The Relationship In Experience Between Various Types of Affect
Journal of Personality and Social Psychology
1986, Vol. 50, No. 5, 1031-1038
Copyright 1986 by the American Psychological Association, Inc.
0022-3514/86/S00.75
The Relationship In Experience Between Various Types of Affect
Ed Diener
University of Illinois
Ashgar Iran-Nejad
University of Michigan
The co-occurence in experience of various emotions was explored, with the focus on positive versus
negative affect. In Study 1,72 subjects read stories designed to produce varying levels of either positive
or negative affect, and then rated their level of both types of affect. In Study 2, 42 participants rated
their feelings during emotional times in everyday life for a period of 6 weeks. We found that emotions
of the same hedonic valence (e.g., fear and anger) tend to co-occur. We also found that positive and
negative affect do not occur together at high levels of intensity. These two facts about the relation of
positive and negative affect are probably responsible for the bipolarity that is often found between
them. These findings represent a challenge to those who postulate that there are unrelated discrete
emotions and that the terms positive affect and negative affect do not describe meaningful clusters of
emotions. We found that if one type of affect is of low intensity, the other type can be at any level
from low to high. Therefore, a truly inverse and linear relation does not characterize positive and
negative affect. This finding represents a challenge to most structural models of emotion. It appears
that mutual exclusion only at high levels of intensity characterizes the relation between positive and
negative affect.
We explore the relation between feeling particular emotions
and experiencing other emotions. For example, if one feels joy,
what is the likelihood of also feeling anger or fear? In recent years
a great deal of discussion has focused on the structure of emotion
(e.g., Daly, Lancee, & Polivy, 1983; Plutchik, 1980; Russell, 1978,
1979). In addition to the structure of momentary emotion, Diener
and his colleagues (Diener, 1984; Diener, & Emmons, 1984;
Diener, Emmons, & Sandvik, 1985; Diener, Larsen, Levine, &
Emmons, 1985) have also examined the long-term structure of
affect across persons. One necessary component in understanding
the structure of affect is to examine how compatible the experience
of varying emotions is. Are certain emotions compatible
and others mutually exclusive? When the levels of an emotion
change, do other emotions also change in intensity? For example,
as negative affect increases, is there a necessary decrease in positive
affect? When an emotion is felt, is the intensity related in
a linear or nonlinear way to other emotions? Gaining knowledge
of the interrelations of emotions as they are experienced will aid
in understanding the structure of emotion, as well as the processes
underlying affect. If particular emotions facilitate the experiencing
of certain other emotions but inhibit yet other emotions, this
will provide valuable insights into how the affect system functions.
In order to narrow the range of the questions asked, we focus
on global positive and negative affect rather than on discrete
emotions such as anger. There were several reasons why focusing
on only global positive and negative affect was desirable. First,
a two-dimensional system of emotion accounts for much of the
variance in affect and has been most replicable (e.g., Russell,
1978, 1979; Schlosberg, 1954; Watson, Clark, &Tellegen, 1984;
Watson & Tellegen, 1985; Zevon & Tellegen, 1982). The two
dimensions are either positive-negative affect and arousal (e.g.,
Correspondence concerning this article should be addressed to Ed
Diener, Department of Psychology, University of Illinois, 603 E. Daniel,
Champaign, Illinois 61820.
Russell, 1978) or a similar system that appears to be a 45° rotation
of the other system, with negative and positive affect as
the two dimensions (see Watson & Tellegen, 1985). In contrast
to the ubiquitous bipolar pleasant-unpleasant and arousal dimensions,
factors of multiple discrete emotions (e.g., fear and
joy) tend not to entirely replicate across investigators (e.g., Clyde,
1963; Ekman, Friesen, & Ellsworth, 1972; Izard, 1972; McNair
& Lorr, 1964; Nowlis & Nowlis, 1956; Plutchik, 1980). Nor has
there been agreement when dimensions in addition to positivenegative
affect and arousal are proposed. A second reason for
focusing on positive and negative affect is to simplify this initial
analysis of the covariation in the experiencing of different emotions.
Zevon and Tellegen (1982) explored the interrelation of positive
and negative affect through within-subject factor analyses. They
found two independent factors, negative and positive affect, which
replicated across virtually all subjects. Diener and Emmons
(1984) also explored the experience of positive and negative affect
within individuals. They found that positive and negative affect
correlated highly inversely when the experience of affect was
reported at particular moments. However, they found that when
longer time intervals were examined, the experiencing of positive
and negative affect tended to be uncorrelated across individuals.
Diener, Emmons, and Sandvik (1985) reviewed the evidence for
the existence of orthogonal negative and positive affect dimensions
when long-term mood was considered. On the other hand,
Russell (1978) found that positive and negative emotions were
not independent, but were at opposite ends of a bipolar continuum.
In this article we intend to extend the findings that in the
earlier studies were based solely on correlational approaches. In
particular, we first examine whether differences in the level of
one type of affect are invariably accompanied by differing levels
of intensity in the other type of affect. It is possible that positive
and negative affect correlate only because they are incompatible
at high levels of intensity. However, this would not necessarily
mean that there was a close link between the two at other less
1031
1032 ED DIENER AND ASHGAR IRAN-NEJAD
intense levels. It might be that rather than show an inverse relation,
as suggested by a negative correlation, or no correlation,
as suggested by independent dimensions, positive and negative
affect may be mutually exclusive; that is, positive affect may be
experienced intensely only when negative affect is not present.
However, it may be that the degree of positive affect can vary
from low to high, as long as the level of negative affect is low. In
addition, it may be that an increase or decrease in positive affect
does not necessitate a change in negative affect as long as the
latter is experienced at relatively low levels. Such a relation would
appear as a simple moderate negative correlation unless one made
a finer grained analysis of emotion reports. Thus the relation
between positive and negative affect may be more complex and
not completely captured by correlational or factor analytic approaches.
A closely related issue is whether positive and negative
affect are related in a truly linear way. The conclusions of Diener
and Emmons (1984) and Zevon and Tellegen (1982) were based
on an exploration of linear relations through Pearson's productmoment
correlations. However, the idea of mutual exclusion
suggests that a nonlinear relation may exist. Such a relation would
produce an inverse correlation if it were analyzed via the Pearson
statistic, although this correlation would hide an underlying curvilinear
relation.
In summary, this article is a further exploration of the relation
of the experience of positive and negative affect. In earlier studies
researchers relied on correlational approaches. We examined
whether there is a necessary inverse relation between positive
and negative affect. If the two are mutually incompatible but not
strictly inverse, it is necessary to use approaches beyond linear
correlation to uncover this fact. The major question is whether
levels of positive affect are invariably and linearly related to levels
of negative affect. A mutual exclusion view suggests that this will
not be the case. A related issue is whether there are patterns of
co-occurrence for various emotions. If emotions are discrete and
independent, they should not show specific patterns of co-occurrence,
and categories such as "positive affect" and "negative
affect" would have little scientific value. However, if certain emotions
tend to be experienced together or are mutually exclusive,
then categories or dimensions of emotions take on scientific import.
In order to explore the relations between various types of affect
in depth, several approaches were used. In Study 1, an experimental
manipulation was used to produce lower or higher levels
of positive and of negative affect. Thus the pattern of correlations
between the two could be explored both within and across conditions.
If individuals' levels of affect correlated only when both
positive and negative conditions were combined, not when the
conditions were examined separately, it would suggest not that
the two types of affect are invariably linked, but rather that they
are only inversely correlated when strong levels of affect are considered
because the two types of affect tend not to co-occur at
intense levels. In addition, the means in each condition can also
be considered determinants of whether a lower level of one type
of affect necessitates a higher level of the other type of affect. In
the second study, subjects reported their affect in everyday situations
at times over a 6-week period when they felt "emotional."
We explored the relation between negative and positive affect in
these reports in several ways. Both linear and nonlinear correlations
were computed. An analysis of the variance in one type
of affect at each level of the other type of affect was also explored.
In this way, we could determine whether the variability in each
type of affect is greatest at low levels of the other type of affect.
If so, this would suggest that there is not a simple inverse relation
in the experience of positive and negative affect, but rather a
more complex relation between the two.
In addition to the in-depth analysis of the relation between
positive and negative affect just described, we also present in the
second study a preliminary analysis of the relation between discrete
emotions such as fear and anger. The major question was
whether emotions of the same hedonic type tend to co-occur. If
so, this would have clear implications for a comprehensive theory
of emotions. If emotions of the same hedonic tone tend to cooccur,
it would indicate that the underlying processes leading to
these emotions in some way link them together.
Study 1
Method
Participants and procedures. Seventy-two college students served as
subjects. These individuals were involved in a long-term study on daily
mood, but were also requested to complete additional tasks. In this study
each subject read two emotional stories, one negative and one positive.
We manipulated intensity of mood with the "surprisingness" of the story
ending, which has proved to be an acceptable method for manipulating
degree of affective arousal (Iran-Nejad, 1983). After reading the stories,
subjects completed an affect questionnaire.
There were two stories in both the good and bad conditions. The first
story, adapted from Thurmond (1978), described an incident in which
a woman was concerned about running out of gas late at night. After
pulling into a service station, Marilyn is asked to come inside by the
attendant. Because a maniac murderer is on the loose in the area, she is
reluctant to go inside, but does so when he insists. In the negative condition,
the gas station attendant strangles her. In the positive condition,
it turns out that the attendant is really a good Samaritan intending to
save her from a man whom he has spotted hiding on the back seat of her
car. In the second story, adapted from Laymon (1975), the main character
is awakened by footsteps outside his tent. He discovers that Sally is missing
and does not respond to his calls. When he comes out of the tent, there
is a man waiting with a gun in one hand and an ax in the other. In the
negative condition, the camper is murdered. In the positive condition,
the bad man is overpowered and arrested. Each subject read the good
version of one of the stories and the bad version of the other story. Because
each subject read both a good and a bad story, subjects were repeated
across the positive-negative manipulation, but were nested within level
of surprise; that is, each subject was in a particular condition for surprise
in both the positive and negative conditions.
For each of the story conditions, there were three levels of surprisingness
that were designed to manipulate cognitive arousal and therefore emotional
intensity. The information given in the high surprise endings gave
no indication of the story ending. For example, in the negative gas station
story, the attendant was portrayed as a very nice fellow, so that his murder
of the woman is unexpected. Similarly, in the positive condition of the
camping story, Sally turns out to be the camper's dog who attacks the
bad man. In each high surprise condition, the story ending is very unpredictable
from the information contained in the story. In the medium
surprise condition, hints are included throughout the story about the
ending. In the low surprise condition, the ending of the story is obvious
from information that is directly stated in the story.
