Content uploaded by Cynthia D. Fisher
Author content
All content in this area was uploaded by Cynthia D. Fisher on Jul 31, 2019
Content may be subject to copyright.
Mood and emotions while working:
missing pieces of job satisfaction?
CYNTHIA D. FISHER*
School of Business, Bond University, Gold Coast, Queensland 4229, Australia
Summary Job satisfaction is often described as an aective response to one's job, but is usually
measured largely as a cognitive evaluation of job features. This paper explores several
hypothesized relationships between real time aect while working and standard
measures of job satisfaction. Experience sampling methodology was used to obtain up
to 50 reports of immediate mood and emotions from 121 employed persons over a two
week period. As expected, real time aect is related to overall satisfaction but is not
identical to satisfaction. Moment to moment aect is more strongly related to a faces
measure of satisfaction than to more verbal measures of satisfaction. Positive and
negative emotions both make unique contributions to predicting overall satisfaction, and
aect accounts for variance in overall satisfaction above and beyond facet satisfactions.
Frequency of net positive emotion is a stronger predictor of overall satisfaction than is
intensity of positive emotion. It is concluded that aect while working is a missing piece
of overall job attitude, as well as a phenomenon worthy of investigation in its own right.
Implications for further research and for improving the conceptualization and measure-
ment of job satisfaction are discussed. Copyright #2000 John Wiley & Sons, Ltd.
Introduction
For many years, researchers in organizational behavior and industrial psychology have studied
job satisfaction as both an independent and a dependent variable. Cranny et al. (1992) estimate
that there have been more than 5000 published articles and dissertations which examine job
satisfaction in some way. Despite all this research, Staw (1984) points out that relatively little
work has been directed at the construct of job satisfaction itself.
Job satisfaction is an attitude. Attitudes are usually described as containing at least two
components: an aective (emotional, feeling) component, and a cognitive (belief, judgment,
comparison) component (Eagly and Chaiken, 1993). Research has shown that both of these
components are important, contribute unique variance to the overall attitude, and may be
dierentially caused and dierentially linked to behavior (Breckler and Wiggins, 1989; Millar and
CCC 0886±9383/2000/020185±18$17.50
Copyright #2000 John Wiley & Sons, Ltd.
Journal of Organizational Behavior
J. Organiz. Behav. 21, 185±202 (2000)
* Correspondence to: Cynthia D. Fisher, School of Business, Bond University, Gold Coast, Queensland 4229, Australia.
E-mail: Cynthia_Fisher@Bond.edu.au
The author wishes to thank Chris Noble for extensive assistance with data analysis.
Contract grant sponsor: Australian Research Council.
Contract grant no: 421 6204 17.
Millar, 1996; Millar and Tesser 1986; see Weiss, 2000, for a review). Job satisfaction is often
de®ned as an aective reaction toward's one's job (Cranny et al., 1992; Porac, 1987), but is
usually measured as an evaluative assessment of job attributes compared to either internal or
external standards (Locke, 1976; Rice et al., 1989; Weiss and Cropanzano, 1996). A number of
researchers have criticized job satisfaction measures as being too cognitive (c.f. Brief, 1998;
Organ and Near 1985; Pekrun and Frese, 1992). Sandelands (1988) points out that most
measures of work attitude assess `cold cognitions' rather than hot emotions, the level at which the
job is actually experienced. Porac (1987) argues that we know next to nothing about how feelings
at work are translated into responses on job satisfaction scales. These criticisms suggest a need to
explore the relationships between `hot' measures of aect while working and measures of
satisfaction.
Mood and emotions
The term `aect' is broad and encompasses two relatively distinct phenomena of interest in this
study: state moods and emotions. Moods tend to be longer lasting but often weaker states of
uncertain origin, while emotions are often more intense, short lived, and have a clear object or
cause (Frijda, 1993). Moods are usually conceptualized as having two dimensions. Depending on
how the dimensions are rotated, they can be labelled hedonic tone/pleasantness and arousal/
activation (Russell, 1980; Larsen and Diener, 1992), or positive aect and negative aect (Watson
and Tellegen, 1985). Weiss and Cropanzano (1996) suggest that the former conceptualization is
most useful for measuring state mood at work. Some scholars believe that hedonic tone is by far
the more important of these two dimensions (Russell, 1978; Warr, 1990). To the extent that job
satisfaction is `an evaluative (good/bad) response', hedonic tone would be expected to be the
more relevant aspect of mood. As predicted, Weiss et al. (1999) found that the average hedonic
tone while working was correlated with job satisfaction while average activation level was not.
Wright and Bonnett (1996) have also found that pleasantness-based measures are more predictive
in organizational research than activation-based measures. Van Katwyk et al. (1995) reported
that the pleasant/unpleasant dimension dominated descriptions of job-related aect.
Like moods, emotions can easily be classi®ed into positive and negative categories. However,
there are many more than two distinct emotions (Diener et al., 1995; Shaver et al., 1987).
Typologies of `basic' emotions usually contain from ®ve to 10 emotion terms, such as fear, anger,
sadness, disgust, joy and love (Plutchik, 1994), while emotional lexicons contain hundreds of
terms (Averill, 1975, Hunt, 1997Ðunpublished doctoral dissertation, University of North
Carolina, Chapel Hill; Ortony et al., 1987).
Experience sampling
Because state moods and emotions are transient, they are dicult to measure accurately long
after they have occurred. People over-estimate the frequency with which they have experienced
both positive and negative emotions when reporting retrospectively compared to reporting in real
time (Diener et al., 1995). Even daily mood reports are demonstrably less accurate than the
average of more frequent reports (Hedges et al., 1985). Clearly, it is preferable to obtain reports
of current mood and emotion at the time they are being experienced. Experience Sampling
Methodology (ESM) has been developed as a means of obtaining real time reports of phenomena
of this nature (c.f. Alliger and Williams, 1993; Hormuth, 1986; Larson and Csikszentmihalyi,
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
186 C. D. FISHER
1983; Wheeler and Reiss, 1991). In the present study, respondents were asked to report on their
current mood and emotions ®ve times each working day for two weeks.
Because emotions and moods are transient, there will be within-person variance in mood and
emotion reports. There should also be meaningful between-person variance in the average
positiveness or negativeness of moods and emotions experienced at work. It is the relationships of
these between-person dierences and job satisfaction that are of interest in the present study.
The relationship of mood and emotions at work to
job satisfaction
Weiss and Cropanzano's (1996) Aective Events Theory proposes that mood and emotions while
working are the raw materials which cumulate to form the aective element of job satisfaction,
while judgments or comparisons of job attributes to standards contribute to the cognitive
element of satisfaction. There is some evidence in the literature that state moods are related to
overall job satisfaction. Brief and Roberson (1989) found that a retrospective report of subjects'
mood at work over the past week (Job Aect Scale) was signi®cantly related to overall
satisfaction as assessed by the Minnesota Satisfaction Questionnaire (Weiss et al., 1967), by the
sum of all 72 Job Descriptive Index (JDI) items (Smith et al., 1969), and by a one item faces scale
(Kunin, 1955). Weiss et al. (1999) found that average pleasantness of mood, assessed four times
per day over three weeks, was signi®cantly related to a ®ve item measure of overall job satisfac-
tion. They also found that instrumentality and valence beliefs about job outcomes contributed to
satisfaction independently of mood.
