and susceptibility to positive and negative emotional states. Journal
of Personality and Social Psychology, 61: 132–140.
Larsen, J. T., McGraw, A. P., & Cacioppo, J. T. 2001. Can people feel happy and sad at the same time? Journal of
Personality and Social Psychology, 81: 684–696.
Lorr, M., Quing Shi, A., & Youniss, R. P. 1989. A bipolar multifactor conception of mood states. Personality and
Individual Differences, 10: 155–159.
Mayer, J. D., & Gaschke, Y. N. 1988. The experience and meta-experience of mood. Journal of Personality and
Social Psychology, 57: 691–706.
McConville, C., & Cooper, C. 1992. The structure of moods. Personality and Individual Differences, 8: 909–919.
Morris,J. A.,& Feldman,D. C.1996. Thedimensions, antecedents,and consequencesof emotionallabor. Academy
of Management Review, 21: 986–1010.
Meyer, G. J., & Shack, J. R. 1989. Structural convergence of mood and personality: Evidence for old and new
directions. Journal of Personality and Social Psychology, 57: 691–706.
Myers, D. G. 1992. The pursuit of happiness. New York: Avon Books.
Pennebaker, J. W. 1982. The psychology of physical symptoms. New York: Springer–Verlag.
Rafaeli, A., & Sutton, R. I. 1990. Busy stores and demanding customers: How do they affect the display of positive
emotion? Academy of Management Journal, 33: 623–637.
Remington, N. A., Fabrigar, K. R., & Visser, P. S. 2000. Reexamining the circumplex model of affect. Journal of
Personality and Social Psychology, 79: 286–300.
Robinson, M. D., & Clore, G. L. 2002. Belief and feeling: Evidence for an accessibility model of emotional self
report. Psychological Bulletin, 128: 934–960.
Russell, J. A. 1978. Evidence of convergent validity on the dimensions of affect. Journal of Personality and Social
Psychology, 36: 1152–1168.
Russell, J. A. 1979. Affective space is bipolar. Journal of Personality and Social Psychology, 37: 345–356.
Russell, J. A. 1980. A circumplex model of affect. Journal of Personality and Social Psychology, 39: 1161–1178.
Russell, J. A., & Carroll, J. M. 1999a. On the bipolarity of positive and negative affect. Psychological Bulletin,
125: 3–30.
Russell,J. A.,& Carroll, J. M. 1999.The phoenixof bipolarity:Replay to Watson& Tellegen(1999).Psychological
Bulletin, 125: 611–617.
Russell, J. A., & Feldman Barrett, L. F. 1999. Core affect, prototypical emotional episodes, and other things called
emotion: Dissecting the elephant. Journal of Personality and Social Psychology, 76: 805–819.
Russell, J. A., Weiss, A., & Mendelsohn, G. A. 1989. Affect Grid: A single-item scale of pleasure and arousal.
Journal of Personality and Social Psychology, 57: 493–502.
Shaver, P., Schwartz, J., Kirson, D., & O’Connor, C. 1987. Emotion knowledge: Further exploration of a prototype
approach. Journal of Personality and Social Psychology, 52: 1061–1086.
Simonsson-Sarnecki, M., Lundh, L.-G., & Törestad, B. 2000. Factor structure and validity of the affect intensity
measure in a Swedish sample. Personality and Individual Differences, 29: 337–350.
Staats,S., Partlo, C.,& Adam, N. 1989. When time frame makes adifference. Paperpresented at theannual meeting
of the Midwestern Psychological Association. Chicago, IL.
Staw, B. M., & Barsade, S. G. 1993. Affect and managerial performance: A test of the sadder-but-wiser vs.
happier-and-smarter hypotheses. Administrative Science Quarterly, 38: 304–331.
Staw, B. M., Sutton, R. I., & Pelled, L. H. 1994. Employee positive emotion and favorable outcomes at the
workplace. Organization Science, 5: 51–71.
Sternberg, R. J. 1987. The triangle of love: Intimacy, passion, commitment. New York: Basic Books.
Sutton,R. I.1991. Maintainingnorms aboutexpressed emotions:The caseof billcollectors. AdministrativeScience
Quarterly, 36: 245–268.
Sutton, R. I., & Rafaeli, A. 1988. Untangling the relationship between displayed emotions and organizational
sales: The case of convenience stores. Academy of Management Journal, 31: 461–487.
