Expanding our thinking about discrete emotions and politics

Abstract

In recent decades, political psychologists have given a lot of attention to the role of emotions in politics. While there have been several different research programs, the dominant paradigm has been set by affective intelligence theory (AIT), developed by George Marcus, Russell Neuman, and Michael Mackuen. AIT has helped explain many puzzles in understanding how emotions influence political decisions, as any good paradigm should. At the same time, I argue it has also had the effect of limiting broader research into the range of discrete emotions, especially contempt. While recognizing the value of AIT, I call for more research that goes beyond its boundaries, showing through several recent studies how a focus on the additional effects of contempt can improve our understanding of voter decision-making.

In recent decades, political science, political psychology, and the politics and the life sciences community have made great strides in bringing emotions into the study of politics after essentially ignoring them for most of the first decades of quantitative political behavior research. We now know a lot more than we did about the various ways in which emotions impact political behavior, especially political decision-making. Scholars of emotions in politics have moved us away from a limited cognitive view of political information processing toward a much more comprehensive understanding of decision-making. Like many research areas as they mature, in this time, we seem to have moved past a sort of Wild West of emotions approaches to a dominant paradigm based on affective intelligence theory (AIT; Marcus et al., 2000). AIT has much to recommend it, and I am not here to take it on. But I would like to suggest that the dominant focus on AIT has made examining emotions beyond its original focus on enthusiasm and anxiety (or fear) and its more recent embrace of anger, has limited stuides of a wider range of specific discrete emotions. I would like to suggest that there may be some value in going beyond these “big three” in the political psychology of emotions. I hope by the end of this not to have taken down AIT, but to have reasonably suggested that we might broaden the scope of the emotions we consider.

For some reason, the movie 1776 is one of my all-time favorites. It’s overly long and plays fast and loose with the facts about the debate over the Declaration of Independence, but the scenery chewing is great and the songs are fun. In the movie, John Adams is a hothead, told to “sit down, John” when he’s agitating for a vote on independence. The opposition, led by John Dickinson, is represented by a song, Cool, Cool, Considerate Men. This juxtaposition strongly suggests the long-standing belief that emotions get in the way of rationality. Putting aside the actual lyrics of the song, the point here is that only by maintaining a cool, dispassionate perspective can we be successful. The hotheads—ruled by their emotions—will get us in trouble. But those who can think and act rationally can save the day. This view of emotions as opposing rationality has a long history in Western thought.

Even old Ben Franklin tried to come up with a way to reason through decisions. His method was to give himself time to think about all the pros and cons of an important decision, writing them down in a ledger with pros on one side and cons on the other. He recommends assessing each one’s weight, striking from each side until the final balance is determined (see Figure 1). There is no obvious space for emotion here; it is very much a process of a cool, cool, considerate style. Take time and think it all through.

Figure 1. Franklin’s Rule

The idea of reason versus emotion has certainly been reflected over and over again in popular culture. As children, many of us likely had parents who told us at an early age: don’t be so emotional; think before you act; count to 10 and cool down; and so on. While this was no doubt good advice in a sense, it may have reinforced the idea that decision-making can somehow be divorced from emotions.

Maybe our parents weren’t completely wrong about trying not to be so emotional when making decisions. But they were mostly wrong in a key way: we cannot divorce emotions from decision-making. I probably don’t have to convince those in this audience of that reality. Our emotions are inevitably tangled up with our cognition; the question really becomes how the two interact. We are particularly concerned about politics here, and there can be little doubt that politics is affect driven. Our emotions are not something we can set aside and ignore while we process politics coolly and rationally, even if we wish we could. But politics is about feeling. Thus, to understand political decisions, we must understand the emotions that work in tandem with cognitive information processing.

Parenthetically, I think this is one of the things that seems to get Democrats in the United States in deep trouble. Their default seems to be to play to “reason” and expect voters to think hard about the issues. Republicans seem to know that voters feel about issues more than they think, to put it bluntly. And this is true of all of us, no matter how much we might think we think more than we feel. Unfortunately for Democrats, the idea that we will make better decisions by somehow pushing aside emotions is just not tenable.