There have been relatively few studies of emotions experienced at work (see Pekrun and Frese,
1992 for a review), and no systematic studies of the relationship between real-time emotions at
work and job satisfaction. Because emotions have a target (one is angry at some one, frustrated
because of an impediment in reaching a goal, proud of an accomplishment), they are likely to be
triggered by actual events in the workplace. As such, emotions should often be directly
attributable to the job, and should be more readily recalled than vague and diuse moods
experienced while on the job but not necessarily due to the job. For these reasons, Weiss and
Cropanzano (1996) suggest that emotions at work may be more relevant to job satisfaction than
are moods, though both should be related to satisfaction.
Hypothesis 1: Mood and positive and negative emotions while working will be signi®cantly
related to overall job satisfaction.
Attitude research suggests that responses to attitude surveys are constructed on demand using
the information which comes to mind at the time (see Hippler et al., 1987). Most multi-item job
satisfaction measures are belief-oriented, and so may not stimulate very much recall and
weighting of emotional content. Thus, the relationship between aect and typical measures of
overall job satisfaction may be far less than perfect. The relationship of aect to a non-verbal
satisfaction measure such as the faces scale (Kunin, 1955) may be stronger. The faces measure
does not constrain respondents to speci®c objective comparisons (e.g., my coworkers talk too
much, the work is hot, respected, tiring, etc.), but simply asks respondents to choose one of
11 drawings of facial expressions which represents their feelings about the attitude object.
Because speci®c cognitions are not primed, and facial expressions may instead cue emotional
recall, aect may be better captured when overall satisfaction is assessed with the faces scale.
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
AFFECT AND SATISFACTION 187
Brief and Roberson (1989) found that mood at work during the past week contributed much
more to predicting a faces measure of overall satisfaction than it did to the MSQ or JDI measures.
Their study used a retrospective report of mood over the past week, which is potentially
susceptible to memory and information processing biases. Their mood measure was also
collected at the same time as the overall satisfaction measures. This may have in¯ated their results
if all variables were in¯uenced by transient mood at the time of responding (Brief et al., 1995) or
other response±response biases. This study will attempt to replicate their ®ndings without these
potential methodological problems. Further, the present research will extend Brief et al.'s study
by assessing the impact of positive and negative emotion measures as well as mood.
Hypothesis 2: Mood and emotion measures will be more strongly related to the Faces scale of
overall job satisfaction than the Job In-General Scale (Ironson et al., 1989) and the Facet-free
Job Satisfaction Scale (Quinn and Staines, 1979).
The relationship of mood and emotions to ®ve facets of job satisfaction will also be explored.
One might expect mood and emotion to be relatively strongly related to satisfaction with the
work itself for two reasons: much of the variance in overall satisfaction seems to be due to
satisfaction with the work itself (Ironson et al., 1989), and for all jobs the work is continuously
present as a potential cause of mood and emotions at work. In some jobs, coworkers and
supervisors may also be continuously present as possible triggers of emotion. Satisfaction with
pay and promotion seem by nature more calculative, comparative, and cognitive. These aspects
of a job also seem less likely to cause emotional responses as frequently as the work itself. When a
comparison regarding pay or promotion is triggered, strong emotion may be generated, but such
comparisons should be relatively infrequent. For instance, one might feel very unhappy about
one's relative pay after seeing a salary survey report, but feel pleased with an accomplishment or
frustrated at interrupted progress on a task many times each day.
Hypothesis 3: Satisfaction with the work itself will be the facet which is most strongly
correlated with mood and emotions while working. Satisfaction with pay and promotion will
be the facets least strongly related to mood and emotions while working.
While emotions easily cluster into positive and negative categories (Diener et al., 1995; Shaver
et al., 1987), there is also unique variance associated with each distinct emotion. For instance, the
negative emotion rage is quite dierent from fear, which is dierent from sadness in both causes
and eects (Shaver et al., 1987), while elation,gladness, and joy have been shown to be empirically
discriminable (de Rivera et al., 1989; see also Harrison, 1986). This suggests that additional
understanding of the dynamics of job satisfaction might be forthcoming from an exploration of
which speci®c emotions are most related to overall satisfaction. The relationship of 16 speci®c
positive and negative emotions, such as pride, happiness, anger, and frustration, with job
satisfaction will be reported. If some emotions are found to be more strongly related to job
satisfaction than others, this may suggest ways of modifying the work environment or work
processes to reduce the incidence of emotions which are most negatively related to satisfaction,
and increase the incidence of those which are most positively related.
Positive and negative aect are usually strongly inversely related at a moment in time, as people
do not feel simultaneously very happy and very unhappy, or both joyful and disgusted. However,
when positive emotions and negative emotions are aggregated over time, the relationship between
the composites is considerably weaker (Diener and Emmons, 1984; Diener et al., 1995). If
positive and negative emotion composites are relatively independent of each other, there is
opportunity for both to add unique variance to the prediction of other variables. In the case
of attitudes toward political ®gures, for instance, both positive and negative emotions add
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
188 C. D. FISHER
signi®cantly to predicting overall attitude toward the target (Ottati, 1997). This suggests that both
positive and negative emotion composites may carry useful information with respect to job
satisfaction.
Hypothesis 4: Positive and negative emotion measures will each contribute unique variance to
the prediction of overall job satisfaction.
There has been a debate in the job satisfaction literature about whether overall job satisfaction
is simply the sum of facet satisfactions. A combination of facet satisfactions generally accounts
for only about 50 per cent of the variance in overall job satisfaction (Ferratt, 1981; Highhouse
and Becker, 1993; Ironson et al., 1989). This has led some to question whether all the important
pieces of job satisfaction have been identi®ed (Scarpello and Campbell, 1983). Clearly, this paper
contends that one of the missing pieces of overall job satisfaction is aect. If facet satisfaction
ratings are by nature primarily cognitive and comparative, it is reasonable to expect aect to
account for additional variance when respondents are asked to report their overall job
satisfaction. Consistent with this idea are two studies which found that mood contributed beyond
beliefs to the prediction of some measures of overall job satisfaction (Brief and Roberson, 1989;
Weiss et al., 1999).
Hypothesis 5: Aect measures will contribute to the prediction of overall job satisfaction above
and beyond the contribution of facet measures of satisfaction.
Respondents in this study reported their mood and emotions up to 50 times each. Some
analyses will utilize mood and emotions measures averaged over time for each person. This seems
to be the most straightforward way to summarize the overall aective experience of work. But is
this the way people actually aggregate their aective experiences? Taber and Alliger (1995, p. 103)
contend that, ``There currently is no `algebra of job satisfaction' that describes how task
experiences and daily job events concatenate into feelings of job satisfaction.'' Researchers in the
area of subjective well-being have explored the relationship between moment-to-moment aect
and overall happiness, and have found that per cent of time people experience net positive aect is
much more important than the intensity of such positive aect when it is experienced (Diener
et al., 1991). In other words, those who are happiest overall are at least slightly happy most of the
time, while being extremely happy some of the time is not sucient to guarantee overall
happiness. This ®nding seems quite robust in the happiness literature, so hypothesis 6 suggests
that the same pattern will occur in predicting job satisfaction.
Hypothesis 6: Frequency of experiencing net positive emotion will be a better predictor of
overall job satisfaction than will intensity of positive emotion when it is experienced.