Tellegen, A., Watson, D., & Clark, L. A. 1999a. On the dimensional and hierarchical structure of affect.
Psychological Science, 10: 297–303.

856 R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857
Tellegen, A., Watson, D., & Clark, L. A. 1999. Further support for a hierarchical model of affect: Reply to Green
and Salovey. Psychological Science, 10: 307–309.
Thayer, R. E. 1989. The biopsychology of mood and arousal. New York: Oxford University Press.
Tsai, W.-C. 2001. Determinants and consequences of employee displayed positive emotions. Journal of
Management, 27: 497–512.
Tsai, W.-C., & Huang, Y.-M. 2002. Mechanisms linking employee affective delivery and customer behavioral
intentions. Journal of Applied Psychology, 87: 1001–1008.
Vosburg, S., & Kaufmann, G. 1999. Mood and creativity research: The view from a conceptual organizing
perspective. In S. W. Russ (Ed.), Affective, creative experience, and psychological adjustment: 19–39.
Philadelphia, PA: Taylor & Francis.
Warr, P. 1999. Well-being in the workplace. In D. Kahneman, E. Diener, & N. Schwarz (Eds.), Well-being: The
foundations of hedonic psychology: 392–412. New York: Russell Sage Foundation.
Warr, P., Barter, J., & Brownbridge, G. 1983. On the independence of positive and negative affect. Journal of
Personality and Social Psychology, 44: 644–651.
Watson, D. 1988a. The vicissitudes of mood measurement: Effects of varying descriptors, time frames, and
response formats on measures of positive and negative affect. Journal of Personality and Social Psychology,
55: 128–141.
Watson, D. 1988. Intraindividual and interindividual analyses of positive and negative affect: Their relation to
health complaints, perceived stress, and daily activities. Journal of Personality and Social Psychology, 55:
128–141.
Watson, D. 2000. Mood and temperament. New York: Guilford Press.
Watson, D., & Clark, L. A. 1984. Negative affectivity: The disposition to experience negative emotional states.
Psychological Bulletin, 96: 465–490.
Watson, D., & Clark, L. A. 1992. On traits and temperaments: General and specific factors of emotional experience
and their relation to the five-factor model. Journal of Personality, 60: 441–476.
Watson, D., & Clark, L. A. 1997. The measurement and mismeasurement of mood: Recurrent and emergent issues.
Journal of Personality Assessment, 86: 267–296.
Watson, D., Clark, L. A., & Tellegen, A. 1984. Cross-cultural convergence in the structure of mood: A Japanese
replication and a comparison with U.S. findings. Journal of Personality and Social Psychology, 47: 127–144.
Watson, D., Clark, L. A., & Tellegen, A. 1988. Development and validation of brief measures of positive and
negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54: 1063–1070.
Watson, D., & Tellegen, A. 1985. Towards a consensual structure of mood. Psychological Bulletin, 98: 127–144.
Watson,D., &Tellegen,A. 1999.Issues inthe dimensionalstructure ofaffect—Effectsof descriptors,measurement
error, and response formats: Comment on Russell and Carroll (1999). Psychological Bulletin, 125: 609–610.
Watson, D., Wiese, D., Vaidya, J., & Tellegen, A. 1999. The two general activation systems of affect: Structural
findings, evolutionary considerations, and psychobiological evidence. Journal of Personality and Social
Psychology, 76: 820–838.
Weiss, H. M. 2002. Conceptual and empirical foundations for the study of affect at work. In R. G. Lord, R. J.
Klimoski, & R. Kanfer (Eds.) Emotions in the workplace. San Francisco: Jossey-Bass.
Weiss, H. M., & Brief, A. P. 2001. Affect at work: An historical perspective. In R. L. Payne & C. L. Cooper (Eds.),
Emotions at work: Theory, research, and application in management. Chichester, UK: John Wiley & Sons.
Weiss, H. M., & Cropanzano, R. 1996. An affective events approach to job satisfaction. In B. M. Staw & L. L.
Cummings (Eds.), Research in organizational behavior: Vol. 18, 1–74. Greenwich, CT: JAI Press.