As Lerner and colleagues (2015, 799) note, “emotions constitute potent, pervasive, predictable, sometimes harmful and sometimes beneficial drivers of decision making.” This nicely sums up how I see the role of emotions in political decision-making.

Of course, all of this begs the question of what emotions actually are. George Marcus recently presented a paper in which he lays out several theories of emotions and asks to what degree they meet the criteria for a neurologically sound falsifiable theory. His take is that none of the ones he looks at—appraisal, regulation, affective intelligence, or valence—meet all the criteria he sets out. As he noted in his abstract for his 2022 Midwest Political Science Association meeting paper, “I find that the worthy ambition to develop a full theory of emotion awaits fulfillment” (Marcus, 2022).

Marcus is undoubtedly right that this is a worthy endeavor, and one that I hope he will continue to pursue. In the meantime, I will continue here with a much more modest goal—to convince you that political psychology should be considering a wider array of specific emotions than we do now. Affective intelligence—George Marcus, Russell Neuman, and Michael McKuen’s (2000) massive contribution to our understanding of emotions and politics—has become the dominant paradigm in political psychology, if not a full-blown theory. Dominant paradigms become such for a reason. Simply put, they explain puzzles that were not previously explained. But they may not explain new puzzles or puzzles outside their specific domains. I think that is the case here.

Broad roles of emotions

Like many who study emotions in politics, I am going to punt today on the crucial question of what exactly we mean by “emotions.” Lerner et al.’s (2015) statement quoted earlier is not a definition of course. Within psychology, there are many definitions and many disagreements. One that strikes me as reasonable is from Damasio (2011): “emotions are complex programs of actions triggered by the presence of certain stimuli, external to the body or from within the body, when such stimuli activate certain neural systems” (p. 1804).

As someone interested in political behavior, I guess I’ll just admit that while the debate over exactly what emotions are is interesting, I’m more interested in the visible outcomes of how we feel about political stimuli, however those feelings arise. I’m not overly concerned about differentiating words like affect, feeling, and emotion, although people more expert than I can certainly provide excellent guidance. At the same time, I do think it is worth considering a few features of emotions—however defined—before talking more directly about political psychology’s dominant paradigm in emotions study and some of its limits.

Richard Lau and I recently finished the third edition of our chapter on behavioral decision-making for the new edition of the Oxford Handbook of Political Psychology (Lau & Redlawsk, in press). The chapter, now titled “Psychological Approaches to Political Decision Making,” has a brief updated section on emotions (there is an entire chapter dedicated to this in the volume, so we only make a few key points). We start by noting that while positive/negative affect models like hot cognition can help explain basic motivated reasoning processes, feelings are more than just positive or negative, as abundant basic research in social psychology has demonstrated.

In particular, as we think about how politics and emotions interact, there are a range of emotions that need to be considered. Anger, fear, contempt, and disgust are all negative emotions that have identifiable specific effects on politics; likewise positive emotions such as enthusiasm, hope, and pride. They are “automatic effort-free inputs that can strongly influence choices without any awareness or knowledge of where those emotions came from” (Lau & Redlawsk, in press).

Emotions can be integral or incidental, and both can influence decision-making. Some emotions arise from the judgment or choice one has to make at the moment. We’ve all likely experienced these “integral emotions” in recent years, such as feeling anxiety at the thought of eating in a restaurant during the COVID-19 pandemic. In such an example, we might be influenced by rising anxiety to agree only to eat outside, or even to change our mind entirely and make a peanut butter and jelly sandwich at home. Thus, integral emotions can serve as guides, as Wyer, Clore, and Isbell (1999) incorporate in the affect-as-information model. Of course, these automatic affective responses don’t always result in making the right choice. As Lau and I point out, fear of flying can lead to choosing a more dangerous way to get to where we are going, such as driving.

Alternatively, the emotions that arise incidentally from events or activities can unconsciously influence decision-making and information processing (Lerner & Keltner, 2000; Pham, 2007); we see the same with chronic moods (Bernardi & Johns, 2021).