The hypotheses above are about state rather than trait aect. However, it is possible that trait
negative aectivity (NA) may be related to both independent variables (momentary aect) and
dependent variables ( job satisfaction) in this study (c.f. Levin and Stokes, 1989; Watson and
Clark, 1984). If so, trait aectivity could be a third variable which spuriously in¯ates relation-
ships between independent and dependent variables (Brief et al., 1988). On the other hand,
Spector et al. (2000) argue that NA `bias' is substantive rather than spurious. To address the
concern that aect±satisfaction relationships may be spurious, positive and negative aectivity
will be assessed and analyses repeated, controlling for dispositional aectivity.
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
AFFECT AND SATISFACTION 189
Method
Research participants
One-hundred-and-twenty-four employed adults from 65 organizations were recruited to part-
icipate in the study. Of these, 121 completed all aspects of the study. 73 per cent were female. The
age distribution was 17±25 years (12 per cent), 26±35 years (32 per cent), 36±45 years (31 per
cent), 46±55 years (20 per cent), 56±65 years (5 per cent). Tenure on the job averaged 4.5 years
and ranged from one month to 23 years. Many occupations were represented, including childcare
worker, hairdresser, outside salesperson, retail clerk, oce worker, supervisor, accountant,
maintenance worker, bank teller, and rehabilitation counsellor.
Procedure
This study utilized experience sampling methodology to collect frequent real-time reports of
aective experiences at work. The study was run in three stages for each participant. Stage one
was a longer questionnaire containing items on demographics and job attitudes. Stage two was a
two week period during which participants wore programmed alarm watches which rang ®ve
times each working day. They were asked to ®ll out a one page questionnaire which assessed
mood and emotions each time the alarm sounded. The watches rang at dierent times each day,
with each alarm no closer than one hour to the previous one. Stage three was another longer
questionnaire containing additional job satisfaction measures.
Each person could potentially respond to 50 alarms. The average number of responses per
person was 37, with a range from 12 to 50. A total of 4507 alarm reports were received.
Participants were instructed to respond as soon as possible after an alarm, but in any case, within
20 min of hearing the alarm. The mean time to respond was 2 min, with 70 per cent of alarms
answered immediately. Only 0.4 per cent of alarms were answered more than 30 min late. Thus,
most people were responding while the memory of their feelings at the time of the alarm was
fresh.
Measures
Emotion
There are no existing measures of emotions at work. Mood measures are inadequate for this
purpose for several reasons. First, because of the prevalence of the two dimensional mood model,
nearly all mood measures contain activation items, such as sleepy, drowsy, and dull, which are
not emotions. Second, the hedonic tone dimension may be too gross for the wide variety of
distinct positive and negative emotions that exist. Third, adjectives that imply an object are not
usually included in mood measures, as moods do not have objects. Thus, mood measures are
both contaminated and de®cient as measures of emotion.
It was necessary to construct a new instrument, the Job Emotions Scale (JES) for this study.
The starting point was Shaver et al.'s (1987) list of 135 prototypical emotion terms. Shaver et al.
obtained similarity ratings on the 135 emotions and conducted hierarchical cluster analyses. At
the most global level, the emotion words could be clustered as positive or negative. At the next
level, ®ve basic emotion categories emerged: love, joy, anger, sadness and fear. Below these ®ve
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
190 C. D. FISHER
were 25 sub-categories. For instance, under love, the sub-categories were named aection
(containing 10 terms), lust ( ®ve terms), and longing (one term).
A pilot study was undertaken to discover which of the 135 emotions were actually experienced
with reasonable frequency while at work. One-hundred-and-seventy-four university students
with work experience rated the frequency (1 never, 5 very often) with which they had felt
each of the 135 emotions while working on their present or last job. Items chosen for inclusion on
the Job Emotion Scales needed to occur reasonably frequently and cover as many of Shaver's
25 sub-categories as seemed relevant in the workplace.
The sub-categoriesÐlonging, lust, suering, relief, enthralment, torment, envy, sympathy, and
surpriseÐcontained only a few emotion words, and/or were uncommon at work so terms
belonging to these categories were excluded. If sub-categories were common, the two most
frequently experienced terms in the sub-category were considered candidates for the scale, and
usually one was retained. Factor analyses, item analyses, and rational judgment were used to
determine terms to retain. Cheerfulness and contentment were the most common positive sub-
categories. Two terms were chosen to represent eachÐhappiness and enjoyment and content-
ment and pleasureÐto increase the number of positive terms to balance the eight needed to cover
the more complex negative emotion domain. Factor analyses on the pilot data suggested one
positive emotion factor and one or two negative emotion factors in the 16 terms. Because the
structure of recalled emotion frequency may be dierent than that of real-time emotion intensity
(Diener and Emmons, 1984; Diener et al., 1995), the factor structure was reassessed on the ESM
data and is reported in the results.
The terms adopted for positive emotions were: liking for someone or something,happy,
enthusiastic,pleased,proud,optimistic,enjoying something and content. The mean frequency of
the items selected was 3.33 (more than `occasionally'). Negative adjectives were: depressed,
frustrated,angry,disgusted,unhappy,disappointed,embarrassed and worried. The mean frequency
was 2.53 (midway between seldom and occasionally). These items represent 6 of Shaver's
10 positive emotion categories and 7 of his 14 negative emotion categories. In the stage 2
questionnaire, each term was rated on a ®ve point scale on the extent to which it was being
experienced when the alarm rang. Anchors ranged from 1 not at all to 5 a great deal.
Mood
Based on the ®ndings of Weiss et al. (1999) that hedonic tone was the most important aspect of
mood, only hedonic tone was assessed. In the interest of space, a single item was used. Each time
the alarm rang, the ®rst substantive item answered by respondents was `How were you feeling as
the alarm rang? What kind of mood were you in?' They answered on an 11 point faces scale where
1 was the most unpleasant/unhappy face and 11 the most pleasant/happy face. Respondents'
average mood scores were the means of the 12 to 50 responses each made to the faces mood scale.
Satisfaction
Overall job satisfaction was measured in three ways. The Job In General Scale (Ironson et al.,
1989) is an 18 item overall job satisfaction instrument similar in format to the Job Descriptive
Index (Smith et al., 1969). It was administered in stage 3. Coecient alpha for this measure was
0.89. The Quinn and Staines (1979) Facet-free Job Satisfaction Scale was also administered at
stage 3. The ®ve items of this scale were coded as recommended by the authors, and yielded a
reliability of 0.80. An 11 point faces scale (Kunin, 1955) for rating satisfaction with the job as a
whole was administered at stages 1 and 3. These two items were averaged to form a faces measure
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
AFFECT AND SATISFACTION 191
of overall satisfaction. The intercorrelation between the items was 0.67. Facet satisfaction was
assessed with the Job Descriptive Index (Smith et al., 1969) at stage 1. Coecient alpha
reliablities were: Work Itself, 0.82; Pay, 0.83; Promotion, 0.85; Supervision, 0.86; and Coworkers,
0.88.
Aectivity
Dispositional positive and negative aectivity were measured at stage 1 with the Positive and
Negative Aect Scales (PANAS, Watson et al., 1988). Respondents were asked to report how
they generally felt, on average, in their life as a whole, not just at work. The 10 item positive aect
scale (PA) had a coecient alpha of 0.83, while the negative aect scale (NA) had a reliability of
0.85.