Mood instrument. After reading each story, subjects completed a
questionnaire. There was one question per page and these questions were
systematically varied in order across subjects so as to eliminate order
effects. Each question was rated on an 8-point scale anchored by Not at
NEGATIVE AND POSITIVE AFFECT 1033
all (0) and Extremely (7). We also included four irrelevant
adjective questions (lonely, ambitious, toothache, and sweet taste) in order
to see whether the manipulations affected them and to examine their
relation with emotions. We included four questions to assess levels of
positive and negative affect: pleasant, unpleasant, happy, and unhappy.
For example, one question read "The conclusion of the story I just read
made me experience:"; the response options ranged from 0 (No unpleasantness
at all) to 7 (Extreme unpleasantness).
Results
Pearson correlations were computed between the affect adjectives
in order to determine the relation of positive and negative
affect across individuals. The correlations were computed for the
72 subjects across all conditions and also separately for the positive
and negative conditions. These correlations are shown in
Table 1. Several things are clear in this table. First, adjectives
correlate much more strongly with those of the same hedonic
type than with those of the opposite hedonic value. This occurs
in all three sets of correlations. Second, across all conditions,
positive and negative affect vary inversely at moderate levels.
This indicates that when one considers conditions in which radically
different levels of positive and negative affect are present,
an inverse correlation will be found between the two. In other
words, when one examines groups in which some persons are
experiencing high-intensity positive affect and other persons are
experiencing high-intensity negative affect, then across all individuals
an inverse correlation between positive and negative affect
will emerge. However, when one turns to the correlations within
the positive and negative affect cells, one is examining the variations
in affect among individuals who are experiencing varying
degrees of positive affect in the positive condition and varying
degrees of negative affect in the negative condition. In this case,
the correlations between positive and negative affect drop to
nonsignificant values close to zero. This indicates that differences
in positive affect between individuals when they are happy are
not necessarily related to corresponding differences in negative
affect. The same pattern is true when negative affect predominates.
These individual difference correlations support the conclusion
that differences in positive affect are not always related
to differences in negative affect. In other words, there is not a
necessary inverse relation between the two. However, if one considers
a broad spectrum of moments in which at times negative
affect is strong and at times positive affect is strong, an inverse
correlation between the two will be found.
One definitely cannot attribute the low correlation between
positive and negative affect to restriction of range or to floor
effects. If the nondominant affect was restricted in range or had
reached a floor, it could not possibly correlate at moderate to
high levels with adjectives of the same type (rs = .65 and .44).
These correlations indicate that sizable relations are still possible
with the variance that exists for the nondominant affect. In addition,
as the means in Table 2 show, the mood adjectives had a
long way to move before encountering the floor effect that is
clearly evident for "Toothache."
The results revealed that in some cases the stories were not
effective in producing differences in affect. In order to assess how
average differences across individuals of one type of affect relate
to differences in the other type of affect, it was necessary to compare
stories that differed significantly in level of affect produced.
Table 1
Correlations of Mood Adjectives
Mood adjectives
All conditions
Happy
Pleasant
Unhappy
Unpleasant
Positive condition only
Happy
Pleasant
Unhappy
Unpleasant
Negative condition only
Happy
Pleasant
Unhappy
Unpleasant
Happy
.77
-.38
-.45
.77
.07
-.08
.44
-.05
-.15
Pleasant
-.33
-.39
.07
-.08
-.09
-.11
Unhappy
.80
.65
.74
Note. For rs(70) of .20 and greater, p < .05; for rs(70) of .28 or greater,
Therefore, the low- and medium-surprise stories were combined
because they produced very similar levels of affect for both stories.
In addition, only one positive story (about the gas station) and
one negative story (about the camper) were retained because the
other two stories showed no mean differences for either positive
or negative affect between any of the levels of surprise. Thus after
deleting the ineffective manipulations, we had a 2 X 2 design
(low/medium vs. high surprise and positive vs. negative ending).
Dropping of the ineffective stories occurred without reference
to the data that would shed light on the various questions under
study.
The check on the surprisingness manipulation produced highly
significant results, F(, 34) = 31.76, p < .001, which did not
interact with the positive versus negative story conditions. The
means for the four cells are shown in Table 2. We computed
preplanned orthogonal comparisons to compare the predictable
and unpredictable cells within the positive and negative conditions
(for all questions). The significance levels of these comparisons
are indicated in the table by the letters a and b. The
level of surprisingness was effective in manipulating level of affect
in both positive and negative conditions.
The key findings in Table 2 are those for the mood adjectives.
The positive stories differed significantly between the predictable
and unpredictable cells in terms of Happy and Pleasant. However,
not only did the unhappy and unpleasant ratings not differ significantly,
but the means are ordered in the same direction as
the positive adjectives. In the negative condition, there is the
same type of pattern, although the significance levels for Unhappy
and Unpleasant were of only borderline value. Once again, the
opposite adjectives (Happy and Pleasant) were almost identical
and, in addition, went in the same direction as the negative adjectives.
What these results clearly show is that positive and negative
affect can be felt at varying levels without necessarily influencing
the value of the opposite type of affect. For example,
greater positive affect in the unpredictable positive condition was
clearly not accompanied by lesser negative affect. If a person has
high affect of one type, this affect can decrease without the other
type of affects' necessarily increasing.
1034 ED DIENER AND ASHGAR IRAN-NEJAD
Table 2
Mean Ratings of Stories
Story type
Positive Negative
Questions
Predictable
Unpredictable
Predictable
Unpredictable
Surprise 2.29" 6.08" 5.83"
Irrelevant
adjectives
Lonely
Ambitious
Toothache
Sweet taste
Mood
adjectives
Happy
Unhappy
Pleasant
Unpleasant
1.92
2.46
0.00
0.54
2.67"
1.42
3.21b
2.00
1.58
2.67
0.17
0.67
5.42"
1.67
5.33"
2.25
2.38
1.46
0.13
0.54
0.71
3.25"
1.08
3.96a
2.42
2.00
0.17
0.33
0.83
4.58"
1.17
5.33'
Note. N = 24 for predictable conditions and 12 for the unpredictable
conditions.
* Predictable and unpredictable conditions differ at p < .10.
b
Predictable and unpredictable conditions differ at p < .01.
The predictability manipulations did not influence the irrelevant
adjectives. However, there were main effect differences between
the positive and negative conditions for Lonely, F(,
34) = 6.51, p < .05, and for Ambitious, F(l, 34) = 7.23, p ^
.05. Thus these adjectives were influenced by the mood manipulations.
Ambitious and Lonely had seemed on an a priori basis
to be irrelevant to the stories. However, the fact that the affect
manipulations influenced them suggests a general "spreading"
of negative or positive affect to other emotions of the same hedonic
type. In other words, increased negative affect tends to
increase other specific negative emotions and positive affect tends
to increase other specific positive emotions.
Discussion
There are several limitations of the first study. The stories may
have produced only modest levels of affect. In addition, stories
are somewhat artificial stimuli and not highly relevant to subjects'
lives. Last, surprise is a nonstandard way to manipulate intensity
of affect, and there could be unique properties of such a manipulation.
To remedy these limitations and obtain more ecologically
valid data, in Study 2 we collected emotion reports from subjects
on a daily basis over a 6-week period. In addition to the ecological
representativeness of such emotion reports (Epstein, 1982), in
Study 2 we had the advantage of being able to examine the actual
co-occurrence of emotions over a broad spectrum of intensity.
Study 2
Method
Participants. Forty-two University of Illinois undergraduates served
as subjects. They were enrolled in a semester-long course on mood research
and were required to complete daily mood forms and many personality
questionnaires (see Diener & Emmons, 1984, for further details). Students
who did not wish to be subjects had the alternative of serving as experimenters
or assistants in other studies. There appeared to be a wide range
of students participating. Some wanted to learn about research, some
wanted to learn about themselves, and others entered the course because
it did not involve tests and papers. The course was graded credit-no
credit; 3 persons received no credits (for completing an insufficient number
of assignments), and their data are not included here.
Procedures. For this study subjects were to complete a mood form
each time they felt "emotional," but no more than one report per day in
order to ensure that a new emotional reaction was being captured by
each report. They were instructed to pick a time when they felt an emotion
that for them was fairly strong. Most participants completed approximately
one mood report per day, but some produced only one report
every several days. We have found that college students experience more
intense emotions on the average than do older persons (Diener, Sandvik,
& Larsen, 1985), and therefore an emotion report almost every day is
not unreasonable. Obviously some subjects set their threshold higher
than others for the level of feeling that was strong enough to warrant a
mood report. Nevertheless, subjects were instructed to complete a report
about once a day for the time that would usually qualify as their most
emotional time that day. Subjects could complete the report immediately
or as late as bedtime of that night, but were encouraged to fill it out as
soon as practicable. The emotion reports had to be turned in the day
after they were completed.
Mood instrument. The emotion measure contained nine mood terms
designed to enable us to measure positive and negative affect. Four were
positive (happy, joyful, pleased, and enjoyment/fun) and five were negative
(depressed/blue, unhappy, frustrated, angry/hostile, and worried/anxious/
fearful). Each term was to be marked on a scale on which 0 = Not at all
and 6 = Extremely much. The intermediate values on this 7-point scale
were also anchored with labels such as Very slight and Much. The nine
adjectives were chosen on the basis of both a previous factor analysis
(Diener & Emmons, 1984) and an effort to include the most important
specific emotions listed by most emotion theorists (e.g., anger, fear, joy,
and sadness). Although in some analyses we considered the single mood
adjectives, the majority were based on composites for positive and negative
mood. The composite for positive mood was the sum of the four adjectives
divided by 4, and the negative composite was computed in an analogous
fashion.
Results
During the course of the 6-week study, participants generated
1,416 emotion reports, or an average of about 34 per subject (in
42 days). A frequency count of positive and negative affect felt
by all subjects at these emotional times is shown in Table 3.
Because the composite affect scores were averages that were based
on a number of mood adjectives, we rounded the averages to the
closest integer to achieve the values shown in Table 3. One obvious
fact is that very high levels of emotion occurred less frequently
than did lower and moderate levels of affect. Another
observation is that positive and negative affect did not occur at
high levels together. Virtually all of the lower right cells are devoid
of entries. Similarly, the upper left cells have fewer entries because
these would not as readily qualify as emotional times, and therefore
no emotion reports would be completed for them. Those
few times when reports were turned in with virtually no affect
can be attributed to several compulsive students who felt obliged
to complete a report every day.