In general, processing styles are influenced by emotions that condition thought content, goal activation, and what and how much information is actually processed. AIT makes this argument, as it describes how positive emotions like enthusiasm are much more limited in information processing than avoidance emotions like fear. Lerner and Keltner (2001) identify the effects of fear compared with anger on risk estimates. Fear makes us more risk averse and increase estimates of risk. The mechanism appears to be the salience it creates for negative consequences. Anger, on the other hand, as an approach emotion, decreases risk estimates. We become more risk seeking when angry, as potential positive outcomes of action are made more salient.

Appraisal theory of emotions

Before continuing on my quest to broaden the scope of the emotions we consider, let me turn briefly to appraisal theory. It is certainly true that appraisal theory may not be the complete theory of emotions that Marcus seeks. He argues that appraisal generally does not recognize the neuroscience underlying how emotions arise, particularly the parallel processing paths that AIT favors. Even so, appraisal theory seems to me to have value in identifying a set of discrete emotions worth considering beyond the three AIT focal points, expanding our understanding of how people make sense of what is happening in the political world.

Kyle Mattes and I prepared a chapter on emotions for the new Cambridge Handbook of Political Psychology (Redlawsk & Mattes, 2022) in which we describe recent research in politics derived from considering a wider range of discrete emotions. I want to use this to make a few points about how we see appraisal theory. The theory is about how we become self-aware of the emotions that we experience. Awareness is generated by generally nonconscious assessments of the environment in which we find ourselves and results in recognizing a range of discrete emotions generated by appraisals across various dimensions. These can include assessments of motivational state or outcome probability. Importantly, the appraisals need not be accurate.

Lerner and Keltner (2000, 2001) propose appraisal-tendency theory, writing that “appraisal tendencies are goal-directed processes through which emotions exert effects on judgement and choice until the emotion eliciting problem is resolved” (Lerner & Keltner, 2000, p. 477). Judgment and choice are core aspects of politics, especially given individual-level models of voting or elite decision-making. This framework seems to do a good job of linking emotions to decision-making where they “serve a coordination role, automatically triggering a set of concomitant responses (physiological, behavioral, experiential, communication) … address[ing] problems quickly” (Lerner et al., 2015, p. 805). The framework explains the connection between different discrete emotions and appraisals made about anticipated efforts and control, certainty, pleasantness, and who should be considered responsible for the outcome. To consider two AIT emotions in this vein, anger has high certainty, low pleasantness, high effort and control, and responsibility attributed to someone else. Fear, on the other hand, can be differentiated by its low pleasantness, certainty, and control, with high effort and unclear attribution.

Roseman’s (2011) emotion system model categorizes 16 of the most-studied discrete emotions in social psychology into four families:

1. 1. All positive emotions are in the Contacting family (as in AIT, there is no differentiation here).

2. 2. Low control, potential negative emotions like fear, sadness, distress, dislike, and regret are in the Distancing family.

3. 3. Rejection family emotions are negative aversive emotions with high control potential. These include disgust, contempt, and shame.

4. 4. Attack emotions like frustration, anger, and guilt are negative, appetitive emotions with high control potential).

More specifically, the significant appraisal dimensions of emotions most relevant to politics are, first, the three AIT-defined emotions:

• • Hope (enthusiasm)—uncertain, appetitive, circumstance-caused, high control potential, positive

• • Fear—uncertain, circumstance-caused, low control potential, negative

• • Anger—other-caused, appetitive, high control potential, negative

There are two other rejection emotions that seem particularly important in politics:

• • Disgust—circumstance-caused, aversive, high control potential, negative

• • Contempt—other-caused, aversive, high control potential, negative

AIT specifically addresses the first three emotions, but it lumps disgust and contempt into an “aversive” anger dimension—more on this momentarily. I just want to set the stage as to why I think we need to consider a wider range of discrete emotions than we have been doing in recent years.

