Mixed Emotions Viewed from the Psychological Constructionist Perspective

Abstract

Feeling bad is one thing, judging something to be bad another. This hot/cold distinction helps resolve the debate between bipolar and bivariate accounts of affect. A typical affective reaction includes both core affect (feeling good or bad) and judgments of the affective qualities of various aspects of the stimulus situation (which can have both good and bad aspects). Core affect is described by a bipolar valence dimension in which feeling good precludes simultaneously feeling bad and vice versa. Judgments of affective quality of opposite valence can occur simultaneously because the stimulus situation has many aspects. Affective reaction can also include an emotional meta-experience, which can, but rarely does, embrace simultaneous emotion categories of opposite valence.

Keywords

ambivalence core affect mixed emotion

Common sense is of two minds on the topic of this special section. On the one hand, common sense says that when you are happy, you are not sad. When you are sad, you are not happy. Semantically, happy is an antonym of sad. Happiness is the antidote to sadness. On the other hand, common sense also says that we can feel ambivalent. We talk of mixed emotions and bittersweet times. We enjoy being frightened by a horror film or being saddened by melancholic music. We can be both happy and sad at the same time.

This contradiction is mirrored in a fundamental controversy—here termed bipolar versus bivariate—concerning the nature of emotion as consciously experienced. The ubiquitous affective valence dimension was traditionally conceptualized as one bipolar continuum with a neutral tipping point between feeling good and bad. Should that dimension be replaced with a bivariate system: two separate dimensions, feeling good as one, feeling bad as the other? That is, does feeling good exclude feeling bad, or can we feel both at the same time? Does inducing a good feeling counteract a bad one, or are the two independent of one another?

Initially, the bipolar view prevailed: psychologists’ introspections indicated that positive and negative feelings were mutually exclusive (Wundt, 1896). Nowlis and Nowlis (1956; to their surprise) found that positive versus negative affect were weakly correlated with each other, showed different correlations with other variables, and were better represented as independent of one another. Such evidence suggested that the bipolar valence dimension should be replaced with a bivariate model. These conclusions of independence, however, were challenged as based on method artifacts and misunderstanding of the predictions of a bipolar model. When the multidimensional nature of affect, systematic and random errors of measurement, response format, time frame of assessment, and item selection were all taken into account, the psychometric case for bipolarity was found strong and the case for independence of seemingly bipolar opposites disappeared (Russell, 1979; Russell & Carroll, 1999a, 1999b). For example, when the affect scales are restricted to semantic opposites, available evidence converged on strong negative correlations; the confidence interval was −.87 to −.92. (Russell & Carroll, 1999b, Table 5). Watson and Tellegen (1999, p. 601) concurred: “we agree that self-rated affect is characterized by a bipolar dimension of pleasant versus unpleasant feeling.”

Larsen and colleagues resurrected a bivariate account: positive and negative affective reactions stem from biologically independent processes, but are reciprocally related under most conditions. Co-occurrence is unstable and short-lived. Larsen and McGraw (2011, p. 14) wrote, “people do not often feel happy and sad at the same time but can do so in bittersweet situations.”

Psychological Constructionist Account

My reconciliation of the bipolar–bivariate controversy begins by faulting the terms in which the debate is articulated. The terms emotion, affect, and feeling each refer to various qualitatively different things. When distinctions within these overly vague, broad, and heterogeneous categories are made, evidence for the bivariate position (cases of ambivalence, mixed emotions, and bittersweet moments) does not contradict the existence of a bipolar valence dimension of primitive affective feelings. Emotion, feeling, and affect are best elevated to chapter titles with no serious scientific job to do and replaced with narrower, better defined concepts (Russell, 2003). My analysis here uses three concepts that are part of a larger framework—psychological construction—for emotion in general and that are defined not by everyday language but stipulatively.