Analyses
There are several alternatives for analysing ESM data, and the question of how best to do it is not
entirely resolved. (See Alliger and Williams 1993; or Jaccard and Wan, 1993 for more informa-
tion). Two approaches will be employed in this study. The ®rst uses average measures of aect
over time for each individual, and assesses the relationship of these single aggregated scores for
each construct (e.g., average positive emotion experienced while at work) to job satisfaction. This
approach has an nof 121 in this study. The second approach assesses relationships between signal
level measures of aect and individual level measures of satisfaction with hierarchical linear
modeling (HLM) (Bryk and Raudenbush, 1992), a technique speci®cally designed to deal with
unbalanced multi-level data. Since between 12 and 50 aect observations are nested within each
subject, HLM is the preferred analysis strategy. In the terminology of HLM, Level 1 data are the
within-individual signal-contingent mood and emotion reports (n4465 after dropping reports
with missing data on one or more items), while Level 2 variables are individual level measures of
job satisfaction (n121). HLM partitions the variance in Level 1 data into that which is
between persons and that which is within persons. Between-persons variance is then predicted by
Level 2 variables in a random intercept regression. This approach only allows Level 1 measures to
be treated as univariate dependent variables. With HLM it is not possible to assess, for instance,
the amount of variance in job satisfaction (a Level 2 variable) which is accounted for jointly by
the independent variables of mood, positive emotion, and negative emotion (Level 1 variables).
When the hypotheses required aect to be the independent variable, the aggregated approach will
be used. When possible, both aggregated and HLM analyses will be conducted and reported.
Results
The factor structure of the 16 emotion items of the Job Emotions Scale was investigated ®rst (see
also Ha
Èrtel et al., 1998). Because the intention was to reduce the 16 variables to a smaller number
of predictors that captured much of the variance in the larger set, principal components analysis
with varimax rotation was utilized (Hair et al., 1995). Analyses were performed on the mean
ratings over time for the eight positive and eight negative emotions, and also on raw emotion
ratings for 10 of the 50 response periods, one per day of the ESM portion of the study. The results
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
192 C. D. FISHER
for the positive emotion items were absolutely consistent across all analysesÐthere was only one
positive emotion factor present, with all items loading very strongly on that factor. These items
were averaged to produce a positive emotion scale with a coecient alpha of 0.95.
As anticipated, there was some evidence that negative emotions were more complex and
dierentiated than positive emotions (Hunt, 1997). Analysis of the mean data showed two
factors accounting fro 78 per cent of the variance. The second factor contained embarrassment
and worry, while all other negative items loaded strongly on the ®rst factor. Analyses of raw
data showed that on two occasions the negative items formed a single factor, once three factors
with eigen values greater than one appeared, and the other seven analyses produced two factors,
most commonly with embarrassment and worry loading on one factor and all other items on
the other. However, embarrassment and worry often had sizeable cross loadings on the ®rst
factor, so for ease of analysis a single negative emotion scale was constructed of all eight items.
The reliability of this scale was 0.90. Descriptive statistics on all study variables appear in
Table 1.
The ®rst step in HLM is a random ANOVA on each Level 1 (aect) variable with person as the
independent variable. This showed that about 39 per cent of the variance in mood was between
persons and the remaining 61 per cent was within persons over time. Twenty-three per cent of the
variance in the negative emotion scale was between persons, with 77 per cent being within person.
About 53 per cent of the variance in positive emotion was between persons.
Hypotheses 1, 2, and 3 were tested with both aggregated and HLM analyses, with very similar
results. Correlations between aect and overall job satisfaction derived from both methods are
shown in Table 2. For the HLM analyses, these correlations were calculated as the square root of
the per cent of the between-person variance in aect which is attributable to the Level 2 predictor.
Clearly, Hypothesis 1 is supported, with all correlations between aect and satisfaction being
signi®cant. The magnitude of the correlations, however, 0.19 to 0.54 (0.22 to 0.53 in the
aggregated data), is less than would be found if aect while working was identical to job
satisfaction.
As predicted in Hypothesis 2, the correlations of aect with the faces scale of overall job
satisfaction are stronger than the correlations of aect with either the Job In General Scale or the
Quinn and Staines Facet-free Satisfaction measure. The dierences between the correlations are
statistically signi®cant. These conclusions were con®rmed by both HLM and aggregated analysis
methods.
Hypothesis 3 suggested that facet satisfactions would not all be equally strongly correlated
with real-time aect. Speci®cally, satisfaction with the work itself was expected to be most
strongly related to the aect measures, and satisfaction with pay and promotion the least strongly
related. Across the three aect measures, satisfaction with the work itself did tend to have the
highest correlations, though very few of the dierences in correlations between facets were
signi®cant (Table 3). Overall, there is little support for Hypothesis 3. The magnitude of the
correlations between aect and facet satisfactions was generally quite low.
Table 4 shows the relationships between the average experience of sixteen dierent emotions
and overall job satisfaction. The ®nding that aect is more strongly related to the faces measure of
satisfaction than more verbal measures of satisfaction holds up at the individual emotion level as
well. The signs of the correlations are as expected, with speci®c positive emotions positively
related to job satisfaction and speci®c negative emotions negatively related to job satisfaction. Of
the eight positive emotions, feeling proud and experiencing liking for something or someone have
the lowest correlations with job satisfaction, while feeling content and enthused tend to have the
highest correlations. On the negative side, feeling embarrassed and worried have the lowest
correlations with job satisfaction.
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
AFFECT AND SATISFACTION 193
Table 1. Descriptive statistics and intercorrelations
Mean S.D. 123456789101112
1. Positive aectivity 3.86 0.49
2. Negative aectivity 1.74 0.54 ÿ0.09
3. Mood 8.00 1.19 0.29{ÿ0.28{
4. Positive emotion 2.85 0.68 0.60{ÿ0.14 0.62{
5. Negative emotion 1.25 0.24 0.00 0.34{ÿ0.46{ÿ0.20*
6. Job in general 43.86 9.69 0.19* ÿ0.12 0.30{0.30{ÿ0.29{
7. Facet-free satisfaction 3.65 1.05 0.12 ÿ0.12 0.28{0.34{ÿ0.22* 0.78{
8. Faces satisfaction 8.70 1.68 ÿ0.23* ÿ0.22* 0.53{0.46{ÿ0.51{0.72{0.69{
9. JDI work 33.49 10.45 0.14 ÿ0.16 0.26{0.18* ÿ0.29{0.66{0.64{0.66{
10. JDI pay 16.63 7.14 ÿ0.18* ÿ0.01 0.09 ÿ0.11 ÿ0.19* 0.18* ÿ0.20* 0.20* 0.30{
11. JDI promotion 9.57 7.10 0.11 0.07 0.11 0.21* ÿ0.18* 0.32* 0.28{0.28{0.36{0.12
12. JDI supervision 38.74 12.35 0.00 ÿ0.07 0.17 0.08 ÿ0.32{0.25{0.20* 0.27{0.32{0.25{0.42{
13. JDI co-workers 39.58 11.52 ÿ0.01 ÿ0.22* 0.22* 0.13 ÿ0.17 0.37{0.27{0.30{0.42{0.05 0.24{0.35{
n121
*p50.05; {p50.01.