Three emotions programs in political psychology

A few years back, Douglas Pierce and I discussed what we saw as the three basic research programs on emotions in politics that have been most prominent in recent years (Redlawsky & Pierce, 2017). AIT is obviously one of them. The others include Lodge and Taber’s (2013) The Rationalizing Voter, which examines hot cognition and online processing, and Lavine, Johnston, and Steenbergen’s (2012) The Ambivalent Partisan, which challenges assumptions about how emotions influence voters. The rationalizing voter is a voter who is influenced by affect before they even know it, with affect already linked to a person or event (hot cognition) in memory triggered and generating a simple “how do I feel” heuristic. All cognitions carry affect, which is easily extracted from new information, stored, and recalled when an evaluation or choice is called for. This is formalized in the JQ Public model, which has been shown to do a very good job of matching what real human participants do in a lab study of voter decision-making (Kim, Taber, & Lodge, 2010).

Lavine and colleagues react to what they see as too little focus on the fact that many people aren’t actually sure of what they feel, or simply cannot easily determine the direction of their feelings. These people are ambivalent and respond more to the violation of expectations than they do to specific emotions. This work places AIT as simply a specific example in a broader expectancy violation framework. Ultimately, this leads ambivalent voters to be more open to deeper information search and thus less partisan. In some ways, this is similar to the argument about fearful/anxious voters in AIT.

Both of these programs on affective processing bring important ideas to the table even as they have different expectations for how emotions influence decision-making. For JQP, unexpected information does not engender the kind of learning that ambivalence suggests, but instead a motivated reasoning process designed to maintain existing evaluations, rather than being motivated to learn more to perhaps challenge them. This may be the clearest way (besides the lack of detailing specific emotions) that Lodge and colleagues differ from Marcus et al.

Speaking of Marcus, let me (finally) turn to affective intelligence theory, since (1) it has become essentially paradigmatic in political psychology, and (2) its position in the field may be limiting our careful consideration of emotions beyond its three focal points: hope/enthusiasm (and positive emotions generally); fear/anxiety activated automatically by the surveillance system under conditions of threat; and, more recently, anger, which Marcus and colleagues (2000) categorize as one of a group of “aversive” emotions—more on this in a moment.

I am not going to describe AIT in detail here, since this audience is no doubt aware of it. AIT arises from a neuroscience view of emotions that Marcus identifies in work by Gray and others and is supported by new knowledge of the workings of the brain evidenced through functional MRI studies. As Zajonc (1980) noted four decades ago, affect is processed faster than and before cognition; cognitive awareness is slow. Emotions are fast and are generated by stimuli before we become aware of them. They may then go on to influence cognitive processing, which may then feed back into emotional states. Thus, in his 2022 paper, Marcus makes clear that he views AIT as a type of appraisal theory. But it is one that places emotional processing in parallel across all three of its emotion categories simultaneously. It isn’t anxiety to the exclusion of anger, or enthusiasm, it is the levels of each that matter.

In many ways, this all seems quite reasonable.

Yet, I think the success of AIT in becoming our dominant paradigm has its problems. Some of them are related to the ways in which its most recent addition—anger—is theorized. But most are related to its influence in limiting the exploration of emotions beyond its three focal emotions.

Anger and aversion

Ted Brader and George Marcus (2013) define anger as “a cluster of feelings that includes anger, disgust, contempt, and hatred” (p. 179). Other AIT-based students have similarly treated aversion and anger as synonyms (e.g., Capelos, 2013). Yet, as Roseman’s emotion system model argues, and as mundane experience suggests, each has distinct characteristics and effects. Anger is not aversive in either the usual meaning of the word or the Roseman model. Instead, it is an aggressive emotion quite different from disgust or contempt. For example, anger increases information seeking (Ryan, 2012) and support for military action (Huddy et al., 2007). These effects are exactly what AIT argues should happen when people are angry, as they try to address the cause of their anger.