Core Affect

Core affect (CA) is the neurophysiological state that gives rise to the most primitive affective feelings: feeling good, feeling bad, feeling energized, feeling enervated. These are hot feelings rather than cold judgments. A person has only one state of core affect at a time. Whereas the typical emotion is directed at something (angry with someone, afraid of something), core affect per se is not directed at anything. The word mood comes closer than either emotion or affective reaction to conveying the concept of core affect—except that core affect can be intense and can change rapidly and that a person always has core affect, although it is not always attended to. Core affect is part (but not the whole) of an emotional episode, giving it its felt hedonic and activation components.

Perception of Affective Quality (PAQ)

Perception of Affective Quality (PAQ) is a cold judgment as to how an event or object can influence core affect. It is part of evaluations, attitudes, liking, and disliking. PAQ and core affect often go hand in hand: the music perceived as pleasant makes you feel pleasure. Still, PAQ and core affect are separable. Schadenfreude is the pleasure felt upon judging an enemy’s situation to be unpleasant. Envy is feeling bad that another’s situation is so pleasant. Guilt can be feeling bad while doing something judged to bring pleasure. A person can have many PAQs simultaneously.

Emotional Meta-Experience (EME)

Emotional Meta-Experience is one’s perception of one’s own emotional state achieved by categorizing that state. Available categories are those labeled, in English, happiness, sadness, fear, anger, and so on. To have an EME of fear is to perceive one’s current state as resembling the script for fear. Resemblance is a matter of degree, and the components of the script are neither necessary nor sufficient. Indeed, actual danger is not a necessary component for fear, as in the feeling during a horror film. A person can perceive his or her own emotional state as resembling more than one emotion script.

Proposed Theoretical Reconciliation: When Feeling Good and Feeling Bad Are Mutually Exclusive and When Mixed Emotions Occur

Mixed Emotions as Changes in Core Affect

Core affect is best described as including, besides a dimension of activation, a bipolar valence dimension ranging from feeling miserable through neutral to feeling ecstatic. It is core affect that is captured by the common sense notion that feeling good precludes feeling bad.

Like the weather, a person’s core affect changes over time, sometimes slowly (as in moods and certain psychiatric disorders), sometimes rapidly (as in sudden emotions). Feeling good now does not guarantee feeling good tomorrow, or even the next moment. Russell and Carroll (1999a) provided an analysis of time changes in core affect. Even if core affect is perfectly bipolar at each point in time, longitudinal studies of core affect cannot test bipolarity and can show apparent independence of positive core affect from negative core affect, simply because core affect at one time is independent of core affect at a different time.

The role of time can also be seen in the vacillation hypothesis. Kahneman (1992) proposed that ambivalence can seem to occur when we vacillate between two distinct states. As with the perception of the Necker cube, only one state occurs at a time, but quick vacillation can be misreported as the two percepts occurring simultaneously.

Mixed Emotions as Multiple PAQs

A person can have many PAQs simultaneously, even of opposite valence. We can judge an event or object or person as having different affective properties because each has different aspects. Much as a refrigerator can be cold in the freezer but hot in the motor, PAQ can be that the refrigerator is pleasant in its ability to freeze things but worrying in its cost. We can simultaneously judge that the cookie tastes good but is bad for health.

Liking, disliking, and other such attitudes include PAQs. Cacioppo and Berntson’s (1994) evaluative space model was proposed as a model of attitudes. One defining feature of an attitude is that it is directed at an attitude object. Thus “mixed emotions” occur when by “emotion” is meant attitude and when the attitude object has multiple aspects, as they all do.

Mixed Emotions as EME, a Resemblance to Multiple Scripts

An EME is a categorization of one’s current state based on its resemblance to the mental script for a specific emotion, in which core affect is neither necessary nor sufficient. Because resemblance is a matter of degree, one’s current state can resemble various scripts albeit typically to different degrees. Indeed, when people are given a list of emotions and asked to check all those they are experiencing, they check about half (Russell & Mehrabian, 1977). Because resemblance among emotion concepts is dominated by core affect, one’s current emotion will typically resemble scripts similar in valence and arousal. Cases of strong clear intense emotions will not show mixed valences (perceiving oneself as panicked precludes perceiving oneself as serene). Still, it’s possible for one’s current state to resemble scripts of opposite valence. A ride on a roller coaster can be both exciting and scary, meaning that it simultaneously resembles both the fear script and the excitement script. Equal resemblance is rare.