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
194 C. D. FISHER
Table 2. Correlations between Aect and Overall Satisfaction based on HLM analyses and aggregate
analyses (in parentheses)
Job in general Facet-free satisfaction Faces overall satisfaction
Positive emotion 0.29{0.33{0.45{
(0.30{) (0.34{) (0.46{)
Negative emotion ÿ0.30{ÿ0.19* ÿ0.52{
(ÿ0.29{)(ÿ0.22{)(ÿ0.51{)
Mood 0.26{0.24{0.54{
(0.30{) (0.28{) (0.53{)
*p50.05; {p50.01.
Table 3. Correlations between Aect and Facet Satisfaction based on HLM analyses and aggregate
analyses (in parentheses)
Mood Positive emotion Negative emotion
JDI work 0.22* 0.15 ÿ0.27{
(0.26{) (0.18*) (ÿ0.29{)
JDI pay 0.02 ÿ0.07 ÿ0.12
(0.09) (ÿ0.11) (ÿ0.19*)
JDI promotion 0.06 0.17* ÿ0.14
(0.11) (0.21{)(ÿ0.18*)
JDI supervision 0.07 0.00 ÿ0.28{
(0.17*) (0.07) (ÿ0.32{)
JDI coworkers 0.18* 0.09 ÿ0.18*
(0.22{) (0.13) (ÿ0.17*)
*p50.05; {p50.01.
Table 4. Correlations of speci®c emotions with overall satisfaction based on HLM analyses and aggregate
analyses (in parentheses)
Job in general Facet-free satisfaction Faces overall satisfaction
Liking 0.16* (0.18*) 0.22{(0.24{) 0.24{(0.26{)
Happy 0.27{(0.28{) 0.30{(0.31{) 0.50{(0.50{)
Enthusiastic 0.34{(0.34{) 0.39{(0.39{) 0.53 (0.52{)
Pleased 0.25{(0.26{) 0.28{(0.29{) 0.40{(0.40{)
Proud 0.11 (0.15) 0.15 (0.16*) 0.22{(0.23{)
Optimistic 0.25{(0.26{) 0.29{(0.28{) 0.37{(0.37{)
Enjoying 0.29{(0.30{) 0.37{(0.37{) 0.47{(0.46{)
Content 0.37{(0.36{) 0.35{(0.36{) 0.58{(0.57{)
Depressed ÿ0.30{(ÿ0.27{)ÿ0.20* (ÿ0.23{)ÿ0.43{(ÿ0.41{)
Frustrated ÿ0.26{(ÿ0.26{)ÿ0.22* (ÿ0.24{)ÿ0.45{(ÿ0.43{)
Angry ÿ0.33{(ÿ0.31{)ÿ0.22* (ÿ0.26{)ÿ0.47{(ÿ0.44{)
Disgusted ÿ0.26{(ÿ0.25{)ÿ0.18* (ÿ0.21*) ÿ0.46{(ÿ0.44{)
Unhappy ÿ0.35{(ÿ0.33{)ÿ0.19 (ÿ0.21*) ÿ0.49{(ÿ0.47{)
Disappointed ÿ0.23* (ÿ0.23{)ÿ0.10 (ÿ0.14) ÿ0.42{(ÿ0.41{)
Embarrassed 0.00 (ÿ0.08) 0.00 (ÿ0.02) ÿ0.27{(ÿ0.28{)
Worried ÿ0.06 (ÿ0.11) 0.00 (ÿ0.08) ÿ0.32{(ÿ0.31{)
*p50.05; {p50.01.
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
AFFECT AND SATISFACTION 195
The more weakly related emotions of pride, liking, and embarrassment were the least
frequently reported emotions in their class in the ESM study. These emotions seem to have quite
speci®c precursors (someone or something to feel liking for, an accomplishment substantial
enough to be proud of, a ga to feel embarrassed about) which may not occur often on many
jobs. Some of the emotions with stronger correlations are more mood-like, may be more easily
triggered by a wider variety of work circumstances, and may be experienced more frequently, thus
contributing more to a summary judgment of overall satisfaction. For instance, one can be
unhappy or enthused or content for any number of reasons. Alternatively, liking may be more
associated with people than jobs, while pride and embarrassment are more likely attributed to the
self than the job, so these emotions may be seen as less relevant to an evaluation of the job.
In testing all hypotheses thus far, HLM has produced very similar results to ordinary
regression on aggregated Level 1 variables, suggesting that regression on aggregates is not an
inappropriate or inaccurate technique for this data set. The remaining hypotheses will be tested
using ordinary regression on aggregate mood and emotion measures. HLM cannot be used for
Hypotheses 4 or 5 because these hypotheses require Level 1 measures to be independent variables
predicting a Level 2 dependent variable, while Hypothesis 6 is a comparison of two methods of
aggregating emotion.
Hypothesis 4 predicted that positive and negative emotion measures would each contribute
unique variance to overall job satisfaction. The correlation between the aggregated positive and
negative emotion variables was relatively small at ÿ0.20, leaving plenty of scope for each to make
a unique contribution. The hypothesis was tested with multiple regression and Darlington's
(1968) usefulness index. The latter is the reduction in R2observed when one variable is removed
from an equation in which both were previously predictors. These analyses can be seen in
Table 5. Aggregate measures of both positive and negative emotion contributed signi®cantly to
the prediction of Job in General Satisfaction and Faces Satisfaction. Positive and negative
emotions were equally useful in predicting Job in General Satisfaction, while negative emotion
was a more useful predictor of Faces Satisfaction. The Quinn and Staines Facet-free measure was
best predicted by positive emotions, with negative emotions adding only marginally (p50.07) to
the equation. Taken together, the results support the usefulness of both positive and negative
emotions in contributing uniquely to most measures of overall satisfaction, though it appears
that some measures may be dierentially sensitive to positive or negative emotions.
Hypothesis 5 suggested that mood and emotion measures would contribute to the prediction
of overall job satisfaction above and beyond facet satisfactions. This was tested with hierarchical
regression. All JDI facet measures were forced into the equation on the ®rst step, then one of the
aggregate aect measures on the second step. Results of these analyses can be seen in Table 6.
Each of the three aect measures contributed signi®cantly beyond facets to the prediction of the
Faces measure of overall satisfaction, con®rming earlier speculation that the Faces measures
contains more than just cold cognitions about aspects of the job. Average positive emotions
Table 5. Useful analyses for positive and negative emotions predicting job satisfaction
Job in general Facet-free satisfaction Faces satisfaction
BetaaUsefulness Beta Usefulness Beta Usefulness
Positive emotion 0.25{0.053 0.30{0.080 0.37{0.126
Negative emotion ÿ0.24{0.050 ÿ0.16* 0.017 ÿ0.43{0.174
Total Adj R20.13{0.12{0.37{
aStandardized Betas when both variables are in the equation.
*p50.10; {p50.01.
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
196 C. D. FISHER
contributed beyond facets to the prediction of all three satisfaction scales. Negative emotions did
not add unique variance beyond facets to the prediction of Job in General Satisfaction or Facet-
free Satisfaction, while mood was only a marginal contributor to Job in General Satisfaction. The
results are generally supportive of the hypothesis, in that an aect measure, positive emotions,
contributed above and beyond facet satisfactions to all three measures of job satisfaction.
Hypothesis 6 concerned the `algebra of satisfaction'. The hypothesis followed Diener et al.