But in classifying anger as an “aversive” emotion, AIT allows it to be conflated with emotions that seem similar but which actually are aversive. This conflation is a problem, as each is distinct in its characteristics and effects. As an aggressive, appetitive emotion, anger is substantially different from the aversive emotions with which AIT groups it. Disgust and contempt move us away from a stimulus instead of confronting it. But angry people take action to try to resolve their anger, moving against that which blocks their goals. Thus, behaviorally, anger is an approach emotion (e.g., Frija, 1986). In politics, one way you hurt the other side is to get out and vote against them. As AIT notes, anger increases people’s motivation to find information bolstering existing attitudes (MacKuen et al., 2010) and decreases motivations to learn about candidates with whom they disagree (Redlawsk et al., 2007). There are plenty of examples of anger’s ability to mobilize. Most recently, the GOP’s efforts to whip up anger at critical race theory in schools may have led to the 2021 Republican gubernatorial victory in Virginia. It was also a stronger predictor than fear of voting against either Barack Obama or John McCain in 2008 (Finn & Glaser, 2010). What is fascinating is that the anger generated by restrictive voters laws was so mobilizing that it offset the effects of the laws (Valentino & Neuner, 2017). Valentino et al. (2011) also find that across various forms of political participation, anger is more mobilizing than enthusiasm, the original mobilizing agent in AIT. However, there may be important conditions—some people may not be mobilized by anger, as Davon Phoenix (2019) argues. Black voters, he finds, are consistently less likely to be willing to express anger compared to Whites. This racial anger gap exists for good reason, but its result is that tamping down of Black anger can also reduce Black turnout, as seen in 2016.

Returning to Roseman’s (2011) model, anger, disgust, and contempt do have similarities in that they all arise under high control potential and certainty is not a factor (as it is for anxiety/fear). But anger stands out in motivating a move against the other, the cause of the anger. Both disgust and contempt instead lead to separation. While disgust is more likely to be about objects than people (though not entirely), contempt is inherently interpersonal. The anticipated behaviors are different as well and the emotions have different facial expressions. Contempt and disgust are in the “rejection” family, while anger is in the “attack family.” So, while anger and contempt (and disgust) occupy similar spaces in Roseman’s model, they have important differences that we must recognize. Politics clearly engages rejection emotions at least as readily as the attack emotions. Anyone who has spent even brief moments watching negative campaign advertising has certainly felt the contempt or disgust that is meant to be engendered toward the target of the ad (Mattes & Redlawsk, 2015). I would argue that these two discrete emotions have particular resonance in the political realm and need more consideration than we have given them so far in the existing research paradigms.

Anger is about unjust action and is outcome focused. Contempt, on the other hand, is trait focused, and it is elicited in part by beliefs that others are inferior. It is focused on the incompetence of others or their stupidity, rather than how goals may be blocked. Perceived unfair treatment will generate anger, not contempt, and these are distinguishable in their consequences for action that might or might not be taken. Politicians, like Donald Trump, can make use of contempt—expressing it themselves or trying to get voters to feel it for an opponent—and in doing so can enhance perceptions of incompetence or corruption or even place those on the other side beneath consideration. The political implications of having voters feel contempt for an opponent differ from what happens when they feel angry. As Mattes and I wrote, “voters may perceive that bad outcomes can be changed, while bad people cannot” (Redlawsk & Mattes, 2022, p. 147). Of course, as with any negative emotion, it is possible to go too far in attempting to generate contempt, as we also discuss. Contempt can easily cross into personal defamation and incivility which may reduce support for the attacker rather than the attacked.

All of this is to say that by lumping together anger and contempt, in particular, AIT has the potential to miss important differences between these emotions and their uses and effects in politics.

Disgust, on the other hand, has been studied in some detail in politics, although not often in the context of how it affects voter turnout and decision-making. Like contempt, disgust is a rejection emotion, but it can be readily differentiated from contempt (Miceli & Castelfranchi, 2018) and shame, the latter being activated when the self or in-group is seen as substandard (Roseman, 2013). Objects of disgust can be human or nonhuman. Damasio (2011) describes disgust as “triggered from a small region of the anterior insular cortex when certain stimuli are present, for instance, the sight of decomposing food or body waste” (p. 1804). One example of a disgust sensitivity study in the voting literature is that by Shook et al. (2017), who find that voting for Obama was conditioned in part on disgust sensitivity; voters with greater sensitivity were less likely to vote for Obama in 2012.

Examining the policy realm, a number of students have found that perceptions of moral purity violations may evoke a disgust reaction in areas such as abortion, same-sex marriage, and even health-related issues (Adams et al., 2014; Clifford & Wendell, 2016; Inbar et al., 2009).