A polynomial regression was computed in order to determine
whether the distribution was completely linear or had a curvilinear
component. Because our computer program could analyze
NEGATIVE AND POSITIVE AFFECT 1035
Table 3
Frequency Count of Combinations of Positive
and Negative Affect at Emotional Times
Level of
reported
negative affect
0
1
2
3
4
5
6
Marginal
totals
Level of reported positive affect
0
1
4
27
55
77
71
39
274
1
0
10
50
105
67
33
10
275
2
9
33
31
31
27
11
0
142
3
43
58
25
21
12
4
0
163
4
93
98
29
3
0
0
0
223
5
93
99
22
2
1
0
0
217
6
57
51
12
2
0
0
0
122
Marginal
totals
296
353
196
219
184
119
49
1,416
a maximum of 500 observations at once, and in order to
cross-validate the results, three separate regressions were calculated
on the emotion times of the subjects randomly divided
into three groups. The linear coefficients were high and significant
in all three groups (rs = —.64, —.71, and —.66). However, there
were also highly significant curvilinear components in each of
the three groups, F(l, 457) = 46.53, F(l, 484) = 25.33, and F(l,
466) = 29.93, all ps < .001. The degrees of freedom vary because
subjects had somewhat different numbers of emotion reports in
the three groups. One might think that the curvilinear effect was
due to combining data from different subjects. However, individual
regression analyses on the data of individual subjects revealed
that a statistically significant curvilinear effect existed in
about one half of the subjects' data (despite much lower degrees
of freedom).
It is interesting that data that we artificially created in a perfect
L shape, with a right angle and with roughly the same marginal
frequencies as appear in Table 3, yielded a linear correlation
coefficient of —.60. Thus even data that have a perfect L shape
will show a strong linear component when they are analyzed via
Pearson's product-moment correlation. In other words, a high
inverse Pearson correlation does not necessarily imply a truly
linear inverse relation.
The average values of each type of affect at each level of the
other type of affect are shown in Table 4. High average levels of
positive affect occur only at very low levels of negative affect.
Similarly, high average levels of negative affect occur only at low
levels of positive affect. However, once a certain level of affect is
reached, the other type of affect does not change dramatically.
A second approach to the question of average values is also
shown in Table 4. If one looks at those times when one type of
affect predominated over the other type, and ignores times when
the two were equal, one can compute the average level of each
type of affect at each level of the other type of affect. In other
words, when a person is predominantly happy, how does negative
affect vary as his or her positive affect increases? Similarly, when
a person is predominantly unhappy, how does his or her positive
affect vary as a function of the level of negative affect? In Table
4 these values are labeled Average subordinate values and are
shown for each level of the dominant affect. Values are not shown
at the 0 and 1 levels of the dominant affect because it is impossible
to have any of the opposite type of affect if the predominant
affect is at these low levels. These subordinate values are interesting
in that they are very low and rather uniform across all
levels of the dominant affect. This suggests, for example, that
when a person is predominantly happy at a particular time, he
or she will experience a very low level of negative affect. More
importantly, the level of negative affect does not decrease dramatically
as he or she becomes happier. Once a person is predominantly
happy, he or she tends to experience low levels of
negative affect across all levels of positive affect. When a predominantly
happy person experiences lower levels of positive
affect, he or she does not have to experience any specific level of
negative affect. The same pattern is evident when negative affect
predominates.
As informative as the average values just presented may be,
the variance in the data is also important. The standard deviations
of each type of affect at each level of the other type of affect were
therefore computed. This analysis will reveal whether one type
of affect is more dispersed or variable at certain levels of the
other type of affect. The standard deviations for negative affect
were computed for each level of positive affect from 0 through
6 and were, in order, 1.27, 1.13, 1.39, 1.33, 0.73, 0.72, and 0.72.
Similarly, the standard deviations for positive affect at each level
of negative affect were 1.08, 1.35, 1.81, 0.93, 0.80, and 0.40.
What is clear is that the standard deviations for one emotion are
much lower at high levels of the other emotion; that is, there is
a large amount of variability in an emotion at low and moderate
levels of the opposite valenced affect, but very low variability of
an emotion when the opposite valenced affect is high.
Although these analyses were all conducted on global positive
and global negative affect, one can also examine specific discrete
emotions. The correlations between specific emotions across all
1,416 emotion times are shown in Table 5. Several things are
apparent. First, emotions of the same valence tend to show high
correlations with each other. This means, for example, that when
Table 4
Average Values of One Type of Affect at Each
Level of the Other Type of Affect
Value
Positive affect
0
1
2
3
4
5
6
Average
4.09
3.39
2.47
1.46
0.74
0.71
0.66
Average
subordinate value
Negative affect
—
.79
.86
.74
.71
.66
Negative affect
0
1
2
3
4
5
6
Positive affect
4.48
4.09
2.47
1.21
0.88
0.56
0.20
.65
.87
.86
.56
.20
1036 ED DIENER AND ASHGAR IRAN-NEJAD
Table 5
Correlations Between Specific Affect Adjectives
Specific emotions 1
1. Happy
2. Joyful
3. Pleased
4. Fun/Enjoyment
5. Depressed/blue
6. Unhappy
7. Frustrated
8. Angry/hostile
9. Worried/anxious
.85
.88
.80
-.66
-.78
-.72
-.64
-.45
.83
.79
-.55
-.67
-.66
-.52
-.40
.77
-.60
-.73
-.70
-.59
-.44
-.55
-.65
-.63
-.51
-.39
.81
.66
.51
.48
.77
.68
.51
.69
.58 .38
people feel frustrated, they are also likely to feel anxious and
angry. These correlations between discrete emotions indicate that,
as suggested by Russell (1978), an overall structure for emotions
exists, even though there may be distinct emotions. Another obvious
pattern in Table 5 is that emotions of opposite valence
tend not to occur together. However, these inverse correlations
are not quite as strong in absolute magnitude as the correlations
between emotions of the same valence (average values were .59
and .69, respectively). This suggests once again that although
opposite-valenced emotions tend to be incompatible, their relation
is not completely inverse.
Discussion
In Study 2 an inverse view receives some support in light of
the strong inverse correlations found between both global and
specific negative and positive emotions. However, the existence
of a truly linear inverse relation is doubtful because of the significant
curvilinear pattern and also the averages shown in Table
4. If the pattern were actually linear and inverse, average
affect would decline in a uniform way without the precipitous
drop that is evident in Table 4. In addition, the values of the
subordinate affect would not be as uniform as they are across
levels of the predominant affect. If the relation between negative
and positive affect were inverse and liner, one would also expect
the standard deviations either to be uniform or to be larger in
the middle categories and smaller on the ends. However, the
clearest pattern is one of small standard deviations at the upper
ends of the other type of affect. There is some tendency for smaller
standard deviations to occur also at lower levels of affect, but
this pattern seems to be due mainly to the fact that subjects only
infrequently completed mood reports with very little of either
type of affect. The averages presented in Table 4 tend to be compatible
with a mutual exclusion view, as long as mutual exclusivity
does not imply an absolute zero level of the nondominant emotion.
It is clear from Table 3 that a strong form of mutual exclusion
is not supported because it appears that intermediate
levels of negative affect can be felt simultaneously with moderate
levels of positive affect. Thus there is not complete mutual exclusivity
between the two types of affect except at strong levels
of intensity. Above the level of Moderate (a 3 on our scale), positive
and negative affect virtually never occurred together.
General Discussion
In these studies we relied not simply on the scaling of emotion
words, but rather on an analysis of self-reported emotions as
they were experienced. Several clear patterns emerged from the
data:
1. People do not simultaneously experience both positive and
negative affect at intense levels.
2. If one type of affect is at low levels, the other type of affect
can be at any level of intensity from low to high. Similarly, when
the predominant affect changes in intensity, the other type reflects
no more change than is evident in irrelevant adjectives. In other
words, the levels of positive and negative affect that one experiences
are not tightly linked. However, a moderate inverse correlation
between the two will emerge if a broad spectrum of
times is considered because the occasional intense positive or
negative affect is incompatible with high levels of the other type
of affect. This reasoning is confirmed by the pattern of results
found by Diener and Emmons (1984). They reported a correlation
of —.57 between positive and negative affect when mood
reports were completed at random times, but a very strong correlation
of —.85 when the ratings were completed during times
of strong emotion. Nevertheless, near independence of average
levels of positive and negative affect will occur when individuals
are compared over time because affect occurs at relatively nonintense
levels most of the time. Thus positive and negative affect
can occur rather independently over time because low levels of
one type of emotion are compatible with any level of the other
type of emotion. In other words, average levels of positive and
negative affect can be independent across persons because most
affect occurs at nonintense levels.
3. Emotions of similar hedonic valence tend to co-occur. For
example, when a negative emotion is present, it is likely that
other negative emotions will also be experienced. The co-occurrence
of emotions of the same hedonic tone is stronger than the
inverse relation between emotions of opposite hedonic tone. This
has interesting implications for theories of emotion. If various
discrete emotions co-occur, this suggests that they must fit into
some larger more all-encompassing structure. It also suggests
that certain underlying processes must be common to a number
of emotions. Theories denying that there are structured interrelations
among emotions and those denying that there are categories
such as "negative emotion" must explain why some
emotions tend to be experienced together and some do not. Watson
and Clark (1984) postulated the existence of a broad "negative
affectivity" factor that rests implicitly on the assumption that
unpleasant emotions tend to co-occur within the same persons.
What the link is between emotions of the same hedonic tone
should be explored in future research.
NEGATIVE AND POSITIVE AFFECT 1037
4. Positive or negative affect were experienced at virtually all
times that were classified by participants as "emotional." This
suggests that positive and negative affect are essential elements
of emotion. In other words, it appears that one necessary component
of emotion is that it be a pleasant or unpleasant experience.
5. Subjects did report times at which they felt moderate
amounts of positive and negative affect together. These times
require more study in order to know whether subjects actually
have both feelings together or switch rapidly back and forth between
the two; that is, can people actually experience positive
and negative affect at the same instant? If they can, why does
this not occur at intense levels of emotion?