(1991) in suggesting that the frequency of experiencing net positive emotion would be a better
predictor of overall job satisfaction than would the intensity of positive emotion when it is
experienced. Diener et al.'s (1991) methods for calculating frequency and intensity of positive
emotion were used. This entails ®rst comparing positive emotion scores and negative emotion
scores at each time period. A count of the number of times positive emotion predominates is then
divided by the number of reporting periods to indicate the per cent of time that the individual
experienced net positive aect. Finally, positive intensity is calculated as the average intensity of
positive emotions across those reports in which the person felt more positive than negative
emotion.
The frequency and intensity measures of positive emotion were correlated 0.42 with each other,
suggesting that they are capturing at least somewhat dierent phenomena. Frequency of positive
emotion correlated 0.40, 0.35 and 0.58 respectively with the Job in General, Facet-free, and Faces
satisfaction measures. Intensity correlated 0.24, 0.30 and 0.37 with the same measures. As
predicted, the frequency correlations were all larger than their respective intensity correlations. In
the case of Job in General, the frequency correlation was signi®cantly larger than the intensity
correlation at the 0.05 level, while the dierence was signi®cantly larger at the 0.01 level for the
faces satisfaction measure. On the whole, the results support the hypothesis that frequency of
positive emotion is more important for job satisfaction than intensity of such emotion.
While not directly of interest in this study, dispositional positive and negative aectivity were
measured as potential nuisance variables which might be correlated with both independent and
dependent variables. As shown in Table 1, PA and NA were weakly and in some cases non-
signi®cantly related to job satisfaction. Positive aectivity was strongly related to reported
positive emotions and moderately related to average mood. Negative aectivity was moderately
related to both negative emotions and average mood. Therefore, all aggregated analyses in the
study were repeated using PA and NA as control variables. The results were quite similar to those
reported aboveÐoccasionally weaker but still signi®cantÐsuggesting that the relationships
between aggregated state aect and satisfaction are much more than simply eects of chronic
aectivity.
Table 6. Hierarchical regressions of facet satisfaction and aect measures on overall job satisfaction
Job in General Facet-free Satisfaction Faces Satisfaction
Adj R2Signi®cance
of change R2
Adj R2Signi®cance
of change R2
Adj R2Signi®cance
of change R2
Step 1
JDI Facets 0.434 0.393 0.411
Step 2
Mood 0.446 0.07 0.401 n.s. 0.552 0.001
Step 2
Positive emotions 0.460 0.05 0.439 0.01 0.532 0.001
Step 2
Negative emotions 0.442 n.s. 0.389 n.s. 0.520 0.001
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
AFFECT AND SATISFACTION 197
Discussion
Hypothesis 1, that measures of real-time mood and emotion at work would be related to job
satisfaction, was supported, with mood, positive emotions, and negative emotions all being
signi®cantly related to each of the three measures of job satisfaction. Hypothesis 4, that positive
and negative emotion measures would each contribute unique variance to the prediction of overall
job satisfaction, was also supported. Analyses for Hypothesis 5 showed that aect, particularly
positive emotion, contributed beyond facet satisfaction to the prediction of overall satisfaction.
These hypotheses bear on the question posed in the title: Is aect a missing piece of job
satisfaction? The results allow a quali®ed answer of yes to be given to this questionÐreal-time
aect certainly is related to job satisfaction, and accounts for some variance beyond facet
satisfactions. However, it is very clear from the size of the correlations that aect while working is
not equivalent to job satisfaction as the latter is typically measured. if job satisfaction is supposed
to be `an aective response to the job', why aren't satisfaction±aect correlations stronger?
Weiss (2000) suggests that we have failed to distinguish between three dierent types of
reactions to the job: aective reactions, cognitive beliefs and overall evaluative judgments. This
argument is based on recent research on attitudes showing that the cognitive and aective
components are empirically as well as conceptually distinct (c.f. Crites et al., 1994), and that both
contribute independently to the overall judgment. Organizational researchers have acted as if job
satisfaction measures capture overall evaluative judgments, when in fact they focus more on
cognitions and less on aect. This may account for the relatively weak correlations between aect
measures and some of the satisfaction measures in this study.
It may be useful to think of job satisfaction instruments as lying on a continuum from
primarily assessing the cognitive component of job attitudes to assessing a combination of
cognitive and aective components which is closer to Weiss' concept of an overall evaluative
judgment. An examination of the ®ndings of this study and the items on the job satisfaction
instruments themselves leads to similar conclusions about where on this continuum various scales
might lie. The JDI facet scales seem close to the cognition end of the continuum. The facet scales
were found to be relatively independent of aect experienced while working (average correlation
with aect measures 0.18), and the items ask for cold cognitive decisions about whether
phrases like `uncomfortable, useful, respected', describe one's work, or `asks my advice,
intelligent, poor planner, or up-to-date' describe one's supervisor. The JIG and Quinn and
Staines Facet Free measures may be further along the continuum toward overall attitude, but
also have substantial cognitive components. The former conclusion is buttressed by the ®nding
that at least one aect variable contributes beyond facets to the prediction of both JIG and Facet
Free satisfaction. The average correlation of JIG and Facet Free satisfaction with the real-time
aect measures is 0.29. Turning to an inspection of the items, the JIG asks whether the job is
`ideal, a waste of time, acceptable, worse than most, disagreeable, good, bad, pleasant, rotten'
etc., while the Facet Free measure asks whether the incumbent would take the same job again,
would recommend the job to a friend, the extent to which the job measures up to what one
wanted when one took the job, and the extent to which one is satis®ed with the job. Answering
these items seems to require quite a lot of cognition and comparison to standards. As predicted in
Hypothesis 2, the Faces scale comes closest to an overall evaluative judgment, in that it taps both
components to a reasonable degree, correlating fairly strongly with the three aect measures
(average correlation 0.50) and the other two overall job satisfaction scales (average correlation
0.71). Brief and Roberson (1989, p. 723) also concluded that `the Faces measure is the most
balanced . . . in terms of capturing positive and negative aect and cognitions.'
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
198 C. D. FISHER
These suggestions have implications for the measurement of job attitudes. It is clear that most
measures of job satisfaction assess aect only to a limited degree. Rather than aect being a
missing piece of job satisfaction, it seems more likely that aect, together with job satisfaction,
are both pieces of a missing constructÐoverall evaluative judgment. We should either develop
new measures of `overall evaluative judgment' that incorporate both aective and cognitive
content, or develop stand-alone measures of overall evaluation and of job aect to use in parallel
with existing (largely cognitive) measures of job satisfaction.
What might be the bene®ts of assessing job aect as a separate construct from job satisfaction?
One might suggest that mood and emotions while working represent true `quality of work life',
and as such deserve to be studied in their own right. As suggested by Weiss and Cropanzano
(1996), real-time aective experiences may be one of the mechanisms by which work context
features (such as job design or superior's leadership style) and individual dierences (such as
negative aectivity) eventually in¯uence cognitions about the job and subsequent judgment-
driven behaviors such as quitting. Recent research in social psychology suggests that the aective
component of attitudes is sometimes more useful in predicting behavior than the cognitive
component (c.f. Bohm and P®ster, 1996; Sappington, 1990). This is particularly true for con-
summatory rather than instrumental behavior (Millar and Millar, 1996; Millar and Tesser, 1989).
Weiss and Cropanzano (1996) suggest that some work behaviors are directly driven by aect.