Disgust and contempt are closely related. But contempt has a necessary social component. Rather than a motivation to expel contamination, it motivates pushing away the object of contempt. In politics, contempt is used to remove the opponent from any consideration as an option. Tools like belittling are used—think of Trump’s constant use of contempt as a political strategy. His goal was not necessarily to make people angry at his opponents; it was to make them seem not even worthy of consideration.

Contempt in politics is not new. In fact, it is likely as old as competitive elections themselves. Making your opponent seem lesser in voters’ eyes seems like a potentially useful strategy. Given this, one would expect to find a large body of research. But this would require that the right questions be asked, and for the most part, they simply have not been.

Investigating contempt in voter decision-making

Ira Roseman, Kyle Mattes, and I, along with students of ours, have been investigating the effects of contempt in politics and contrasting it with anger over the past few years. We started this work before Donald Trump began his run for president and made clear his active contempt for all things not Trump. This work arose out of Kyle’s and my book project, The Positive Case for Negative Campaigning (Mattes & Redlawsk, 2015), in which we found that voters are quite accepting of negative advertising, as long as it doesn’t rise to levels of defaming the opponent. This seemed particularly interesting and led us to think about the contemptuous nature of such ads.

To our knowledge, until recently, there were just two voting studies focused on contempt. The American National Election Study (ANES) routinely asks questions about anger, fear, pride, and hope, but not contempt. However, in 1995, an ANES pilot study asked about contempt felt toward presidential candidates. Johnston, Roseman, and Katz (2014) analyzed the relationship between candidate ratings on leadership, specific negative emotions, and feeling thermometer evaluations. They found that for Bill Clinton, both anger and contempt (but not fear) independently mediated the effects of leadership perceptions and thermometer evaluations. In contrast, for Republican candidate Bob Dole, anger and fear (but not contempt) mediated these relationships. Another study by Roseman et al. (2013) asked undergraduates to view excerpts from the 2008 presidential campaign debates between Senator Barack Obama and Senator John McCain. For Democratic candidate Obama, both viewer contempt and anger mediated the relationship between perceived undesirable qualities and thermometer evaluations. However, only anger mediated the same relationship for McCain. Thus, in two different elections, contempt, as well as anger, influenced perceptions of a Democratic presidential candidate, while for Republican candidates, anger, but not contempt, was important. This seems like an important distinction.

Many other politicians besides Trump clearly foment contempt, and its use may be on the rise (Stohr, 2017). Grimmer and King (2011) document that more than a quarter of U.S. senators’ press releases from 2005 to 2007 used “exaggerated language to put them down or devalue their ideas” (p. 2649). A better understanding of when and how contempt works in politics would be valuable. So, we set out to try to get this understanding. But so far, we have been limited to data we could personally collect.

Let me quickly run through three studies we have completed on contempt and its effects in elections, focused on advertising in 2014 Senate races, perceptions of candidate expressions of emotions and voters feelings in the 2016 Iowa caucuses, and the role of contempt in the 2016 general election. I highlight these studies to provide some sense of what we can learn if we go beyond a focus on the three emotion groups of AIT and particularly differentiate contempt from anger.

Since Iowa has been first in the nation, I’ll talk first about the 2016 GOP caucuses in that state. In this study (Redlawsk et al., 2018), we look at voters’ perceptions of the use of contempt by candidates and voters’ own feelings of contempt generated by the candidates. We first found that voters perceived a great deal of contempt (as well as other emotions) being expressed by candidates. Then, using a traditional ANES-style emotions question asking whether candidates “ever made you feel” contempt, voters also reported feeling significant contempt. Recognizing reality on the ground, voters saw Trump express contempt at a level dramatically higher than any other candidate. In fact, perceptions of contempt by Trump rival perceptions of anger. Trump also made GOP voters feel contempt more than any other candidate, although feelings of contempt ran behind generated feelings of anger or hope.

We then examined the effects of four emotions—hope, fear, anger, and contempt—on the vote choice. Those who said that Trump made them feel contempt were significantly less likely to vote for him and more likely to vote for Ted Cruz, controlling for the other emotions. Anger generated by Trump also reduced voting for Trump while increasing vote intentions for Cruz and Marco Rubio. The evidence is that both anger and contempt reduced support for the candidate generating the emotion.