These five conclusions are important for several reasons. First,
patterns found in factor analytic and other approaches to uncovering
the "structure" of emotion perhaps arise because of
these conclusions. For example, the incompatibility of higher
levels of negative and positive affect may be responsible for the
bipolar positive-negative factor that is found so often. Second,
these conclusions suggest research directions in addition to those
aimed at uncovering the structure of emotion; that is, each of
these conclusions represents a phenomenon of interest in its own
right and should be explored further. Why is it that emotions of
the same valence tend to co-occur? Why do strong negative and
positive emotions not co-occur? In determining the structure of
emotion, our findings suggest several important considerations.
The curvilinearity that we uncovered should be taken into account
in structural models. The intensity of emotions that are
examined should be carefully considered because different patterns
can emerge at various levels of intensity. A simple inverse
linear view of positive and negative affect must be rejected, despite
sizable inverse Pearson correlations. Indeed, if only strong levels
of emotion are considered, a highly inverse linear relation will
be found. Nevertheless, when varying levels of intensity are closely
examined, it becomes apparent that a true inverse relation between
positive and negative affect does not exist. It is clear that
one hedonic tone can be at varying levels without affecting low
levels of the opposite hedonic tone. Furthermore, it appears that
discrete emotions can "spread" so that different emotions of the
same quality often co-occur. One can explain this by postulating
that there are interconnections between these discrete emotions
so that they are not completely independent of one another. The
connection that couples emotions of similar hedonic tone could
be mediated by the limbic system, by autonomic arousal, or by
the cognitive processes that lead to emotions of a certain type.
In terms of cognitive processes, it might be that situations that
lead to a certain set of cognitive labels are also amenable to other
cognitive labels that eventuate in emotions of the same hedonic
tone. The correlations for specific emotions found here support
the idea that specific emotions definitely have relations with one
another in terms of co-occurrence. Therefore, the discrete emotions
cannot be considered as totally independent of one another
and can be fit into a larger structure as suggested by Russell
(1979) and others.
One can analyze what various relations between positive and
negative affect tell us about the underlying affect system. If they
were completely independent, this would suggest that two separate
affect systems are involved and that the activity of one type of
affect system is independent of the activity of the other. If the
two types of affect vary inversely, this would imply that there are
two separate systems but that they are related by direct inhibition
paths. This means that the degree of activation in one system is
proportionally related to the degree of inhibition in the other
system. Last, a mutual exclusion relation might suggest that there
are not totally separate positive and negative affect systems localized
in the brain, such that when one is active the other remains
. structurally intact. Rather, the experiences of positive and negative
affect arise as a result of the functional interaction among
a large number of distributed neuronal microsystems, much in
the same fashion that a large number of suitably color-coded
light bulbs might generate two different patterns of light (see
Iran-Nejad, 1980; Iran-Nejad, Clore, & Vondruska, in press; IranNejad
& Ortony, in press). Certain neuronal systems may be
connected to certain types of affect, but the experience of emotion
may arise only from the interaction of a number of microsystems,
some of which are common to all types of affect. Mutual exclusion
occurs to the extent that the experience of one type of affect
requires the engagement of the same set of microsystems that
must also be engaged if the experience of the other type of affect
is to occur. In other words, there are common microsystems or
elements that are necessary for both positive and negative affect.
Whether positive or negative affect emerges depends on which
of the other sets of microsystems are also engaged and dominant
at the time. In this sense, positive and negative affect are mutually
exclusive in the same way that a smile and a frown are mutually
exlusive—that is, to the extent that some of the same muscles
and the lips must be engaged in both. It is conceivable that a
person can make a smiling frown or a frowning smile only when
the frown or smile are not making full or intense use of the
muscles involved. In the same fashion, at low levels of intensity,
one might be able to switch back and forth from pleasure to
displeasure imagery, but as the intensity of one type of affect
increases, so does its degree of mutual exclusivity with the other.
This occurs because the common elements needed for an emotion
are increasingly committed to the dominant emotion as intensity
increases, and the nondominant emotion cannot occur without
these elements.
There are several limitations to our studies. We relied solely
on self-report data, and it would be valuable to examine other
types of data in the same way. Do positive and negative facial
expressions, for example, yield findings that are parallel to those
that we found? In addition, our findings could be extended by
the use of experimental designs in which more powerful mood
manipulations are used.
The data on the relations between discrete emotions are interesting,
but preliminary. We used only correlational analyses,
although we found for the other types of data that these were
inadequate to completely understand the relations involved.
However, we thought that a more thorough analysis should await
future studies. First, the discrete emotions in Study 2 were measured
via only a single mood adjective and therefore may be
unreliable. Second, several of these emotions occurred somewhat
less frequently at intense levels than global affect, so that a detailed
analysis would be based on very small numbers of cases. In addition,
no manipulations were carried out in order to alter levels
of specific discrete emotions. Last, it would be desirable to include
a larger sample of emotions in such an indepth analysis. A thorough
examination of the relation between the experience of var-
1038 ED DIENER AND ASHGAR IRAN-NEJAD
ious specific emotions would be a valuable direction for future
research.
A deep rift between those who argue for discrete emotions
and those who postulate an overall structure of emotions has
divided the field of emotion. Our findings present a challenge to
both sides. In terms of emotional structure, our findings indicate
that simple and linear relations as suggested by most dimensional
systems do not fully capture the interrelations among emotions.
On the other hand, there do appear to be systematic relations
between emotions that are not explored in most theories that
postulate the existence of independent discrete emotions. It will
be more important in the long run to uncover why emotions
show certain patterned interrelations than it will be to simply
describe the structure of emotions.
References
Clyde, D. J. (1963). Manual for the Clyde Mood Scale. Coral Gables,
FL: Biometric Laboratory, University of Miami.
Daly, E. M., Lancee, W. J., & Polivy, J. (1983). A canonical model for
the taxonomy of emotional experience. Journal of Personality and Social
Psychology, 45, 443-457.
Diener, E. (1984). Subjective well-being. Psychological Bulletin, 95, 542-
575.
Diener, E., & Emmons, R. A. (1984). The independence of positive and
negative affect. Journal of Personality and Social Psychology, 47, 1105-
1117.
Diener, E., Emmons, R. A., & Sandvik, E. (1985). The dual nature of
happiness: The independence of positive and negative affect. Manuscript
submitted for publication.
Diener, E., Larsen, R. J., Levine, S., & Emmons, R. A. (1985). Intensity
and frequency: The underlying dimensions of positive and negative
affect. Journal of Personality and Social Psychology, 48, 1253-1265.
Diener, E., Sandvik, E., & Larsen, R. J. (1985). Age and sex effects for
emotional intensity. Developmental Psychology, 21, 542-546.
Ekman, P., Friesen, W. V., Ellsworth, P. (1972). Emotion in the human
face. New York: Pergamon.
Epstein, S. (1982). A research paradigm for the study of personality and
emotions. In M. M. Page (Ed.), Personality—Current theory and research:
1982 Nebraska Symposium on Motivation (pp. 91-154). Lincoln:
University of Nebraska Press.
Iran-Nejad, A. (1980). The schema: A structural or a functional pattern
(Tech. Rep. No. 159). Urbana: University of Illinois, Center for the
Study of Reading. (ERIC Document Reproduction Service No. ED
181 449)
Iran-Nejad, A. (1983). Qualitative and quantitative causes of the experience
of affect (Doctoral dissertation, University of Illinois at Champaign-Urbana,
1983). Dissertation Abstracts International, 43, 3847A.
Iran-Nejad, A., Clore, G. L., & Vondruska, R. J. (in press). Affect: A
functional perspective. The Journal of Mind and Behavior.
Iran-Nejad, A., & Ortony, A. (in press). A biofunctional model of distributed
mental content, mental structures, awareness, and attention.
The Journal of Mind and Behavior.
Izard, C. E. (1972). Patterns of emotions. New York: Academic Press.
Laymon, R. (1975). Out of the woods. Ellery Queen's Mystery Magazine,
66, 133-135.
McNair, D. M., & Lorr, N. (1964). An analysis of mood in neurotics.
Journal of Abnormal and Social Psychology, 69, 620-627.
Nowlis, V., & Nowlis, H. H. (1956). The description and analysis of
mood. Annals of the New York Academy of Sciences, 65, 345-355.
Plutchik, R. (1980). Emotion: A psychoevolutionary synthesis. New'Vbrk:
Harper & Row.
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.
Schlosberg, H. (1954). Three dimensions of emotion. Psychological Review,
61, 81-88.
Thurmond, P. J. (1978). If cornered, scream. Ellery Queen's Mystery
Magazine, 71, 66-68.
Watson, D., & Clark, L. A. (1984). Negative affectivity: The disposition
to experience aversive emotional states. Psychological Bulletin, 96, 465-
490.
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., & Tellegen, A. (1985). Toward a consensual structure of
mood. Psychological Bulletin, 98, 219-235.
Zevon, M. A., & Tellegen, A. (1982). The structure of mood change: An
idiographic/nomothetic analysis. Journal of Personality and Social
Psychology, 43, 111-122.
Received April 23, 1985
Revision received October 14, 1985 •
All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.
1986, Vol. 50, No. 5, 1031-1038
Copyright 1986 by the American Psychological Association, Inc.
0022-3514/86/S00.75
The Relationship In Experience Between Various Types of Affect
Ed Diener
University of Illinois
Ashgar Iran-Nejad
University of Michigan
The co-occurence in experience of various emotions was explored, with the focus on positive versus
negative affect. In Study 1,72 subjects read stories designed to produce varying levels of either positive
or negative affect, and then rated their level of both types of affect. In Study 2, 42 participants rated
their feelings during emotional times in everyday life for a period of 6 weeks. We found that emotions
of the same hedonic valence (e.g., fear and anger) tend to co-occur. We also found that positive and
negative affect do not occur together at high levels of intensity. These two facts about the relation of
positive and negative affect are probably responsible for the bipolarity that is often found between
them. These findings represent a challenge to those who postulate that there are unrelated discrete
emotions and that the terms positive affect and negative affect do not describe meaningful clusters of
emotions. We found that if one type of affect is of low intensity, the other type can be at any level
from low to high. Therefore, a truly inverse and linear relation does not characterize positive and
negative affect. This finding represents a challenge to most structural models of emotion. It appears
that mutual exclusion only at high levels of intensity characterizes the relation between positive and
negative affect.