Weiss, H. M., Nicolas, J. P., & Daus, C. S. 1999. An examination of the joint effects of affective experience and
job beliefs on job satisfaction and variations in affective experience over time. Organizational Behavior and
Human Decision Processes, 78: 1–24.
Wertheimer, M. 1987. A brief history of psychology (3rd ed.). Fort Worth, TX: Holt, Rinehart & Winston, Inc.
Wessman, A.E. & Ricks, D.F. 1966. Mood and personality. New York: Holt, Rinehart & Winston.
Wiggins, J. S. 1980. Circumplex models of interpersonal behavior. In L. Wheeler (Ed.), Review of personality and
social psychology: Vol. 1, 265–294. Beverly Hills, CA: Sage.
Wiggins, J. S., Phillips, N., & Trapnell, P. 1989. Circular reasoning about interpersonal behavior: Evidence
concerning some untested assumptions underlying diagnostic classification. Journal of Personality and Social
Psychology, 56: 297–305.

R. Cropanzano et al./ Journal of Management 2003 29(6) 831–857 857
Wiggins, J. S., Steiger, J. H., & Gaelick, L. 1981. Evaluating circumplexity in personality data. Multivariate
Behavioral Research, 16: 263–289.
Wright, T. A., & Bonett, D. G. 1997a. The role of pleasantness and activation-based well-being in performance
prediction. Journal of Occupational Health Psychology, 2: 212–219.
Wright, T. A., & Bonett, D. G. 1997. The contribution of burnout to work performance. Journal of Organizational
Behavior, 18: 491–499.
Wright,T.A.,Bonett, D.G., &Sweeney,D. A. 1993.Mental healthand workperformance: Resultsof a longitudinal
field study. Journal of Occupational and Organizational Psychology, 66: 277–284.
Wright, T. A., & Cropanzano, R. 1998. Emotional exhaustion as a predictor of job performance and voluntary
turnover. Journal of Applied Psychology, 83: 486–493.
Wright, T. A., & Cropanzano, R. 2000. Psychological well-being and job satisfaction as predictors of job
performance. Journal of Occupational Health Psychology, 5: 84–89.
Wright, T. A., Cropanzano, R., Denny, P. J., & Moline, G. L. 2002. When a happy worker is a productive worker:
A preliminary examination of three models. Canadian Journal of Behavioural Science, 34: 146–150.
Wright, T. A., & Doherty, E. M. 1998. Organizational behavior “rediscovers” the role of emotional well-being.
Journal of Organizational Behavior, 19: 481–485.
Wright, T. A., & Staw, B. M. 1999. Affect and favorable work outcomes: Two longitudinal tests of the
happy-productive worker thesis. Journal of Organizational Behavior, 20: 1–23.
Wundt, W. 1897. In C. H. Judd (Trans.), Outline of psychology. Leipzig: Wilhelm Englemann.
Yik, M. S. M., Russell, J. A., & Feldman Barrett, L. 1999. Structure of self-reported current affect: Integration
and beyond. Journal of Personality and Social Psychology, 77: 600–619.
Zautra, A. J., Potter, P. T., & Reich, J. W. 1997. The independence of affect is context-dependent: An integrative
model of the relationship between positive and negative affect. In M. P. Lawton & K. W. Schaie (Eds.), Annual
review of gerontology and geriatrics: Vol. 17, 75–102. Palo Alto, CA: Annual Reviews, Inc.
Zevon, M. A., & Tellegen, A. 1982. The structure of mood change: An idiographic/nomothetic analysis. Journal
of Personality and Social Psychology, 43: 111–122.
Russell Cropanzano is a Professor of Organizational Behavior in the Department of Man-
agement and Policy of the University of Arizona. His research interests include workplace
emotion and organizational justice.
Howard M. Weiss is professor of psychological sciences at Purdue University. He is also
co-directorof Purdue’s Military Family Research Institute. He received his Ph.D. from New
York University. His research interests focus on the emotions in the workplace and on job
attitudes.
Jeff M.S. Hale is a doctoral candidate in the Department of Management at the University
of Arizona. His primary areas of scholarly interest are organizational justice and business
ethics.
JochenRebis adoctoralcandidateintheManagementand Policy Department at the Univer-
sity of Arizona’s Eller College of Business and Public Administration. His current research
interests include emotions in decision making, organizational justice, and decisions in ne-
gotiations.