A second study used data from the 2016 Cooperative Congressional Election Study to investigate the role of contempt in the general election (reported in Matteset al., 2018). We asked voters about six discrete emotions: anger, fear, contempt, hope, pride, and enthusiasm. We found that both Trump and Hillary Clinton had a knack for bringing out the negative. Each of the negative emotions for both candidates had a higher score than any of the positive emotions. The most felt emotion in the election was anger generated by Trump, followed by fear and contempt. Reactions to Clinton were similar. They key point is that voters perceived not just anger and fear, but also contempt when given a chance to express that emotion.

We went beyond asking about generally perceived emotions by presenting respondents with two frequently aired real-life negative campaign ads—one related to each campaign—to directly gauge emotional responses to the candidates. We believe a key purpose of many negative ads is to push voters toward holding the opponent in contempt—not just to make voters feel anxious or angry or afraid, but to make them dismissive of the very idea of the opponent. Given the nature of contempt, if candidates were successful in generating it, we expected voters to respond and the effects on candidate support to be potentially greater than those of anger. At the same time, we were also interested in whether Trump and Clinton ads generating contempt might also have resulted in backlash for going too far in their attacks. The goal of these ads appeared to be lowering the viewer’s opinion of the opposing candidate—not only in leadership ability, but also as a person. But did these ads make viewers feel contempt? The results suggest they did. They generated negative responses that were stronger than the emotions expressed before viewing the ads, led not by anger, but by contempt. On average, viewers perceived contempt most in both candidates’ ads, followed by anger and fear.

Finally, we examined opinions about the target of each ad, after watching it, predicting attitudes from the emotions we examined. We found that anger did as expected; when greater anger was perceived in an ad, opinions about the ad’s target became more negative. But, unexpectedly, perceived contempt led to an improved opinion of the target. What should we make of this difference? As we wrote,

Anger seems to make sense—express anger about the target and voters respond accordingly, reducing their support. But contempt appears to be different. The fact that we see a backlash effect may reflect the nature of contempt in politics. Perhaps voters who see a candidate express contempt give more thought to the target (since contempt is an other-regarding emotion) and respond with something akin to “what a jerk, talking about the other candidate that way” resulting in backlash. This mechanism is, of course, speculation on our part, but the existence of differing effects does support our claim of the importance of differentiating the emotions of contempt and anger (Mattes, et al., 2018, p. 110).

Before I close, let me turn briefly to one other study that we have published (Roseman et al., 2020). In this study focused on two U.S. Senate campaigns in 2020. One, in Iowa, was highly competitive. The other, in New Jersey, was not expected to be competitive at all, although the outcome was much closer than most had anticipated. We worked from the expectation that anger among one’s own supporters, as predicted by AIT, should motivate their turnout to “hurt” the other side, which presumably generated the anger. Contempt, on the other hand, can be about aiming at undecideds—swing voters—who might be convinced to remove a candidate from their choice set entirely as unworthy of consideration. Conflating the two means not being able to differentiate their use—not only important to those of us studying politics, but to those actually doing politics.

In examining responses from Iowa and New Jersey voters, we found they perceived contempt more than anxiety or anger in four negative campaign ads or candidates’ statements we randomly showed them. In response, voters’ feelings of contempt, though less intense than feelings of anger, were of the same or greater importance than anger or anxiety in predicting vote choice for three of the four Senate candidates in these elections.

Before we showed respondents the videos, we asked about candidate-generated emotions using the 2012 ANES emotions format. Of the three negative emotions, anger was the most felt, followed by contempt and anxiety (in most cases). As with our other studies, we see contempt as a regular response to candidates, independent of other discrete emotions. After viewing each video, voters were asked to report their emotional responses. In all cases, voters perceived greater contempt than either anger or fear. Reported contempt was higher in the competitive Iowa campaign than in the uncompetitive New Jersey election. Again, we find support for our belief that negative ads are as much or more about generating contempt for the opponent as engaging fear or anger.

When we turned to testing the effects of the emotions on vote preference, we found little happening in New Jersey for contempt, but there were also few effects for anger or fear. While contempt was perceived in the videos, it made little difference to voters in their candidate preference. The only exception was contempt generated by the GOP candidate noticeably reduced his vote preference, another apparent backlash example.

