Emotions Emerge from More Basic Psychological Ingredients: A Modern Psychological Constructionist Model

Psychological Constructionism Past to Present: A Model for Explaining Variability in Emotion

As indicated by this special section, the field of psychology is currently experiencing a surge in psychological constructionist models of emotion (also see Barrett, 2006b; Clore & Ortony, 2013; Cunningham, Dunfield, & Stillman, 2013; Cunningham & Kirkland, 2012; Lindquist & Barrett, 2012; Lindquist & Gendron, 2013; Roy, Shohamy, & Wager, 2012; Russell, 2003; Wilson-Mendenhall, Barrett, Simmons, & Barsalou, 2011). This surge is attributable to two sources. First, the body of evidence amassed over the last century has led researchers to conclude that emotions are not natural kinds. Recent reviews demonstrate that emotions such as “anger,” “disgust,” “fear,” etcetera, are not associated with consistent and specific peripheral physiological responses (Barrett, 2006a; Cacioppo, Berntson, Larsen, Poehlmann, & Ito, 2000; Mauss & Robinson, 2009), facial muscle movements (Barrett, 2006a; Russell, Bachorowski, & Fernández-Dols, 2003), vocal patterns (Barrett, 2006a; Russell et al., 2003), or behaviors (Baumeister, 2012). Second, several decades of neuroscience research have amplified doubt that emotions are natural kinds. Emotions do not correspond consistently and specifically to anatomically given neural circuits (Barrett, Lindquist, Bliss-Moreau et al., 2007; LeDoux, 2012a, 2012b; Touroutoglou, Lindquist, Hollenbeck, Dickerson, & Barrett, 2013) or functional brain activity (Lindquist, Wager, Kober, Bliss-Moreau, & Barrett, 2012). These data simultaneously point to the involvement of other psychological processes (e.g., memory/semantic knowledge; Kober et al., 2008; Lindquist, Wager, Kober et al., 2012) in emotions that cannot be accounted for by natural kinds approaches. Together, the data suggest that it is time for psychology to adopt a new explanatory framework for the mental states that people call “emotions.”

Psychological Constructionism as a Reaction to Natural Kinds Approaches in the 19th Century

The recent rise of psychological constructionism seems unprecedented, unless you take a careful look at the history of the emotion literature. Psychological constructionist models often emerged throughout history to explain validation failures of natural kinds models of emotion. Indeed, James’ own psychological constructionist model was crafted in contrast to Darwin’s writings on emotion (1872/1965). Darwin argued that certain mammalian behaviors reflect inherited mental states that he named with English emotion categories. ² In reaction, James (1890) argued that “the trouble with emotions in psychology is that they are regarded too much as absolutely individual things. But if we regard them as products of more general causes . . . then the mere distinguishing and cataloguing becomes of subsidiary importance” (James, 1890, p. 449). To James’ mind there are as many varieties of emotions as shapes of rocks on a New England farm (with the implication that detailing the varieties of either was equally trivial). James’ contemporary, Wilhem Wundt (Wundt, 1897/1998), argued that emotions are not modular packages, but mental “compounds” comprised of more basic psychological elements. Foreshadowing the constructionists to follow, Wundt hypothesized two basic elements of the mind: simple feelings (possessing hedonic tone and activation) and a form of interpretation he called “ideation.” Unfortunately, James and Wundt possessed a limited methodological toolbox (introspection only got them so far in hypothesis testing) and psychological constructionist ideas soon lost influence.

The early 20th century was marked by the emergence of several natural kinds approaches, which argued that emotions derive from specific neural structures (Cannon, 1921) and produce specific and consistent outcomes (e.g., facial and physiological responses, Allport, 1924; Cannon, 1921; overt behaviors, Watson, 1919). These approaches incorporated Darwin’s ideas about evolution and hypothesized, for instance, that certain facial behaviors were specifically associated with certain emotion categories because they served an adaptive function (Allport, 1924). In response, the next wave of psychological constructionism (Duffy, 1934, 1941; Dunlap, 1932; Harlow & Stagner, 1932; Hunt, 1941) pointed out that emotions were highly variable events and lacked sufficient facial, physiological, or behavioral regularity to give evidence of natural kinds categories such as anger, disgust, fear, etcetera (Duffy, 1934, 1941; Dunlap, 1932; Hunt, 1941). To account for such variability in emotion responding, early constructionists instead concluded that emotions are states tailor-made to a given context, which emerge when more elemental processes such as basic hedonic feelings (Harlow & Stagner, 1932) or feelings of arousal (Duffy, 1941, 1957) are made meaningful using cognitive interpretation (Duffy, 1941, 1957; Dunlap, 1932; Harlow & Stagner, 1932).