Other Mixed Emotions

Emotional reactions have multiple components, and different components can have different valences. Emotional reactions are typically a mix of core affect, PAQs, an EME, and other components (cognitive, somatosensory, and behavioral), although in varying proportions, including some lacking one or more. A “mixed emotion” can thus appear because core affect, PAQ, and EME are separate processes. As an emotional episode unfolds over time, sometimes rapidly, each of these can change at a different pace. Although in our mental scripts for most emotions all the components fit a stereotypical pattern with the same valence, in reality, components are but weakly correlated. Core affect and behavior might be simultaneous but of different valences: Ordinarily, pleasure is a cue to behavioral approach, displeasure to avoidance. Yet, in certain circumstances, we seek displeasure or avoid pleasure (Tamir, 2011).

One example of mixed emotions is the joy of listening to sad music. Hunter, Schellenberg, and Schimmack (2008) found that the emotion felt while listening to a piece of music could be distinguished from the emotion conveyed by that piece. Liking for the piece is yet another part of the reaction, significantly correlated with the former, but not the latter. Each particular experience of a piece of music is likely unique, but will generally be a composite of a change in core affect, a series of PAQs of various aspects of the piece, liking or disliking of the piece, appreciation of the composer and the performers, and so on. These different parts of an “affective reaction” can all be of the same valence, but they need not be. So, one can have a core affect of sadness, but tremendously like the piece and admire the composer.

Virtual reality allows a mixed emotion in which the simultaneous core affect, PAQ, and behavioral approach–avoidance are of different valences. Violations of the principle of hedonism (that we always seek pleasure) seem common when reality is virtual. Poems, films, television, novels, imagery, music, daydreaming, and other forms of virtual reality are powerful forces on core affect. Some examples of virtual reality create experiences of great displeasure, from sadness to dread. Yet, we approach them, purchase them, and repeat the experience, and we can evaluate the poem, film, and so on as excellent. Much of virtual reality conveys information; novels tell stories, and music conveys emotion. We can like the story of someone’s joy but also of the bad guy’s defeat. Key is knowing that the reality is virtual; a horror film presents no real danger; the imagined fire does not burn anyone. Even very young children distinguish fantasy from reality. A typical reaction to virtual reality is thus a complex event not well characterized as having but one valence.

Review of Evidence

A series of experimental studies was aimed at establishing the existence of mixed emotions. Representative experiments are detailed in Table 1. Here I argue that the empirical results of those studies are consistent with the formulation I presented in the previous lines.

Table 1.

Selected examples of experimental studies purported to show mixed emotions.