These may include impulsive `consummatory' behaviors such as eort exerted at a given moment,
choosing to be absent on a given day, or deciding whether or not to perform an organizational
citizenship behavior. Van Katwyk et al. (1995) have suggested that aect is likely to be more
closely related to phenomena like stress, strain, and burnout than are cold cognitive measures.
Isen and Baron (1991) provide a further review of the possible impact of aect on work behavior.
In sum, there is good reason to pursue the measurement of job aect as a separate construct
which is related to, but not isomorphic with, job satisfaction.
There are a number of possibilities for measuring job aect, and research will be needed on
which are most eective and construct valid. One option would be signal-contingent experience
sampling of mood and emotion while working, as was done in the present study, though this
method is cumbersome and intrusive. A slightly less intrusive method would assess aect via a
once daily diary. Clearly, experience sampling does get at real-time aect, uncontaminated by
memory and recall biases. As such, perhaps ESM-based measures should provide the criterion
against which easier-to-use single-administration scales of job aect are validated. Such scales
might follow the example of Crites et al. (1994) who selected aectively toned adjective pairs such
as joy/sorrow, love/hateful, and delighted/sad to assess the aective component of attitudes. The
present study oers the Job Emotions Scale, which might be useful for rating the frequency with
which each emotion has been experienced during the past day, week, or longer period. Burke
et al.'s (1989) Job Aect Scale (JAS) provides a retrospective measure of mood (but not
emotions) at work. Van Katwyk et al. (1995) describe the development of the Job-Related
Aective Well-Being Scale (JAWS), which asks for a rating of how often the job has made one
feel each of 30 mood and emotion states such as annoyed, bored, angry, proud, happy, calm, and
so on. Like the JAS, the JAWS was built around a two dimensional model of pleasure/displeasure
and high/low arousal. Clearly, there is more to be done in terms of developing measures of job
aect, and in determining their utility in predicting important work behaviors.
The ®ndings of this study may have practical implications for organizations seeking to improve
satisfaction and quality of work life. The consistent and signi®cant correlations found between
aect measures and job satisfaction, together with the attitude literature conclusion that aect
in¯uences overall evaluations independently of beliefs, suggest that eorts to improve moods
and emotions at work may pay o in better job attitudes. While moods may not be directly
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
AFFECT AND SATISFACTION 199
controllable given their somewhat vague and diuse causes, events that provoke speci®c positive
and negative emotions should be more amenable to organizational intervention. The support for
Hypothesis 4, that both positive and negative emotions account for unique variance in satisfac-
tion, suggests a two-pronged approach of reducing events that provoke negative emotions and
increasing events that cause positive emotions.
The ®nding that it is the per cent of time that one feels net positive aect at work, more so than
the intensity of that aect that matters for satisfaction (Hypothesis 6), also suggests that
employers should concentrate on providing a work environment free of the minor irritations and
hassles which tip the balance toward more frequent, if mild, negative aect (Kanner et al., 1981).
They might also build in small frequent positive reinforcements or `uplifts', perhaps through job
design and informal reward systems, rather than relying on possibly more intense but less
frequent positive emotions created by formal rewards, promotions, or public celebrations, in
order to enhance job attitudes.
In conclusion, aect while working is not the same as job satisfaction, though the two are
modestly correlated. There is a need to develop clear measures of job aect, job cognitions, and
overall job evaluation, and to better understand the relationship of these more pure measures to
commonly used indices of job satisfaction. Breaking down reactions to jobs in these ways may
facilitate our understanding of the mechanisms by which job characteristics in¯uence job
attitudes and quality of work life, job stressors aect health, and employees choose to perform
judgment-driven and aect-driven behaviors.
References
Alliger GM, Williams KJ. (1993). Using signal-contingent experience sampling methodology to study work
in the ®eld: a discussion and illustration examining task perceptions and mood. Personnel Psychology 46:
525±549.
Averill JR. (1975). A semantic atlas of emotional concepts. JSAS: Catalog of Selected Documents in
Psychology 5: 330, ms. no. 421.
Bohm G, P®ster HR. (1996). Instrumental or emotional evaluations: what determines preferences? Acta
Psychologica 93: 135±148.
Breckler SJ, Wiggins EC. (1989). Aect versus evaluation in the structure of attitudes. Journal of
Experimental Social Psychology 25: 253±271.
Brief AP. (1998). Attitudes In and Around Organizations. Sage Publications: Thousand Oaks, CA.
Brief AP, Burke MJ, George GM, Robinson BS, Webster J. (1988). Should negative aectivity remain an
unmeasured variable in the study of job stress? Journal of Applied Psychology 73: 193±198.
Brief AP, Butcher AH, Roberson L. (1995). Cookies, disposition, and job attitudes: The eects of positive
mood-inducing events and negative aectivity on job satisfaction in a ®eld experiment. Organizational
Behavior and Human Decision Processes 62: 55±62.
Brief AP, Roberson L. (1989). Job attitude organization: an exploratory study. Journal of Applied Social
Psychology 19: 717±727.
Bryk AS, Raudenbush SW. (1992). Hierarchical Linear Models: Applications and Data Analysis Methods.
Sage Publications: Newbury Park, CA.
Burke MJ, Brief AP, George JM, Roberson L, Webster J. (1989). Measuring aect at work: con®rmatory
analyses of competing mood structures with conceptual linkage to cortical regulatory systems. Journal of
Personality and Social Psychology 57: 1091±1102.
Cranny CJ, Smith PC, Stone EF. (Eds) (1992). Job Satisfaction: Advances in Research and Applications. The
Free Press: New York.
Crites SL Jr, Fabrigar LR, Petty RE. (1994). Measuring the aective and cognitive properties of attitudes:
conceptual and methodological issues. Personality and Social Psychology Bulletin 20: 619±634.
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
200 C. D. FISHER
Darlington RB. (1968). Multiple regression in psychological research and practice. Psychological Bulletin
69: 161±182.
de Rivera J, Possell L, Verette JA, Weiner B. (1989). Distinguishing elation, gladness, and joy. Journal of
Personality and Social Psychology 57: 1015±1023.
Diener E, Emmons RA. (1984). The independence of positive and negative aect. Journal of Personality
and Social Psychology 47: 1105±1117.
Diener E, Sandvik E, Pavot W. (1991). Happiness is the frequency, not the intensity, of positive versus
negative aect. In: Subjective Well-Being, Strack F, Argyle M, Schwarz N (eds). Pergamon Press: Oxford.
Diener E, Smith H, Fujita F. (1995). The personality structure of aect. Journal of Personality and Social
Psychology 69: 130±141.
Eagly AH, Chaiken S. (1993). The Psychology of Attitudes. Harcourt Brace Jovanovich: Fort Worth, TX.
Ferratt TW. (1981). Overall job satisfaction: is it a linear function of facet satisfaction? Human Relations 34:
463±473.
Frijda NH. (1993). Moods, emotion episodes, and emotions. In: Handbook of Emotions, Lewis M,
Haviland IM (eds); Guilford: New York; 381±403.
Hair JF Jr, Anderson RE, Tatham RL, Black WC. (1995). Multivariate Data Analysis with Readings.
Prentice-Hall: Englewood Clis, NJ.
Harrison RH. (1986). The grouping of aect terms according to the situations that elicit them: a test of a
cognitive theory of emotion. Journal of Research in Personality 20: 252±266.