We explore the relation between feeling particular emotions
and experiencing other emotions. For example, if one feels joy,
what is the likelihood of also feeling anger or fear? In recent years
a great deal of discussion has focused on the structure of emotion
(e.g., Daly, Lancee, & Polivy, 1983; Plutchik, 1980; Russell, 1978,
1979). In addition to the structure of momentary emotion, Diener
and his colleagues (Diener, 1984; Diener, & Emmons, 1984;
Diener, Emmons, & Sandvik, 1985; Diener, Larsen, Levine, &
Emmons, 1985) have also examined the long-term structure of
affect across persons. One necessary component in understanding
the structure of affect is to examine how compatible the experience
of varying emotions is. Are certain emotions compatible
and others mutually exclusive? When the levels of an emotion
change, do other emotions also change in intensity? For example,
as negative affect increases, is there a necessary decrease in positive
affect? When an emotion is felt, is the intensity related in
a linear or nonlinear way to other emotions? Gaining knowledge
of the interrelations of emotions as they are experienced will aid
in understanding the structure of emotion, as well as the processes
underlying affect. If particular emotions facilitate the experiencing
of certain other emotions but inhibit yet other emotions, this
will provide valuable insights into how the affect system functions.
In order to narrow the range of the questions asked, we focus
on global positive and negative affect rather than on discrete
emotions such as anger. There were several reasons why focusing
on only global positive and negative affect was desirable. First,
a two-dimensional system of emotion accounts for much of the
variance in affect and has been most replicable (e.g., Russell,
1978, 1979; Schlosberg, 1954; Watson, Clark, &Tellegen, 1984;
Watson & Tellegen, 1985; Zevon & Tellegen, 1982). The two
dimensions are either positive-negative affect and arousal (e.g.,
Correspondence concerning this article should be addressed to Ed
Diener, Department of Psychology, University of Illinois, 603 E. Daniel,
Champaign, Illinois 61820.
Russell, 1978) or a similar system that appears to be a 45° rotation
of the other system, with negative and positive affect as
the two dimensions (see Watson & Tellegen, 1985). In contrast
to the ubiquitous bipolar pleasant-unpleasant and arousal dimensions,
factors of multiple discrete emotions (e.g., fear and
joy) tend not to entirely replicate across investigators (e.g., Clyde,
1963; Ekman, Friesen, & Ellsworth, 1972; Izard, 1972; McNair
& Lorr, 1964; Nowlis & Nowlis, 1956; Plutchik, 1980). Nor has
there been agreement when dimensions in addition to positivenegative
affect and arousal are proposed. A second reason for
focusing on positive and negative affect is to simplify this initial
analysis of the covariation in the experiencing of different emotions.
Zevon and Tellegen (1982) explored the interrelation of positive
and negative affect through within-subject factor analyses. They
found two independent factors, negative and positive affect, which
replicated across virtually all subjects. Diener and Emmons
(1984) also explored the experience of positive and negative affect
within individuals. They found that positive and negative affect
correlated highly inversely when the experience of affect was
reported at particular moments. However, they found that when
longer time intervals were examined, the experiencing of positive
and negative affect tended to be uncorrelated across individuals.
Diener, Emmons, and Sandvik (1985) reviewed the evidence for
the existence of orthogonal negative and positive affect dimensions
when long-term mood was considered. On the other hand,
Russell (1978) found that positive and negative emotions were
not independent, but were at opposite ends of a bipolar continuum.
In this article we intend to extend the findings that in the
earlier studies were based solely on correlational approaches. In
particular, we first examine whether differences in the level of
one type of affect are invariably accompanied by differing levels
of intensity in the other type of affect. It is possible that positive
and negative affect correlate only because they are incompatible
at high levels of intensity. However, this would not necessarily
mean that there was a close link between the two at other less
1031
1032 ED DIENER AND ASHGAR IRAN-NEJAD
intense levels. It might be that rather than show an inverse relation,
as suggested by a negative correlation, or no correlation,
as suggested by independent dimensions, positive and negative
affect may be mutually exclusive; that is, positive affect may be
experienced intensely only when negative affect is not present.
However, it may be that the degree of positive affect can vary
from low to high, as long as the level of negative affect is low. In
addition, it may be that an increase or decrease in positive affect
does not necessitate a change in negative affect as long as the
latter is experienced at relatively low levels. Such a relation would
appear as a simple moderate negative correlation unless one made
a finer grained analysis of emotion reports. Thus the relation
between positive and negative affect may be more complex and
not completely captured by correlational or factor analytic approaches.
A closely related issue is whether positive and negative
affect are related in a truly linear way. The conclusions of Diener
and Emmons (1984) and Zevon and Tellegen (1982) were based
on an exploration of linear relations through Pearson's productmoment
correlations. However, the idea of mutual exclusion
suggests that a nonlinear relation may exist. Such a relation would
produce an inverse correlation if it were analyzed via the Pearson
statistic, although this correlation would hide an underlying curvilinear
relation.
In summary, this article is a further exploration of the relation
of the experience of positive and negative affect. In earlier studies
researchers relied on correlational approaches. We examined
whether there is a necessary inverse relation between positive
and negative affect. If the two are mutually incompatible but not
strictly inverse, it is necessary to use approaches beyond linear
correlation to uncover this fact. The major question is whether
levels of positive affect are invariably and linearly related to levels
of negative affect. A mutual exclusion view suggests that this will
not be the case. A related issue is whether there are patterns of
co-occurrence for various emotions. If emotions are discrete and
independent, they should not show specific patterns of co-occurrence,
and categories such as "positive affect" and "negative
affect" would have little scientific value. However, if certain emotions
tend to be experienced together or are mutually exclusive,
then categories or dimensions of emotions take on scientific import.
In order to explore the relations between various types of affect
in depth, several approaches were used. In Study 1, an experimental
manipulation was used to produce lower or higher levels
of positive and of negative affect. Thus the pattern of correlations
between the two could be explored both within and across conditions.
If individuals' levels of affect correlated only when both
positive and negative conditions were combined, not when the
conditions were examined separately, it would suggest not that
the two types of affect are invariably linked, but rather that they
are only inversely correlated when strong levels of affect are considered
because the two types of affect tend not to co-occur at
intense levels. In addition, the means in each condition can also
be considered determinants of whether a lower level of one type
of affect necessitates a higher level of the other type of affect. In
the second study, subjects reported their affect in everyday situations
at times over a 6-week period when they felt "emotional."
We explored the relation between negative and positive affect in
these reports in several ways. Both linear and nonlinear correlations
were computed. An analysis of the variance in one type
of affect at each level of the other type of affect was also explored.
In this way, we could determine whether the variability in each
type of affect is greatest at low levels of the other type of affect.
If so, this would suggest that there is not a simple inverse relation
in the experience of positive and negative affect, but rather a
more complex relation between the two.
In addition to the in-depth analysis of the relation between
positive and negative affect just described, we also present in the
second study a preliminary analysis of the relation between discrete
emotions such as fear and anger. The major question was
whether emotions of the same hedonic type tend to co-occur. If
so, this would have clear implications for a comprehensive theory
of emotions. If emotions of the same hedonic tone tend to cooccur,
it would indicate that the underlying processes leading to
these emotions in some way link them together.
Study 1
Method
Participants and procedures. Seventy-two college students served as
subjects. These individuals were involved in a long-term study on daily
mood, but were also requested to complete additional tasks. In this study
each subject read two emotional stories, one negative and one positive.
We manipulated intensity of mood with the "surprisingness" of the story
ending, which has proved to be an acceptable method for manipulating
degree of affective arousal (Iran-Nejad, 1983). After reading the stories,
subjects completed an affect questionnaire.
There were two stories in both the good and bad conditions. The first
story, adapted from Thurmond (1978), described an incident in which
a woman was concerned about running out of gas late at night. After
pulling into a service station, Marilyn is asked to come inside by the
attendant. Because a maniac murderer is on the loose in the area, she is
reluctant to go inside, but does so when he insists. In the negative condition,
the gas station attendant strangles her. In the positive condition,
it turns out that the attendant is really a good Samaritan intending to
save her from a man whom he has spotted hiding on the back seat of her
car. In the second story, adapted from Laymon (1975), the main character
is awakened by footsteps outside his tent. He discovers that Sally is missing
and does not respond to his calls. When he comes out of the tent, there
is a man waiting with a gun in one hand and an ax in the other. In the
negative condition, the camper is murdered. In the positive condition,
the bad man is overpowered and arrested. Each subject read the good
version of one of the stories and the bad version of the other story. Because
each subject read both a good and a bad story, subjects were repeated
across the positive-negative manipulation, but were nested within level
of surprise; that is, each subject was in a particular condition for surprise
in both the positive and negative conditions.
For each of the story conditions, there were three levels of surprisingness
that were designed to manipulate cognitive arousal and therefore emotional
intensity. The information given in the high surprise endings gave
no indication of the story ending. For example, in the negative gas station
story, the attendant was portrayed as a very nice fellow, so that his murder
of the woman is unexpected. Similarly, in the positive condition of the
camping story, Sally turns out to be the camper's dog who attacks the
bad man. In each high surprise condition, the story ending is very unpredictable
from the information contained in the story. In the medium
surprise condition, hints are included throughout the story about the
ending. In the low surprise condition, the ending of the story is obvious
from information that is directly stated in the story.
Mood instrument. After reading each story, subjects completed a
questionnaire. There was one question per page and these questions were
systematically varied in order across subjects so as to eliminate order
effects. Each question was rated on an 8-point scale anchored by Not at
NEGATIVE AND POSITIVE AFFECT 1033
all (0) and Extremely (7). We also included four irrelevant
adjective questions (lonely, ambitious, toothache, and sweet taste) in order
to see whether the manipulations affected them and to examine their
relation with emotions. We included four questions to assess levels of
positive and negative affect: pleasant, unpleasant, happy, and unhappy.
For example, one question read "The conclusion of the story I just read
made me experience:"; the response options ranged from 0 (No unpleasantness
at all) to 7 (Extreme unpleasantness).
Results
Pearson correlations were computed between the affect adjectives
in order to determine the relation of positive and negative
affect across individuals. The correlations were computed for the
72 subjects across all conditions and also separately for the positive
and negative conditions. These correlations are shown in
Table 1. Several things are clear in this table. First, adjectives
correlate much more strongly with those of the same hedonic
type than with those of the opposite hedonic value. This occurs
in all three sets of correlations. Second, across all conditions,
positive and negative affect vary inversely at moderate levels.