It was different in Iowa. High levels of contempt were perceived in the ads we showed respondents, and that contempt had independent strong effects on vote preference, controlling for anger and fear. More contempt for the GOP candidate clearly moved voters away from her and toward her Democratic opponent. Contempt toward the Democrat had even stronger effects in reducing his support. Interestingly, we found few such effects for either anger or anxiety in Iowa. Contempt appeared to uniquely reduce vote preferences for the candidates who were its targets.

Very briefly, there is one final reason to give a lot more consideration of contempt than we do. It may help us understand the political effects of incivility. In a book chapter we recently published (Roseman et al., 2021), which sadly I do not have time to go into here, we discuss how contempt in politics may uniquely (among the emotions usually studied) connect to incivility. Whether contempt drives incivility or it is the other way around (or both), there is clear evidence that we should be looking closely at the emotions of contempt if we want to understand some of the wellsprings of incivility.

Conclusion

In the end, we can only study the emotions for which we have data. The AIT paradigm has become embedded in political psychology for better or worse, often for better. It has given us a number of valuable insights into how emotions seem to affect voters (and people more generally in politics). But, at the same time, it seems to have limited exploration of emotions outside of anger, fear (anxiety), and enthusiasm (hope). AIT adherents may well suggest this isn’t really a problem, but I suggest that more detailed discrete approaches to emotions might give us some pause in assuming we know what we need to know.

Unfortunately, I don’t think we are moving in the right direction. For many scholars, the ANES represents the pool of data for studying questions such as these. In 2020, the ANES revised the emotions questions, changing the reference point from candidates to the country as a whole. This is unfortunate. Lau and Funck (2022), in a paper presented here at the Midwest Political Science Association, take a comprehensive look at the new items. The paper is worth a read, as it highlights both how these items do and do not work compared to the prior items. Respondents were asked “How hopeful/angry/outraged/afraid/happy/proud/irritated/nervous/worried [they were] about how things are going in the country” (emphasis added). The intention was to measure emotional experiences divorced from the specifics of candidates. The old questions were certainly problematic—in particular, there have been issues of endogeneity—but at least they focused on candidates in an election study.

An underappreciated aspect of this change is that it is more complicated than it seems. The old questions asked whether a candidate ever made one feel an emotion without any reference to who the emotion was to be felt about. For example, “how often would you say Donald Trump made you feel angry … ” (2016 format). An answer indicating anger could be angry at Trump or angry at his opponents, or angry about something else Trump directed the anger toward. This would seem to matter but was never resolved. In particular, with an other-focused emotion such as contempt, knowing the point of reference is important, and the old question did not make this possible.

The new question format actually resolves this problem (“things are going in the country”) but of course drops any reference to who generated the emotion. Also, even with the change in reference, ANES has not significantly changed the set of “emotions” included in the question. The new version used afraid, angry, happy, hopeful, irritated, nervous, outraged, proud, and worried, while the 2016 version included angry, hopeful, afraid, proud, and disgusted. On the positive side, more emotions are in fact included. On the negative side, we see a lost opportunity to broaden what is covered (and in fact, one emotion—disgusted—was dropped). The questions asked are still influenced primarily by AIT, and so unsurprisingly continue to support its basic model without going beyond it.

A chance has been missed to expand the emotions we consider in our primary large-scale data collection tool. For AIT, three categories are likely sufficient. Appraisals would seem to be more complex. I see no reason, though, that both approaches cannot be accommodated; the immediate noncognitive responses that color decision-making and response, along with appraisals of what we find ourselves feeling. But it will continue to be hard to know, since we continue not to ask.

To close, we all know that effectively studying emotions is hard. Those who have done this work have added substantially to our understanding of the political psychology of voting behavior and political decision-making broadly. Important work has been accomplished over nearly four decades by George Marcus and his colleagues, as well as by others working in this domain. We have learned a lot more than we knew when I started out as a graduate student so many years ago. Still, there is always more to learn, and an important way to do so is to remain open to a wider range of possibilities for understanding more about how emotions influence politics.