Following a brief wave of behaviorism (during which emotion was mostly out of vogue as a scientific topic), history repeated itself again, and a new crop of natural kinds and psychological constructionist views on emotion emerged. The natural kinds models again argued that emotions emerged from dedicated mechanisms in the nervous system (e.g., “facial affect programs”; Tomkins, 1962) and had consistent and specific facial, physiological, experiential, and behavioral concomitants (Ekman, 1972; Izard, 1971; Tomkins, 1962). Contemporary psychological constructionist models argued that emotions were underlain by basic affective dimensions and that the situation determined how affective feelings were experienced as instances of different emotions (Mandler, 1975; Schachter & Singer, 1962). Although this period witnessed the first empirical evidence for a psychological constructionist view (Schachter & Singer, 1962), the natural kinds approach overwhelmingly held sway for the next 40 years or so, with researchers continuing the search for the peripheral physiological (e.g., Ekman, Levenson, & Friesen, 1983), facial (e.g., Ekman, Frank, & Ancoli, 1980; Matsumoto, 1990), vocal (e.g., Scherer, Banse, Wallbott, & Goldbeck, 1991), and then neural (e.g., Sprengelmeyer et al., 1996) basis of natural kinds of emotion. As in the beginning half of the century, other researchers critically questioned whether the data were sufficient to validate natural kinds models of emotion (Fernández-Dols, Sánchez, Carrera, & Ruiz-Belda, 1997; Ortony & Turner, 1990; Russell, 1980, 1983, 1994, 1995; Turner & Ortony, 1992), but the approach remained strong even in the face of conflicting evidence.

Natural Kinds, Appraisal Models, and Psychological Constructionism: Competing Views in the Late 20th Century

The sustained success of the natural kinds approach during the latter half of the 20th century can be attributed in part to the (re)emergence of causal appraisal approaches during that time. ³ Although the first causal appraisal model of emotion emerged early in the history of psychology (Irons, 1894), and again in the mid-20th century (Arnold, 1960), causal appraisal approaches really began to exert an influence after the cognitive revolution of the 1970s (for a discussion, see Gendron & Barrett, 2009). According to causal appraisal approaches, a cognitive appraisal (Arnold, 1960; Irons, 1894; Lazarus & Folkman, 1984), or a suite of appraisals (Roseman, 1984, 1991; Scherer, 1984; Smith & Ellsworth, 1985) makes meaning of the stimulus situation, which in turn triggers the emotion. As per basic emotion approaches, many early causal appraisal approaches assumed that the resulting emotion triggers a set of consistent and specific physiological changes, facial muscle movements, behavior, feelings, and so on (e.g., when a person appraises that a dark alley is uncertain, unpleasant, and that she lacks control of the situation, she might experience fear which in turn generates an increased heart rate, sweating, widened eyes, and the tendency to run away; for a discussion on causal appraisal models, see Gross & Barrett, 2011). This hypothesis amounts to the idea that emotions are natural kinds by virtue of homology: all instances of the same category (e.g., anger) emerge from the same causal mechanism (a specific appraisal[s]). By hypothesizing that the appraisal is an intervening mechanism between the stimulus and emotion, causal appraisal models made it easier to accommodate evidence of variability in physiological, facial, and behavioral patterns into the natural kinds framework: such variability could occur because different people appraise the same stimulus in a different way, and thus experience different emotions (Roseman, 2011). ⁴ The historical effect was to shift the empirical emphasis away from observing within-emotion category similarity in objective measurements (of physiology, facial muscle movements, vocal acoustics, behavior), and towards observing the cognitive dimensions that characterized emotions in self-report (e.g., Roseman, 1991; Smith & Ellsworth, 1985).

At first blush, causal appraisal models appear similar to psychological constructionist approaches, especially since both can accommodate greater variability in emotional responding than basic emotion approaches. Both causal appraisal and psychological constructionist approaches also hypothesize a role for “interpretation” in emotion, leading to the false perception of similarity. Yet causal appraisal approaches differ from psychological constructionist approaches in two important ways. First, causal appraisal approaches view appraisals as a specific mechanism that is itself distinct from the emotion (for a review of causal appraisal models, see Ellsworth & Scherer, 2003; see Figure 1). To the contrary, psychological constructionist approaches assume that the interpretive component is constitutive of the emotion—just as specific values of hue, brightness, and saturation are constitutive of perceiving colors such as “cardinal red” versus “claret” versus “poppy.” Second, causal appraisal models assume that emotions include distinct steps: appraisals evaluate the stimulus situation (see Ellsworth & Scherer, 2003), which then causes the emotion, which causes associated bodily changes. To the contrary, psychological constructionist approaches hypothesize that an interpretive process makes meaning of on-going bodily changes in light of information about the stimulus situation (see Figure 2). Neither bodily changes nor interpretation are thought to “turn on” in a given context, but instead are ever-present processes that are continually interacting with one another in a conscious brain.