Reference	Measure of affect	Control		Experimental		Comment
		Stimulus	Results	Stimulus	Results
Larsen, McGraw, & Cacioppo (2001, Study 1)	Separate strictly unipolar self-report scales of happiness and sadness “right now.”	Before film	10% of 70 participants checked yes to both happy and sad.	After viewing Life is Beautiful.	44% of 107 participants checked yes to both happy and sad.	Because the affect measure was taken as participants left a theatre, it likely included an assessment of judgments about the film. Thus both core affect and PAQ might have occurred. The words happy and sad denote prototypical emotions, and so some of the participants might have reported EMEs.
Larsen, McGraw, & Cacioppo (2001, Study 2)	Separate strictly unipolar self-report scales of happiness, sadness, and bittersweet affect “right now.”	Typical day	16% of 92 participants checked yes to both happy and sad.	Day moving out of college dormitory.	54% of 100 participants checked yes to both happy and sad. Mean bittersweet affect ratings did not differ significantly from typical day.	The affect measure likely also included some assessment of the loss of college and the opportunities of the future. Thus both core affect and PAQ might have occurred. The words happy and sad denote prototypical emotions, and so some of the participants might have reported EMEs.
Larsen, McGraw, & Cacioppo (2001, Study 3)	Separate strictly unipolar self-report scales of happiness, sadness, bittersweet affect, and ambivalence “right now.”	Typical day	20% of 49 participants checked yes to both happy and sad.	Graduation day at college.	50% of 66 participants checked yes to both happy and sad. Mean bittersweet affect, but not ambivalent, ratings differed significantly from typical day.	The affect measure likely also included some assessment of the graduation. Thus both core affect and PAQ might have occurred. The words happy and sad denote prototypical emotions, and so some of the participants might have reported EMEs.
Larsen, McGraw, Mellers, & Cacioppo (2004, Study 1)	Separate strictly unipolar verbal scales of positive and negative reaction to the outcome, scored 0 to 6. Measure of simultaneous positive and negative was the minimum of the two.	Outright wins and losses during 16 gambles.	Mean minimum rating (score on lower of two ratings) = 0.0.	Disappointing wins and relieving losses during 16 gambles.	Mean minimum rating (score on lower of two ratings) varied from 0.3 to 0.8. 60% of subjects showed greater minimum scores in experimental than control condition.	Participants rated their emotional feelings about the outcome of the gamble. Thus, the measure likely included PAQ.
Larsen, McGraw, Mellers, & Cacioppo (2004, Study 2)	Button press	Announcement of win vs. loss; 16 gambles; 3-sec. window.	Mean duration of simultaneously pressing both buttons (= 0.1 to 0.2 sec.) did not vary significantly with alternative outcomes.	Disappointing wins and relieving losses during 16 gambles.	Mean duration of simultaneously pressing both buttons (= 0.5 to 1.2 sec.) varied with alternative outcome. 70% of subjects showed greater duration during experimental condition.	Participants rated emotional feelings about the wins and losses. Thus, the measure likely included PAQ.
Larsen & Stastny (2011)	Button press	48 musical excerpts each with either consistently happy or consistently sad elements.	3.9% of time both buttons pressed.	48 musical excerpts each with mixed happy and sad elements heard simultaneously.	5.8% of time both buttons pressed.	Participants rated the music. Thus the measure was of PAQ.
Larsen & McGraw (2011, Study 1a)	Button press	20 minutes of film clips with some happy and some sad scenes.	Mode and median was 0% of time pressing both buttons.	20 minutes of film clips with bittersweet scenes.	Mode was 0% pressing both buttons. Median = .41 sec.	Participants rated the events depicted in the film. Thus the measure was of PAQ.
Larsen & McGraw (2011, Study 4)	Retrospective free listing of emotions experienced plus indication whether positive and negative emotions occurred simultaneously or at different times.	Either no film clip or a clip with nonoverlapping happy and sad moments.	In both control conditions, 0% reported simultaneous mixed emotions.	Clip from Life Is Beautiful that contains bittersweet moments.	24% reported simultaneous mixed emotions.	Participants rated the events depicted in the film. Thus the measure was of PAQ.
Larsen & Green (2013, Study 1)	Button press	20 film scenes without bittersweet elements.	Median time spent pressing both buttons = 0.0. 11% of participants pressed both buttons at some point.	20 film scenes with bittersweet elements.	Median time spent pressing both buttons = 0.0. 40% of participants pressed both buttons at some point.	Participants rated the events depicted in the film. Thus the measure was of PAQ.
Larsen & Green (2013, Study 2)	Button press	One 60-sec. film scene without bittersweet elements.	Median time spent pressing both buttons = 0.0.	One 60-sec. film scene with bittersweet elements.	Median time spent pressing both buttons = 3.5 sec.	Participants rated the events depicted in the film. Thus the measure was of PAQ.

Note. Button press measure consists of two buttons; participants are instructed to press one button if feeling happy, the other button if feeling sad, neither button if feeling neither happy nor sad, or both buttons if feeling both happy and sad. The buttons could also be labeled “feel good” and “feel bad.”