This indicates that when one considers conditions in which radically
different levels of positive and negative affect are present,
an inverse correlation will be found between the two. In other
words, when one examines groups in which some persons are
experiencing high-intensity positive affect and other persons are
experiencing high-intensity negative affect, then across all individuals
an inverse correlation between positive and negative affect
will emerge. However, when one turns to the correlations within
the positive and negative affect cells, one is examining the variations
in affect among individuals who are experiencing varying
degrees of positive affect in the positive condition and varying
degrees of negative affect in the negative condition. In this case,
the correlations between positive and negative affect drop to
nonsignificant values close to zero. This indicates that differences
in positive affect between individuals when they are happy are
not necessarily related to corresponding differences in negative
affect. The same pattern is true when negative affect predominates.
These individual difference correlations support the conclusion
that differences in positive affect are not always related
to differences in negative affect. In other words, there is not a
necessary inverse relation between the two. However, if one considers
a broad spectrum of moments in which at times negative
affect is strong and at times positive affect is strong, an inverse
correlation between the two will be found.
One definitely cannot attribute the low correlation between
positive and negative affect to restriction of range or to floor
effects. If the nondominant affect was restricted in range or had
reached a floor, it could not possibly correlate at moderate to
high levels with adjectives of the same type (rs = .65 and .44).
These correlations indicate that sizable relations are still possible
with the variance that exists for the nondominant affect. In addition,
as the means in Table 2 show, the mood adjectives had a
long way to move before encountering the floor effect that is
clearly evident for "Toothache."
The results revealed that in some cases the stories were not
effective in producing differences in affect. In order to assess how
average differences across individuals of one type of affect relate
to differences in the other type of affect, it was necessary to compare
stories that differed significantly in level of affect produced.
Table 1
Correlations of Mood Adjectives
Mood adjectives
All conditions
Happy
Pleasant
Unhappy
Unpleasant
Positive condition only
Happy
Pleasant
Unhappy
Unpleasant
Negative condition only
Happy
Pleasant
Unhappy
Unpleasant
Happy
.77
-.38
-.45
.77
.07
-.08
.44
-.05
-.15
Pleasant
-.33
-.39
.07
-.08
-.09
-.11
Unhappy
.80
.65
.74
Note. For rs(70) of .20 and greater, p < .05; for rs(70) of .28 or greater,
Therefore, the low- and medium-surprise stories were combined
because they produced very similar levels of affect for both stories.
In addition, only one positive story (about the gas station) and
one negative story (about the camper) were retained because the
other two stories showed no mean differences for either positive
or negative affect between any of the levels of surprise. Thus after
deleting the ineffective manipulations, we had a 2 X 2 design
(low/medium vs. high surprise and positive vs. negative ending).
Dropping of the ineffective stories occurred without reference
to the data that would shed light on the various questions under
study.
The check on the surprisingness manipulation produced highly
significant results, F(, 34) = 31.76, p < .001, which did not
interact with the positive versus negative story conditions. The
means for the four cells are shown in Table 2. We computed
preplanned orthogonal comparisons to compare the predictable
and unpredictable cells within the positive and negative conditions
(for all questions). The significance levels of these comparisons
are indicated in the table by the letters a and b. The
level of surprisingness was effective in manipulating level of affect
in both positive and negative conditions.
The key findings in Table 2 are those for the mood adjectives.
The positive stories differed significantly between the predictable
and unpredictable cells in terms of Happy and Pleasant. However,
not only did the unhappy and unpleasant ratings not differ significantly,
but the means are ordered in the same direction as
the positive adjectives. In the negative condition, there is the
same type of pattern, although the significance levels for Unhappy
and Unpleasant were of only borderline value. Once again, the
opposite adjectives (Happy and Pleasant) were almost identical
and, in addition, went in the same direction as the negative adjectives.
What these results clearly show is that positive and negative
affect can be felt at varying levels without necessarily influencing
the value of the opposite type of affect. For example,
greater positive affect in the unpredictable positive condition was
clearly not accompanied by lesser negative affect. If a person has
high affect of one type, this affect can decrease without the other
type of affects' necessarily increasing.
1034 ED DIENER AND ASHGAR IRAN-NEJAD
Table 2
Mean Ratings of Stories
Story type
Positive Negative
Questions
Predictable
Unpredictable
Predictable
Unpredictable
Surprise 2.29" 6.08" 5.83"
Irrelevant
adjectives
Lonely
Ambitious
Toothache
Sweet taste
Mood
adjectives
Happy
Unhappy
Pleasant
Unpleasant
1.92
2.46
0.00
0.54
2.67"
1.42
3.21b
2.00
1.58
2.67
0.17
0.67
5.42"
1.67
5.33"
2.25
2.38
1.46
0.13
0.54
0.71
3.25"
1.08
3.96a
2.42
2.00
0.17
0.33
0.83
4.58"
1.17
5.33'
Note. N = 24 for predictable conditions and 12 for the unpredictable
conditions.
* Predictable and unpredictable conditions differ at p < .10.
b
Predictable and unpredictable conditions differ at p < .01.
The predictability manipulations did not influence the irrelevant
adjectives. However, there were main effect differences between
the positive and negative conditions for Lonely, F(,
34) = 6.51, p < .05, and for Ambitious, F(l, 34) = 7.23, p ^
.05. Thus these adjectives were influenced by the mood manipulations.
Ambitious and Lonely had seemed on an a priori basis
to be irrelevant to the stories. However, the fact that the affect
manipulations influenced them suggests a general "spreading"
of negative or positive affect to other emotions of the same hedonic
type. In other words, increased negative affect tends to
increase other specific negative emotions and positive affect tends
to increase other specific positive emotions.
Discussion
There are several limitations of the first study. The stories may
have produced only modest levels of affect. In addition, stories
are somewhat artificial stimuli and not highly relevant to subjects'
lives. Last, surprise is a nonstandard way to manipulate intensity
of affect, and there could be unique properties of such a manipulation.
To remedy these limitations and obtain more ecologically
valid data, in Study 2 we collected emotion reports from subjects
on a daily basis over a 6-week period. In addition to the ecological
representativeness of such emotion reports (Epstein, 1982), in
Study 2 we had the advantage of being able to examine the actual
co-occurrence of emotions over a broad spectrum of intensity.
Study 2
Method
Participants. Forty-two University of Illinois undergraduates served
as subjects. They were enrolled in a semester-long course on mood research
and were required to complete daily mood forms and many personality
questionnaires (see Diener & Emmons, 1984, for further details). Students
who did not wish to be subjects had the alternative of serving as experimenters
or assistants in other studies. There appeared to be a wide range
of students participating. Some wanted to learn about research, some
wanted to learn about themselves, and others entered the course because
it did not involve tests and papers. The course was graded credit-no
credit; 3 persons received no credits (for completing an insufficient number
of assignments), and their data are not included here.
Procedures. For this study subjects were to complete a mood form
each time they felt "emotional," but no more than one report per day in
order to ensure that a new emotional reaction was being captured by
each report. They were instructed to pick a time when they felt an emotion
that for them was fairly strong. Most participants completed approximately
one mood report per day, but some produced only one report
every several days. We have found that college students experience more
intense emotions on the average than do older persons (Diener, Sandvik,
& Larsen, 1985), and therefore an emotion report almost every day is
not unreasonable. Obviously some subjects set their threshold higher
than others for the level of feeling that was strong enough to warrant a
mood report. Nevertheless, subjects were instructed to complete a report
about once a day for the time that would usually qualify as their most
emotional time that day. Subjects could complete the report immediately
or as late as bedtime of that night, but were encouraged to fill it out as
soon as practicable. The emotion reports had to be turned in the day
after they were completed.
Mood instrument. The emotion measure contained nine mood terms
designed to enable us to measure positive and negative affect. Four were
positive (happy, joyful, pleased, and enjoyment/fun) and five were negative
(depressed/blue, unhappy, frustrated, angry/hostile, and worried/anxious/
fearful). Each term was to be marked on a scale on which 0 = Not at all
and 6 = Extremely much. The intermediate values on this 7-point scale
were also anchored with labels such as Very slight and Much. The nine
adjectives were chosen on the basis of both a previous factor analysis
(Diener & Emmons, 1984) and an effort to include the most important
specific emotions listed by most emotion theorists (e.g., anger, fear, joy,
and sadness). Although in some analyses we considered the single mood
adjectives, the majority were based on composites for positive and negative
mood. The composite for positive mood was the sum of the four adjectives
divided by 4, and the negative composite was computed in an analogous
fashion.
Results
During the course of the 6-week study, participants generated
1,416 emotion reports, or an average of about 34 per subject (in
42 days). A frequency count of positive and negative affect felt
by all subjects at these emotional times is shown in Table 3.
Because the composite affect scores were averages that were based
on a number of mood adjectives, we rounded the averages to the
closest integer to achieve the values shown in Table 3. One obvious
fact is that very high levels of emotion occurred less frequently
than did lower and moderate levels of affect. Another
observation is that positive and negative affect did not occur at
high levels together. Virtually all of the lower right cells are devoid
of entries. Similarly, the upper left cells have fewer entries because
these would not as readily qualify as emotional times, and therefore
no emotion reports would be completed for them. Those
few times when reports were turned in with virtually no affect
can be attributed to several compulsive students who felt obliged
to complete a report every day.
A polynomial regression was computed in order to determine
whether the distribution was completely linear or had a curvilinear
component. Because our computer program could analyze
NEGATIVE AND POSITIVE AFFECT 1035
Table 3
Frequency Count of Combinations of Positive
and Negative Affect at Emotional Times
Level of
reported
negative affect
0
1
2
3
4
5
6
Marginal
totals
Level of reported positive affect
0
1
4
27
55
77
71
39
274
1
0
10
50
105
67
33
10
275
2
9
33
31
31
27
11
0
142
3
43
58
25
21
12
4
0
163
4
93
98
29
3
0
0
0
223
5
93
99
22
2
1
0
0
217
6
57
51
12
2
0
0
0
122
Marginal
totals
296
353
196
219
184
119
49
1,416
a maximum of 500 observations at once, and in order to
cross-validate the results, three separate regressions were calculated
on the emotion times of the subjects randomly divided
into three groups. The linear coefficients were high and significant
in all three groups (rs = —.64, —.71, and —.66). However, there
were also highly significant curvilinear components in each of
the three groups, F(l, 457) = 46.53, F(l, 484) = 25.33, and F(l,
466) = 29.93, all ps < .001. The degrees of freedom vary because
subjects had somewhat different numbers of emotion reports in
the three groups. One might think that the curvilinear effect was
due to combining data from different subjects. However, individual
regression analyses on the data of individual subjects revealed
that a statistically significant curvilinear effect existed in
about one half of the subjects' data (despite much lower degrees
of freedom).