OPEN IN VIEWER

Figure 1.

Causal appraisal models hypothesize that the cognitive appraisal (or a suite of appraisals) of the stimulus situation triggers a discrete emotion. In this regard, they are similar to basic emotion approaches, and are natural kinds approaches by nature of the fact that they hypothesize that all emotions of the same category (e.g., anger) have shared homology.

OPEN IN VIEWER

Figure 2.

Psychological constructionist approaches hypothesize that core affective changes (caused in part by interactions with the stimulus situation) are made meaningful using representations of prior experiences that are tied to the context.

Further blurring the line between appraisal and psychological constructionist approaches, recent “constitutive appraisal” approaches have moved away from making strong causal hypotheses about the role of appraisals in emotion (e.g., Clore & Ortony, 2000, 2008; Moors, 2013; Scherer, 2009a, 2009b). Like causal appraisal approaches (and unlike psychological constructionism), constitutive appraisal models (Moors, 2013; Scherer, 2009a, 2009b) still assume that emotion categories refer to distinct states with specific functional importance. Yet unlike causal appraisal approaches (and like psychological constructionism), constitutive appraisal approaches (e.g., Clore & Ortony, 2000, 2008; Scherer, 2009a, 2009b) highlight the “informational” (Moors, 2013) content of appraisals rather than viewing them as mechanisms of the emotion, per se. Some constitutive appraisal models (e.g., Clore & Ortony, 2000, 2008) are thus quite similar to psychological constructionist approaches (e.g., Kirkland & Cunningham, 2012; Lindquist & Barrett, 2008a) because they view appraisals as descriptions of what it is like to experience an emotion (for discussions, see Gross & Barrett, 2011; Lindquist, Wager, Bliss-Moreau, Kober, & Barrett, 2012). ⁵ For constitutive appraisal and psychological constructionist approaches to remain theoretically distinct, researchers taking a constitutive appraisal approach thus need to be clear about what the extra level of psychological description (i.e., the appraisals) contributes to an understanding of emotion above and beyond the domain-general conceptual knowledge hypothesized by constructionist accounts. If appraisals are merely conceptual content (i.e., what it is like to experience an emotion), then it is more parsimonious to hypothesize a domain-general conceptual process that is a basic ingredient of the mind and contributes to all varieties of mental states (as in Lindquist & Barrett, 2012).

A Modern Psychological Constructionist Approach

My own lab is part of the endeavor to validate a psychological constructionist model that can explain the great variability in emotional experience and perception that exists both within and between individuals. Like previous constructionist models, the approach we take emerged from the observation that 100 years of psychological research has yet to identify the discrete bodily (Barrett, 2006a; Cacioppo et al., 2000; Mauss & Robinson, 2009), facial (Barrett, 2006a; Cacioppo et al., 2000; Mauss & Robinson, 2009; Russell et al., 2003), behavioral (Barrett, 2006a; Mauss & Robinson, 2009), or neural (Barrett & Wager, 2006; Kober et al., 2008; Lindquist, Wager, Kober et al., 2012) basis of English emotion categories such as “anger,” “disgust,” “fear,” “happiness,” and “sadness.” Rather than revealing evidence for natural kinds of emotion, measures of the peripheral nervous system (e.g., cardiovascular reactivity, skin conductance, respiration) instead indicate whether a person is feeling positive or negative, activated or deactivated, or whether the person will approach or avoid something (Barrett, 2006a; Cacioppo et al., 2000; Mauss & Robinson, 2009). Facial muscle movements, as measured by electromyography, indicate whether someone feels positive or negative (Barrett, 2006a; Cacioppo et al., 2000; Mauss & Robinson, 2009). Brain activity that is correlated with emotion experience and perception does not reveal evidence for modular brain regions that have consistent and specific activity during instances of any one emotion category (Lindquist, Wager, Kober et al., 2012). Instead, these brain regions include areas that are involved in more basic psychological processes such as visceromotor control, the representation of hedonic value, the representation of concepts and knowledge, and executive control (Kober et al., 2008; Lindquist & Barrett, 2012; Lindquist, Wager, Kober et al., 2012). The networks that support these more basic psychological processes are also involved during myriad other mental states (including “cognitions”; Oosterwijk et al., 2012) and comprise the intrinsic functional organization of the human brain (Lindquist & Barrett, 2012).