Larsen and colleagues compared an experimental condition with a control condition. Co-occurrence of positive and negative affect in the control condition was expected to be rare rather than nonexistent because of measurement error—shown to be important in the psychometric studies. Statistical comparison of control with experimental groups thus had the considerable advantage of taking measurement error into account.

Researchers also developed measures of affect that are sensitive to time. When a participant completes a questionnaire measure of affect, the time period for the affect is often vaguely defined. Even specifying “right now” might be interpreted as asking for a report of how one feels right now about an entire film or entire day or whatever is salient. To overcome this obstacle, J. T. Larsen, McGraw, Mellers, and Cacioppo (2004) used a button press. The participant was told to hold down one button when feeling happy, a different button when feeling sad, neither button when feeling neither happy nor sad, and both buttons when feeling both happy and sad.

A third feature was that the researchers carefully examined various alternative explanations for co-occurrence and carried out empirical studies to rule them out. J. T. Larsen, McGraw, and Cacioppo (2001) considered the possibility that watching the film Life Is Beautiful in their experimental condition induced more systematic and random errors than were induced in the control condition, thus creating the appearance of simultaneous happy and sad reactions as an artifact. Analysis of response items other than happy and sad, however, argued against that interpretation.

Rarity of Co-Occurrence

Results in the control conditions were as expected. As Table 1 shows, the typical response was either positive or negative but not both, most of the time. There were some co-occurrences, but authors of these studies consistently interpreted the rare occurrence of both as likely due to measurement error. This repeated finding is consistent with the correlational results supporting bipolarity.

Surprisingly, even in experimental conditions designed to create co-occurrence of positive and negative affect, a substantial number of participants failed to report co-occurrence. J. T. Larsen et al. (2001) found that 56%, 46%, and 50% of participants in experimental conditions in their three studies, respectively, failed to report such co-occurrence. In a later series of six studies, J. T. Larsen and McGraw (2011, p. 13) pointed to similar data:

At first glance, the results of all six studies indicate that the incidence of mixed emotions is low. In Studies 1b and 3, for instance, the majority of subjects never reported mixed emotions for so much as a second over the course of a nearly 20-min film clip. In Studies 4–6, 84% did not report experiencing simultaneously mixed emotions after the clips. Thus, our data may provide the strongest evidence to date that people cannot feel happy and sad at the same time. A closer look at the data, however, provides even stronger evidence for a more nuanced conclusion.

That more nuanced conclusion was that significantly more participants showed co-occurrence in the experimental than control conditions.

When Co-Occurrence Was Observed

In every study of Table 1, co-occurrence of positive and negative affect was reported significantly more frequently under bittersweet circumstances. The first question is what is co-occurring with what. My interpretation is that what co-occur are likely PAQs or PAQs with CA. This interpretation is given in Table 1 for specific experiments. The common central feature of all the experimental conditions in Table 1 was that a stimulus event was highly salient. Participants were leaving a movie theatre just having seen a powerful film, were moving out of their college dormitory, were graduating from college, had just won or lost a gamble, were listening to musical excerpts, or were watching intense film clips. J. T. Larsen et al. (2004) instructed participants to rate how they felt “about the outcome.” The salience of the stimulus and the instructions in some of the studies made it likely that at least some participants in the experimental conditions included in their report their perception of the affective quality of the stimulus (a PAQ). At least some participants perceived the win, loss, film, or music to have both positive and negative aspects, as they were indeed designed to do.