It is interesting that data that we artificially created in a perfect
L shape, with a right angle and with roughly the same marginal
frequencies as appear in Table 3, yielded a linear correlation
coefficient of —.60. Thus even data that have a perfect L shape
will show a strong linear component when they are analyzed via
Pearson's product-moment correlation. In other words, a high
inverse Pearson correlation does not necessarily imply a truly
linear inverse relation.
The average values of each type of affect at each level of the
other type of affect are shown in Table 4. High average levels of
positive affect occur only at very low levels of negative affect.
Similarly, high average levels of negative affect occur only at low
levels of positive affect. However, once a certain level of affect is
reached, the other type of affect does not change dramatically.
A second approach to the question of average values is also
shown in Table 4. If one looks at those times when one type of
affect predominated over the other type, and ignores times when
the two were equal, one can compute the average level of each
type of affect at each level of the other type of affect. In other
words, when a person is predominantly happy, how does negative
affect vary as his or her positive affect increases? Similarly, when
a person is predominantly unhappy, how does his or her positive
affect vary as a function of the level of negative affect? In Table
4 these values are labeled Average subordinate values and are
shown for each level of the dominant affect. Values are not shown
at the 0 and 1 levels of the dominant affect because it is impossible
to have any of the opposite type of affect if the predominant
affect is at these low levels. These subordinate values are interesting
in that they are very low and rather uniform across all
levels of the dominant affect. This suggests, for example, that
when a person is predominantly happy at a particular time, he
or she will experience a very low level of negative affect. More
importantly, the level of negative affect does not decrease dramatically
as he or she becomes happier. Once a person is predominantly
happy, he or she tends to experience low levels of
negative affect across all levels of positive affect. When a predominantly
happy person experiences lower levels of positive
affect, he or she does not have to experience any specific level of
negative affect. The same pattern is evident when negative affect
predominates.
As informative as the average values just presented may be,
the variance in the data is also important. The standard deviations
of each type of affect at each level of the other type of affect were
therefore computed. This analysis will reveal whether one type
of affect is more dispersed or variable at certain levels of the
other type of affect. The standard deviations for negative affect
were computed for each level of positive affect from 0 through
6 and were, in order, 1.27, 1.13, 1.39, 1.33, 0.73, 0.72, and 0.72.
Similarly, the standard deviations for positive affect at each level
of negative affect were 1.08, 1.35, 1.81, 0.93, 0.80, and 0.40.
What is clear is that the standard deviations for one emotion are
much lower at high levels of the other emotion; that is, there is
a large amount of variability in an emotion at low and moderate
levels of the opposite valenced affect, but very low variability of
an emotion when the opposite valenced affect is high.
Although these analyses were all conducted on global positive
and global negative affect, one can also examine specific discrete
emotions. The correlations between specific emotions across all
1,416 emotion times are shown in Table 5. Several things are
apparent. First, emotions of the same valence tend to show high
correlations with each other. This means, for example, that when
Table 4
Average Values of One Type of Affect at Each
Level of the Other Type of Affect
Value
Positive affect
0
1
2
3
4
5
6
Average
4.09
3.39
2.47
1.46
0.74
0.71
0.66
Average
subordinate value
Negative affect
—
.79
.86
.74
.71
.66
Negative affect
0
1
2
3
4
5
6
Positive affect
4.48
4.09
2.47
1.21
0.88
0.56
0.20
.65
.87
.86
.56
.20
1036 ED DIENER AND ASHGAR IRAN-NEJAD
Table 5
Correlations Between Specific Affect Adjectives
Specific emotions 1
1. Happy
2. Joyful
3. Pleased
4. Fun/Enjoyment
5. Depressed/blue
6. Unhappy
7. Frustrated
8. Angry/hostile
9. Worried/anxious
.85
.88
.80
-.66
-.78
-.72
-.64
-.45
.83
.79
-.55
-.67
-.66
-.52
-.40
.77
-.60
-.73
-.70
-.59
-.44
-.55
-.65
-.63
-.51
-.39
.81
.66
.51
.48
.77
.68
.51
.69
.58 .38
people feel frustrated, they are also likely to feel anxious and
angry. These correlations between discrete emotions indicate that,
as suggested by Russell (1978), an overall structure for emotions
exists, even though there may be distinct emotions. Another obvious
pattern in Table 5 is that emotions of opposite valence
tend not to occur together. However, these inverse correlations
are not quite as strong in absolute magnitude as the correlations
between emotions of the same valence (average values were .59
and .69, respectively). This suggests once again that although
opposite-valenced emotions tend to be incompatible, their relation
is not completely inverse.
Discussion
In Study 2 an inverse view receives some support in light of
the strong inverse correlations found between both global and
specific negative and positive emotions. However, the existence
of a truly linear inverse relation is doubtful because of the significant
curvilinear pattern and also the averages shown in Table
4. If the pattern were actually linear and inverse, average
affect would decline in a uniform way without the precipitous
drop that is evident in Table 4. In addition, the values of the
subordinate affect would not be as uniform as they are across
levels of the predominant affect. If the relation between negative
and positive affect were inverse and liner, one would also expect
the standard deviations either to be uniform or to be larger in
the middle categories and smaller on the ends. However, the
clearest pattern is one of small standard deviations at the upper
ends of the other type of affect. There is some tendency for smaller
standard deviations to occur also at lower levels of affect, but
this pattern seems to be due mainly to the fact that subjects only
infrequently completed mood reports with very little of either
type of affect. The averages presented in Table 4 tend to be compatible
with a mutual exclusion view, as long as mutual exclusivity
does not imply an absolute zero level of the nondominant emotion.
It is clear from Table 3 that a strong form of mutual exclusion
is not supported because it appears that intermediate
levels of negative affect can be felt simultaneously with moderate
levels of positive affect. Thus there is not complete mutual exclusivity
between the two types of affect except at strong levels
of intensity. Above the level of Moderate (a 3 on our scale), positive
and negative affect virtually never occurred together.
General Discussion
In these studies we relied not simply on the scaling of emotion
words, but rather on an analysis of self-reported emotions as
they were experienced. Several clear patterns emerged from the
data:
1. People do not simultaneously experience both positive and
negative affect at intense levels.
2. If one type of affect is at low levels, the other type of affect
can be at any level of intensity from low to high. Similarly, when
the predominant affect changes in intensity, the other type reflects
no more change than is evident in irrelevant adjectives. In other
words, the levels of positive and negative affect that one experiences
are not tightly linked. However, a moderate inverse correlation
between the two will emerge if a broad spectrum of
times is considered because the occasional intense positive or
negative affect is incompatible with high levels of the other type
of affect. This reasoning is confirmed by the pattern of results
found by Diener and Emmons (1984). They reported a correlation
of —.57 between positive and negative affect when mood
reports were completed at random times, but a very strong correlation
of —.85 when the ratings were completed during times
of strong emotion. Nevertheless, near independence of average
levels of positive and negative affect will occur when individuals
are compared over time because affect occurs at relatively nonintense
levels most of the time. Thus positive and negative affect
can occur rather independently over time because low levels of
one type of emotion are compatible with any level of the other
type of emotion. In other words, average levels of positive and
negative affect can be independent across persons because most
affect occurs at nonintense levels.
3. Emotions of similar hedonic valence tend to co-occur. For
example, when a negative emotion is present, it is likely that
other negative emotions will also be experienced. The co-occurrence
of emotions of the same hedonic tone is stronger than the
inverse relation between emotions of opposite hedonic tone. This
has interesting implications for theories of emotion. If various
discrete emotions co-occur, this suggests that they must fit into
some larger more all-encompassing structure. It also suggests
that certain underlying processes must be common to a number
of emotions. Theories denying that there are structured interrelations
among emotions and those denying that there are categories
such as "negative emotion" must explain why some
emotions tend to be experienced together and some do not. Watson
and Clark (1984) postulated the existence of a broad "negative
affectivity" factor that rests implicitly on the assumption that
unpleasant emotions tend to co-occur within the same persons.
What the link is between emotions of the same hedonic tone
should be explored in future research.
NEGATIVE AND POSITIVE AFFECT 1037
4. Positive or negative affect were experienced at virtually all
times that were classified by participants as "emotional." This
suggests that positive and negative affect are essential elements
of emotion. In other words, it appears that one necessary component
of emotion is that it be a pleasant or unpleasant experience.
5. Subjects did report times at which they felt moderate
amounts of positive and negative affect together. These times
require more study in order to know whether subjects actually
have both feelings together or switch rapidly back and forth between
the two; that is, can people actually experience positive
and negative affect at the same instant? If they can, why does
this not occur at intense levels of emotion?
These five conclusions are important for several reasons. First,
patterns found in factor analytic and other approaches to uncovering
the "structure" of emotion perhaps arise because of
these conclusions. For example, the incompatibility of higher
levels of negative and positive affect may be responsible for the
bipolar positive-negative factor that is found so often. Second,
these conclusions suggest research directions in addition to those
aimed at uncovering the structure of emotion; that is, each of
these conclusions represents a phenomenon of interest in its own
right and should be explored further. Why is it that emotions of
the same valence tend to co-occur? Why do strong negative and
positive emotions not co-occur? In determining the structure of
emotion, our findings suggest several important considerations.
The curvilinearity that we uncovered should be taken into account
in structural models. The intensity of emotions that are
examined should be carefully considered because different patterns
can emerge at various levels of intensity. A simple inverse
linear view of positive and negative affect must be rejected, despite
sizable inverse Pearson correlations. Indeed, if only strong levels
of emotion are considered, a highly inverse linear relation will
be found. Nevertheless, when varying levels of intensity are closely
examined, it becomes apparent that a true inverse relation between
positive and negative affect does not exist. It is clear that
one hedonic tone can be at varying levels without affecting low
levels of the opposite hedonic tone. Furthermore, it appears that
discrete emotions can "spread" so that different emotions of the
same quality often co-occur. One can explain this by postulating
that there are interconnections between these discrete emotions
so that they are not completely independent of one another. The
connection that couples emotions of similar hedonic tone could
be mediated by the limbic system, by autonomic arousal, or by
the cognitive processes that lead to emotions of a certain type.
In terms of cognitive processes, it might be that situations that
lead to a certain set of cognitive labels are also amenable to other
cognitive labels that eventuate in emotions of the same hedonic
tone. The correlations for specific emotions found here support
the idea that specific emotions definitely have relations with one
another in terms of co-occurrence. Therefore, the discrete emotions
cannot be considered as totally independent of one another
and can be fit into a larger structure as suggested by Russell
(1979) and others.