The scientific observation that specific bodily, facial, vocal, behavioral, or brain activity does not correlate with discrete emotions appears to be inconsistent with the (very real) experiences that all healthy human beings have on a daily basis: discrete emotions appear on the faces and bodies of our spouses, children, and even pets, and characterize our own physical and mental reactions to the world. Psychological constructionism thus describes the mechanisms by which basic psychological ingredients combine to give us such discrete perceptions and experiences of emotion. According to our approach, emotions emerge in consciousness when people categorize ambiguous internal (i.e., bodily) and external (i.e., visual, auditory, tactile, etc.) sensations as instances of discrete emotion categories (e.g., anger). This could easily be called a “perception” of the body insofar as every visual perception involves making meaning of ambiguous visual sensations through the use of context (Bar, 2009). ⁷ Because emotion construction is so heavily dependent on the context, emotions are sometimes referred to as situated conceptualizations (cf. Barrett, 2009; Barsalou, 2009; see Figure 2), meaning that the experiences are tailor-made for interacting with current situations. ⁸ The ability to transform core affective changes into emotion likely evolved so that humans could better communicate, predict, and perhaps regulate the underlying core affective reactions of themselves and others.

There are multiple metaphors that describe the sort of emergentism that my lab’s constructionist model proposes, but the one that I find most useful is a culinary metaphor. Just as gastronomic delights such as croissants, brioche, tarts, cookies, sauces, and puddings emerge from the combination of basic ingredients (flour, water, salt, etc.), we hypothesize that emotions such as anger, disgust, fear, happiness, and sadness emerge from the combination of more basic “psychological ingredients.” The two psychological ingredients that our model emphasizes are core affect and conceptualization. Elsewhere, I have hypothesized other basic ingredients, such as sensations from outside the body (exteroceptive sensations) and executive control (for an extended discussion of all hypothesized ingredients, see Lindquist & Barrett, 2012; Lindquist, Wager, Bliss-Moreau et al., 2012; Lindquist, Wager, Kober et al., 2012).

Empirical Support for a Modern Psychological Constructionist Account

Growing evidence supports the hypothesis that a person experiences an emotion when core affect and conceptualization combine in a given context. For instance, participants who feel unpleasant core affect in the presence of accessible knowledge about “fear” are more likely to experience the world as full of risk than participants who feel unpleasant in the presence of accessible knowledge about anger, participants who feel just unpleasant, or participants who feel neutral in the presence of accessible knowledge about “fear” (Lindquist & Barrett, 2008a). I have replicated this effect in several additional studies, in which core affect and conceptual knowledge were combined to produce the experience of pride and anger (Lindquist, Mendes, & Barrett, 2013). My colleagues and I have also demonstrated that the brain networks I associated with core affect and conceptualization in the previous lines show increased activity when a person experiences an emotion in real time in the fMRI scanner (Oosterwijk et al., 2012; Satpute, Shu, Weber, Roy, & Ochsner, 2013; Wilson-Mendenhall et al., 2011).

If core affect and conceptualization are necessary to experience a discrete emotion, it follows that disruption of one or both will prevent experiences of discrete emotions. Although the relevant research has yet to be done in the domain of emotion experience, findings on the perception of emotion on other people’s faces are instructive. Temporarily removing access to conceptual knowledge about emotion by having participants repeat a relevant emotion word (e.g., “anger”) out loud 30 times (a technique called “semantic satiation”) hinders participants’ subsequent ability to judge that two angry faces match in emotional content (Lindquist et al., 2006). Semantically satiating a relevant emotion word (e.g., “anger”) prior to the perception of a face (e.g., an angry face) also impairs that face’s ability to perceptually prime itself on a subsequent presentation (Gendron et al., 2012). Conceptually replicating these findings, patients with semantic dementia (who have impaired access to conceptual knowledge following neurodegeneration in the left anterior temporal lobe) cannot perceive discrete emotion on faces. Rather than sorting posed facial behaviors depicting six emotion categories into six distinct piles in a free-sort task, patients instead sort faces into piles representing positive, negative, and neutral valence (Lindquist, Gendron, Barrett, & Dickerson, 2013). Unlike the neuroimaging findings showing that brain regions involved in language are also involved in emotion, these behavioral tasks more clearly document the role of language in emotion by showing that language influences emotion perception, even on perceptual tasks that do not explicitly require the use of language.