J. T. Larsen and McGraw (2011) themselves raised the question whether the affect assessed in their research might sometimes include a PAQ. They rejected that interpretation by citing Andrade and Cohen (2007):

It is especially unlikely that subjects in a recent study of mixed emotions by Andrade and Cohen (2007) reported their perceptions of affective quality. In that study, continuous [Evaluative Space Grid] ratings indicated that fans of horror films experienced mixed emotions of happiness and fear while they watched horror clips. The horror clips were terrifying and contained no happy material, so it is doubtful that subjects merely reported their appraisals of the clips’ emotional content. A more plausible explanation is that subjects genuinely experienced happiness and, by extension, mixed emotions of happiness and fear. Thus, the most parsimonious interpretation for our studies is that subjects who watched the bittersweet Life Is Beautiful clip reported mixed emotions of happiness and sadness because they genuinely experienced mixed emotions of happiness and sadness. (2011, p. 1108)

There is an alternative interpretation of Andrade and Cohen’s (2007) results. In their Study 1, the measure of those feelings was the PANAS (Watson & Tellegen, 1985). PANAS does not assess the bipolar valence dimension. Watson and Tellegen (1999) acknowledged this point and therefore renamed the PA scale as “positive activation” and the NA scale as “negative activation” to emphasize that each contains a component of activation. That shared component accounts for the typically weak correlation between the two scales and precludes their use in tests of bipolarity.

Andrade and Cohen’s (2007) Study 2 used a different affect measure. Participants watched 4.5 minutes of a horror film followed by a short neutral scene. The affect measure was a grid with two dimensions, with one labeled happy, joyful, and glad and the other labeled afraid, scared, and alarmed. Participants who disliked horror films (fear-avoiders) reported considerable fear during the clip but negligible happiness. Their ratings were consistent with a bipolar account. However, participants who like horror films (fear-approachers) yielded results not consistent with bipolarity. They reported considerable fear and, simultaneously, more happiness than the fear-avoiders. The question is just what the affect measures measured. Terms such as happy and afraid tap EMEs, and so some fear-approachers might have reported that their emotional state did, to a small degree, resemble scripts for happy/joy/glad in that they were watching the kind of film that they were glad to watch. The word glad might have suggested a reaction to the experiment and the film, feeling glad that the experiment involves watching the kind of film that they like to watch.

It might be argued that the measures used in the studies of Table 1 tapped core affect exclusively. Even so, these studies do not necessarily speak against bipolarity because the results might reflect vacillation (Kahneman, 1992). The measures included might have been insufficiently sensitive to the time period to rule out a vacillation between positive and negative core affect. That is, each vacillation could produce a short period in which the measure would appear to show co-occurrence. J. T. Larsen and McGraw (2011, Study 1a) sought to examine this vacillation hypothesis. They used their button press measure and assessed how much time both buttons were pressed for more than 1 sec. More time was spent holding down both buttons during the experimental clip than during the control clip. The experimental clip allowed a large number of occasions for vacillation. The control condition did not. The control clip consisted largely of two sections, a happy one followed by a sad one, with perhaps only one opportunity to vacillate from the one to the other. It is reasonable to assume that the more vacillations, the more opportunities there are for holding down both buttons and hence more time spent doing so. Roughly, the number of vacillations would be proportional to the amount of time spent holding down both buttons. If so, this control condition did not provide an adequate comparison for how much time is spent holding down both buttons when the person vacillates frequently.

The vacillation hypothesis was explored in two studies by Carrera and Oceja (2007) with a new measure more sensitive to time. More reports of simultaneous opposite values of core affect were found with longer periods of time and may disappear when the time periods are sufficiently short.

Conclusion

My review of the evidence here has not resolved the bipolarity–bivariate debate. Both sides remain viable. And that is a good thing, for the debate has been most useful. The debate has forced the development of new measures, and further development of measures is needed. Retrospective measures are subject to distortions of memory; on-line measures are reactive, influencing what is to be measured.

The debate also led researchers to be more specific. J. T. Larsen et al. (2001), Russell and Carroll (1999a, 1999b), and Watson and Tellegen (1999) converge in predicting that in most circumstances a person feels either good or bad but not both. Different predictions occur only for special circumstances, what Larsen called bittersweet situations. Such moments occur when a stimulus event is highly salient (a powerful film, college graduation, and so on).