Ha
Èrtel CEJ, Fisher CD, Ha
Èrtel GF. (1998). Measuring work emotions in real time. Paper presented in the
symposium `Emotion at work: New research directions' at the Society of Australasian Social
Psychologists, Christchurch, New Zealand, April 16.
Hedges SM, Jandorf L, Stone AA. (1985). Meaning of daily mood assessments. Journal of Personality and
Social Psychology 48: 428±434.
Highhouse S, Becker AS. (1993). Facet measures and global job satisfaction. Journal of Business and
Psychology 8: 117±127.
Hippler HJ, Schwarz N, Sudman S. (1987). Social Information Processing and Survey Methodology.
Springer-Verlag: New York.
Hormuth SE. (1986). The sampling of experiences in situ.Journal of Personality 54; 262±293.
Hunt. (1992). Unpublished doctoral dissertation. University of North Carolina, Chapel Hill.
Ironson GH, Smith PC, Brannick MT, Gibson WM, Paul KB. (1989). Construction of a Job in General
Scale: a comparison of global, composite, and speci®c measures. Journal of Applied Psychology 74:
193±200.
Isen AM, Baron RA. (1991). Positive aect as a factor in organizational behavior. Research in
Organizational Behavior 13: 1±53.
Jaccard J, Wan CK. (1993). Statistical analysis of temporal data with many observations: issues for
behavioral medicine data. Annals of Behavioural Medicine 15: 41±50.
Kanner AD, Coyne JC, Schaefer C, Lazarus RS. (1981). Comparison of two modes of stress measurement:
daily hassles and uplift versus major life events. Journal of Behavioral Medicine 4: 1±39.
Kunin T. (1995). The construction of a new type of attitude measure. Personnel Psychology 9: 65±78.
Larsen RJ, Diener E. (1992). Promises and problems with the circumplex model of emotions. Review of
Personality and Social Psychology 13: 25±29.
Larson R, Csikszentmihalyi M. (1983). The experience sampling method. In: Naturalistic Approaches to
Studying Social Interaction. New Directions for Methodology of Social and Behavioral Science, Reis HT
(ed.); 15: Jossey-Bass: San Francisco; 41±56.
Levin I, Stokes JP. (1989). Dispositional approach to job satisfaction: role of negative aectivity. Journal of
Applied Psychology 74: 752±758.
Locke EA. (1976). The nature and causes of job satisfaction. In: Handbook of Industrial and Organizational
Psychology, Dunnette MD (ed.); Rand-McNally: Chicago; 1297±1349.
Millar MG, Millar KU. (1996). The eects of direct and indirect experience on aective and
cognitive responses and the attitude±behavior relation. Journal of Experimental Social Psychology 32:
561±579.
Millar MG, Tesser A. (1986). Eects of aective and cognitive focus on the attitude±behavior relation.
Journal of Personality and Social Psychology 51: 270±276.
Organ DW, Near JP. (1985). Cognition versus aect in measures of job satisfaction. International Journal of
Psychology 20: 241±253.
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
AFFECT AND SATISFACTION 201
Ortony A, Clore GL, Foss MA (1987). The referential structure of the aective lexicon. Cognitive Science
11: 361±384.
Ottati V. (1997). When the survey question directs retrieval: implications for assessing the cognitive and
aective predictors of global evaluation. European Journal of Social Psychology 27: 1±21.
Pekrun R, Frese M. (1992). Emotions in work and achievement. International Review of Industrial and
Organizational Psychology 7: 153±200.
Plutchik R. (1994). The Psychology and Biology of Emotion. Harper-Collins: New York.
Porac JF. (1987). The job satisfaction questionnaire as a cognitive event: ®rst- and second-order processes
in aective commentary. Research in Personnel and Human Resource Management 5: 51±102.
Quinn RP, Staines GL (1979). The 1997 Quality of Employment Survey. Institute for Social Research,
University of Michigan: Ann Arbor, Michigan.
Rice RW, McFarlin DB, Bennett DE. (1989). Standards of comparison and job satisfaction. Journal of
Applied Psychology 74: 591±598.
Russell JA. (1978). Evidence of convergent validity on the dimensions of aect. Journal of Personality and
Social Psychology 36: 1152±1168.
Russell JA. (1980). A circumplex model of aect Journal of Personality and Social Psychology 39:
1161±1178.
Sandelands LE. (1988). The concept of work feeling. Journal for the Theory of Social Behavior 18: 437±457.
Sappington AA. (1990). The independent manipulation of intellectually and emotionally based beliefs.
Journal of Research in Personality 24: 487±509.
Scarpello V and Campbell JP. (1983). Job satisfaction: are all the parts there? Personnel Psychology 36:
577±600.
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.
Smith PC, Kendall LM, Hulin CL. (1969). The Measurement of Satisfaction in Work and Retirement. Rand-
McNally: Chicago.
Spector PE, Zapf D, Chen PY, Frese M. (2000). Why negative aectivity should not be controlled in job
stress research: Don't throw out the baby with the bath water. Journal of Organizational Behavior 21:
79±95.
Staw B. (1984). Organizational behavior: a review and reformulation of the ®eld's outcome variables.
Annual Review of Psychology 35: 627±666.
Taber TD, Alliger GM. (1995). A task-level assessment of job satisfaction. Journal of Organizational
Behavior 16: 101±121.
Van Katwyk PT, Fox S, Spector PE, Kelloway EK. (1995). Determining the cognitive structure of job-
related aective well-being: how does my job make me feel? Annual Meeting of the Society for Industrial
and Organizational Psychology, May 18±21, Orlando.
Warr P. (1990). The measurement of well-being and other aspects of mental health. Journal of Occupational
Psychology 63: 193±210.
Watson D, Clark LA. (1984). Negative aectivity: the disposition to experience aversive emotional states.
Psychological Bulletin 96: 465±490.
Watson D, Clark LA, Tellegen A. (1988). Development of brief measures of positive and negative aect: the
PANAS scale. Journal of Personality and Social Psychology 54: 1063± 1070.
Watson D, Tellegen A. (1985). Toward a consensual structure of mood. Psychological Bulletin 98: 219±202.
Weiss HM. (2000). Deconstructing job satisfaction: separating evaluations, beliefs, and aective experi-
ences. Human Resource Management Review; in press.
Weiss HM, Cropanzano R. (1996). Aective events theory: a theoretical discussion of the structure, causes
and consequences of aective experiences at work. Research in Organizational Behavior 8: 1±74.
Weiss DJ, Dawis RV, England GW, Lofquist LH. (1967). Manual for the Minnesota Satisfaction
Questionnaire. University of Minnesota Press: Minneapolis, MN.
Weiss HM, Nicholas JP, Daus CS. (1999). An examination of the joint eects of aective experiences and
job beliefs on job satisfaction and variations in aective experiences over time. Organizational Behavior
and Human Decision Processes 78: 1±24.
Wheeler L, Reis HT. (1991). Self-recording of everyday life events: origins, types, and uses. Journal of
Personality 59: 339±354.
Wright TA, Bonett DG. (1996). The role of activation and pleasantness-based aect in performance
prediction. Presented at the Academy of Management Annual Meeting, August, Cincinnati.
Copyright #2000 John Wiley & Sons, Ltd. J. Organiz. Behav. 21, 185±202 (2000)
202 C. D. FISHER