One can analyze what various relations between positive and
negative affect tell us about the underlying affect system. If they
were completely independent, this would suggest that two separate
affect systems are involved and that the activity of one type of
affect system is independent of the activity of the other. If the
two types of affect vary inversely, this would imply that there are
two separate systems but that they are related by direct inhibition
paths. This means that the degree of activation in one system is
proportionally related to the degree of inhibition in the other
system. Last, a mutual exclusion relation might suggest that there
are not totally separate positive and negative affect systems localized
in the brain, such that when one is active the other remains
. structurally intact. Rather, the experiences of positive and negative
affect arise as a result of the functional interaction among
a large number of distributed neuronal microsystems, much in
the same fashion that a large number of suitably color-coded
light bulbs might generate two different patterns of light (see
Iran-Nejad, 1980; Iran-Nejad, Clore, & Vondruska, in press; IranNejad
& Ortony, in press). Certain neuronal systems may be
connected to certain types of affect, but the experience of emotion
may arise only from the interaction of a number of microsystems,
some of which are common to all types of affect. Mutual exclusion
occurs to the extent that the experience of one type of affect
requires the engagement of the same set of microsystems that
must also be engaged if the experience of the other type of affect
is to occur. In other words, there are common microsystems or
elements that are necessary for both positive and negative affect.
Whether positive or negative affect emerges depends on which
of the other sets of microsystems are also engaged and dominant
at the time. In this sense, positive and negative affect are mutually
exclusive in the same way that a smile and a frown are mutually
exlusive—that is, to the extent that some of the same muscles
and the lips must be engaged in both. It is conceivable that a
person can make a smiling frown or a frowning smile only when
the frown or smile are not making full or intense use of the
muscles involved. In the same fashion, at low levels of intensity,
one might be able to switch back and forth from pleasure to
displeasure imagery, but as the intensity of one type of affect
increases, so does its degree of mutual exclusivity with the other.
This occurs because the common elements needed for an emotion
are increasingly committed to the dominant emotion as intensity
increases, and the nondominant emotion cannot occur without
these elements.
There are several limitations to our studies. We relied solely
on self-report data, and it would be valuable to examine other
types of data in the same way. Do positive and negative facial
expressions, for example, yield findings that are parallel to those
that we found? In addition, our findings could be extended by
the use of experimental designs in which more powerful mood
manipulations are used.
The data on the relations between discrete emotions are interesting,
but preliminary. We used only correlational analyses,
although we found for the other types of data that these were
inadequate to completely understand the relations involved.
However, we thought that a more thorough analysis should await
future studies. First, the discrete emotions in Study 2 were measured
via only a single mood adjective and therefore may be
unreliable. Second, several of these emotions occurred somewhat
less frequently at intense levels than global affect, so that a detailed
analysis would be based on very small numbers of cases. In addition,
no manipulations were carried out in order to alter levels
of specific discrete emotions. Last, it would be desirable to include
a larger sample of emotions in such an indepth analysis. A thorough
examination of the relation between the experience of var-
1038 ED DIENER AND ASHGAR IRAN-NEJAD
ious specific emotions would be a valuable direction for future
research.
A deep rift between those who argue for discrete emotions
and those who postulate an overall structure of emotions has
divided the field of emotion. Our findings present a challenge to
both sides. In terms of emotional structure, our findings indicate
that simple and linear relations as suggested by most dimensional
systems do not fully capture the interrelations among emotions.
On the other hand, there do appear to be systematic relations
between emotions that are not explored in most theories that
postulate the existence of independent discrete emotions. It will
be more important in the long run to uncover why emotions
show certain patterned interrelations than it will be to simply
describe the structure of emotions.
References
Clyde, D. J. (1963). Manual for the Clyde Mood Scale. Coral Gables,
FL: Biometric Laboratory, University of Miami.
Daly, E. M., Lancee, W. J., & Polivy, J. (1983). A canonical model for
the taxonomy of emotional experience. Journal of Personality and Social
Psychology, 45, 443-457.
Diener, E. (1984). Subjective well-being. Psychological Bulletin, 95, 542-
575.
Diener, E., & Emmons, R. A. (1984). The independence of positive and
negative affect. Journal of Personality and Social Psychology, 47, 1105-
1117.
Diener, E., Emmons, R. A., & Sandvik, E. (1985). The dual nature of
happiness: The independence of positive and negative affect. Manuscript
submitted for publication.
Diener, E., Larsen, R. J., Levine, S., & Emmons, R. A. (1985). Intensity
and frequency: The underlying dimensions of positive and negative
affect. Journal of Personality and Social Psychology, 48, 1253-1265.
Diener, E., Sandvik, E., & Larsen, R. J. (1985). Age and sex effects for
emotional intensity. Developmental Psychology, 21, 542-546.
Ekman, P., Friesen, W. V., Ellsworth, P. (1972). Emotion in the human
face. New York: Pergamon.
Epstein, S. (1982). A research paradigm for the study of personality and
emotions. In M. M. Page (Ed.), Personality—Current theory and research:
1982 Nebraska Symposium on Motivation (pp. 91-154). Lincoln:
University of Nebraska Press.
Iran-Nejad, A. (1980). The schema: A structural or a functional pattern
(Tech. Rep. No. 159). Urbana: University of Illinois, Center for the
Study of Reading. (ERIC Document Reproduction Service No. ED
181 449)
Iran-Nejad, A. (1983). Qualitative and quantitative causes of the experience
of affect (Doctoral dissertation, University of Illinois at Champaign-Urbana,
1983). Dissertation Abstracts International, 43, 3847A.
Iran-Nejad, A., Clore, G. L., & Vondruska, R. J. (in press). Affect: A
functional perspective. The Journal of Mind and Behavior.
Iran-Nejad, A., & Ortony, A. (in press). A biofunctional model of distributed
mental content, mental structures, awareness, and attention.
The Journal of Mind and Behavior.
Izard, C. E. (1972). Patterns of emotions. New York: Academic Press.
Laymon, R. (1975). Out of the woods. Ellery Queen's Mystery Magazine,
66, 133-135.
McNair, D. M., & Lorr, N. (1964). An analysis of mood in neurotics.
Journal of Abnormal and Social Psychology, 69, 620-627.
Nowlis, V., & Nowlis, H. H. (1956). The description and analysis of
mood. Annals of the New York Academy of Sciences, 65, 345-355.
Plutchik, R. (1980). Emotion: A psychoevolutionary synthesis. New'Vbrk:
Harper & Row.
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.
Schlosberg, H. (1954). Three dimensions of emotion. Psychological Review,
61, 81-88.
Thurmond, P. J. (1978). If cornered, scream. Ellery Queen's Mystery
Magazine, 71, 66-68.
Watson, D., & Clark, L. A. (1984). Negative affectivity: The disposition
to experience aversive emotional states. Psychological Bulletin, 96, 465-
490.
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., & Tellegen, A. (1985). Toward a consensual structure of
mood. Psychological Bulletin, 98, 219-235.
Zevon, M. A., & Tellegen, A. (1982). The structure of mood change: An
idiographic/nomothetic analysis. Journal of Personality and Social
Psychology, 43, 111-122.
Received April 23, 1985
Revision received October 14, 1985 •
All in-text references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.
Konular
- PARS DERGİSİ
- ŞU’ARÂ HOCASI MÂDER-ZÂD BİR ŞÂİR: ZÂTÎ
- KLÂSİK TÜRK EDEBİYATINDA ŞEM’Ü PERVÂNELER VE LÂMİ’Î ÇELEBÎ’NİN ŞEM’Ü PERVÂNE MESNEVİSİ
- FARS EDEBİYATINDA METAFİZİK YOLCULUKLAR
- شاعران فارسی سرای وفارسینويس ارزرومی
- تعلیم وتربیت ازمنظر سعدی
- توازن موسیقايی غزلهای سعدی
- YAŞAR KEMAL’İN İNCE MEMED ROMANI İLE SADIK ÇUBEK’İN TENGSİR ADLI ROMANININ KARŞILAŞTIRMASI
- YAVUZ SULTAN SELİM’İN DÎVÂNINDA OLMAYAN FARSÇA ŞİİRLERİ
- KÜÇÜKASYA’DA İSLAMİYET (DER İSLAM IN KLEIN ASIEN)
- ERKEN DÖNEM FARSÇA MESNEVİLERDE BEZM - IYRD.
- RÛDEKÎ-Yİ SEMERKANDÎ (Ö. 329/940)
- NAZÎRÎZÂDE EMÎN’İN ŞEYHÜLİSLÂM FEYZULLAH EFENDİ’YE FARSÇA METHİYELERİ
- HAYRETÎ DİVANINDA GEÇEN “GAM” KELİMELERİNİN TASARIMLARI
- BÂBÂ TÂHİR-İ HEMEDÂNÎ DİVANININ MEHDÎ-İ HAMÎDÎ NÜSHASINDA GEÇEN DOBEYTÎLERİ VE TÜRKÇE TERCÜMESİ
- EŞREFOĞLU RÛMÎ’NİN GAZELLERİNDE NASİHAT VE NEFİS MUHASEBESİ
- HÂB-I HAYÂL, AYINTABLI HÜSNÜ
- شاعران فارسیسرای و فارسینويس ارزرومی*
- مسئلة »مضمون« در شعر کودکان و نوجوانان
- مأخذ اصلی تمثیل خورندگان پیلبچهدرمثنوی
- وگرايی درهنر ايران
- NEF’Î’NİN TUHFETU’L-UŞŞÂK ADLI FARSÇA KASİDESİ
- ERKEN DÖNEM FARSÇA MESNEVİLERDE BEZM - IIYRD.
- DAKİKÎ-Yİ TUSÎ (Ö. 366/976)
- ÂRİF ÇELEBİ’NİN FARSÇA KASİDESİ VE TÜRKÇE ÇEVİRİSİ
- SÂİB-İ TEBRİZÎ’NİN ŞİİRLERİNDE GEÇEN “HÂB-I BAHÂR” TAMLAMASI ÜZERİNE
- ROMEN DİLİNDE KULLANILAN FARSÇA KELİMELER
- سینمای ایران
- آداب حرب مغول درتاریخ جهانگشای جوینی
- بررسی تطبیقی ضرب المثل های ترکی سنقر با ضرب المثل های زبان فارسی