The models further agree—and empirical demonstrations show—that, if we speak loosely, ambivalence, mixed emotions, and bittersweet times occur. The key remaining issue on “mixed emotions” is conceptual. The concepts of emotion, feeling, affect, and mood continue to be used as if they are adequate scientifically when in fact they are vaguely defined folk concepts. These terms oversimplify a person’s affective life and result in conceptual confusion. Conceptual advances and establishment of empirical generalizations are not independent processes, but complementary. Each requires the other. On my conceptualization, “mixed emotions” typically include a mix of PAQs, a change in core affect, and often an EME, all potentially different in valence. The position here advocated is thus not bivariate but multivariate and bipolar.

Footnotes

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

References

Andrade

E. B.

Cohen

J. B.

(2007). On the consumption of negative feelings. Journal of Consumer Research, 34, 283–300. doi:10.1086/519498

Cacioppo

J. T.

Berntson

G. G.

(1994). Relationship between attitudes and evaluative space: A critical review, with emphasis on the separability of positive and negative substrates. Psychological Bulletin, 115, 401–423.

Carrera

Oceja

(2007). Drawing mixed emotions: Sequential or simultaneous experiences? Cognition and Emotion, 21, 422–441. doi:10.1080/02699930600557904

Hunter

P. G.

Schellenberg

E. G.

Schimmack

(2008). Mixed affective responses to music with conflicting cues. Cognition & Emotion, 22, 327–352.

Kahneman

(1992). Reference points, anchors, norms, and mixed feelings. Organizational Behavior and Human Decision Processes, 51, 296–312. doi:10.1016/0749–5978(92)90015-Y

Larsen

J. T.

Green

J. D.

(2013). Evidence for mixed feelings of happiness and sadness from brief moments in time. Cognition & Emotion, 27, 1469–1477.

Larsen

J. T.

McGraw

A. P.

(2011). Further evidence for mixed emotions. Journal of Personality and Social Psychology, 100(6), 1095–1110.

Larsen

J. T.

McGraw

A. P.

Cacioppo

J. T.

(2001). Can people feel happy and sad at the same time? Journal of Personality and Social Psychology, 81, 684–696. doi:10.1037/0022–3514.81.4.684

Larsen

J. T.

McGraw

A. P.

Mellers

B. A.

Cacioppo

J. T.

(2004). The agony of victory and thrill of defeat: Mixed emotional reactions to disappointing wins and relieving losses. Psychological Science, 15, 325–330. doi:10.1111/j.0956–7976.2004.00677.x

10.

Larsen

J. T.

Stastny

B. J.

(2011). It’s a bittersweet symphony: Simultaneously mixed emotional responses to music with conflicting cues. Emotion, 11, 1469–1473.

11.

Nowlis

H. H.

(1956). The description and analysis of mood. Annals of the New York Academy of Sciences, 65, 345–355.

12.

Russell

J. A.

(1979). Affective space is bipolar. Journal of Personality and Social Psychology, 37, 345–356.

13.

Russell

J. A.

(2003). Core affect and the psychological construction of emotion. Psychological Review, 110, 145–172. doi:10.1037/0033–295X.110.1.145

14.

Russell

J. A.

Carroll

J. M.

(1999a). On the bipolarity of positive and negative affect. Psychological Bulletin, 125, 3–30. doi:10.1037/0033–2909.125.1.3

15.

Russell

J. A.

Carroll

J. M.

(1999b). The phoenix of bipolarity: Reply to Watson and Tellegen. Psychological Bulletin, 125, 611–617.

16.

Russell

J. A.

Mehrabian

(1977). Evidence for a three-factor theory of emotions. Journal of Research in Personality, 11, 273–294.

17.

Tamir

(2011). The maturing field of emotion regulation.Emotion Review, 3, 3–7.

18.

Watson

Tellegen

(1985). Toward a consensual structure of mood. Psychological Bulletin, 98, 219–235.

19.

Watson

Tellegen

(1999). Issues in the dimensional structure of affect—Effects of descriptors, measurement error, and response formats: Comment on Russell and Carroll (1999). Psychological Bulletin, 125, 601–610.

20.

Wundt

(1896). Outlines of psychology. Leipzig, Germany: Engelmann.