Novel Predictions and Future Directions

Psychological constructionist approaches such as the one I have outlined here provide a promising paradigm for understanding emotion, but much work remains to validate and refine this approach. For instance, future research in my lab, and I hope elsewhere, will continue to explore the dynamics of emotion construction. Although constructionist models traditionally discussed meaning making as occurring well after the experience of an affective state (Duffy, 1941; Russell, 2003; Schachter & Singer, 1962), my model relies on the principles of constraint satisfaction (Read, Vanman, & Miller, 1997), assuming that core affect and conceptualization have a nonlinear influence on each other. This means that changes in core affective representations and conceptual representations are often co-occurring and mutually constraining one another. Although the relevant research has yet to be conducted, the literature on vision is instructive. Within 100 ms of light waves contacting the retina, visual sensations are fed forward to prefrontal areas involved in conceptualization (e.g., the OFC; Lamme & Roelfsema, 2000). This means that conceptualization is occurring even before a person has a conscious experience of visual sensations, at which point it has the opportunity to feed back and potentially even change the visual representation. If the same is the case in emotion, then conceptualization is constraining core affect well before a person is consciously aware of having internal sensations.

Another remaining question is how alterations in one or several psychological ingredients can alter emotional experiences in any given instance (just as increasing the amount of salt and flour in a recipe changes the resulting product). Because the core affective system operates in part via the same neural and neurochemical systems that subserve immune function and behavior, a novel prediction of my psychological constructionist approach is that body changes that are unrelated to the immediate situation (e.g., neuropeptide release related to infection; heart rate changes related to a baroreflex response; hormonal changes related to diurnal rhythms) can contribute to the experience of emotion. Existing evidence suggests that changes in proprioception (Stepper & Strack, 1993), facial muscle movements (Strack, Martin, & Stepper, 1988), peripheral physiology (Dutton & Aron, 1974), and even individual differences in awareness of one’s own body changes (Barrett, Quigley, Bliss-Moreau, & Aronson, 2004; Hantas, Katkin, & Blascovich, 1982) alter a person’s experience of emotion. It thus follows that any neurochemical changes related to homeostasis have the potential to alter emotion experiences (even if those changes are not necessarily related to an emotional context). This avenue of research has important implications for a host of conditions in which there are alterations in core affective responding such as mental illness (e.g., anxiety disorders), physical illnesses (e.g., cancer), and even aging (in which people become less aware of bodily changes; Mendes, 2010).

Another question important to validating and extending a psychological constructionist hypothesis concerns how individual differences in conceptual knowledge shape emotional experiences. Evidence shows that children who do not yet know certain linguistic emotion concepts cannot perceive those emotions on others’ faces, even in tasks that do not explicitly require language (Widen & Russell, 2008). It follows that the content and complexity of an adult’s concept knowledge about emotion will shape how he or she experiences his or her core affective states as instances of discrete emotions. It is possible that individuals who possess more complex knowledge of emotions (perhaps acquired in part from discussions with a parent as a child; Fivush, Berlin, Sales, Mennuti-Washburn, & Cassidy, 2003) will experience more nuanced and complex emotions that are more appropriate to the context. This avenue of research has important implications for emotional well-being, insofar as experiencing affective states as instances of discrete emotions is related to self-regulation (Lindquist & Barrett, 2008b).

Implications of a Constructionist Approach for the Science of Emotion and Beyond

A common critique of my psychological constructionist approach, and constructionism in general, is that it explains emotions out of existence. Although some psychological constructionists have in fact argued that emotions are illusions that scientists should do away with (Duffy, 1934), my constructionist approach takes a more productive approach. It does not assume that emotions are illusions—they do not emerge out of thin air, because there are physical constraints that reliably underlie them (i.e., all measures of emotion find evidence for core affect; for reviews, see Barrett, 2006a; Lindquist, Wager, Kober et al., 2012; Mauss & Robinson, 2009). The role of psychology is thus to explain how the emotions that people experience emerge from these physical constraints. In doing so, the psychological constructionist approach that I take thus cautions us not to take our own experiences of the world too seriously—as scientists, we should not confuse the content of mental experience (e.g., the experience of anger when the dishwasher breaks) for a process (e.g., evidence that there is a mechanism for all instances of “anger”). In the same vein, although instances of emotion correspond to a brain state, this does not mean that the brain regions with increased activation during an emotion experience represent the anatomically given circuit for that particular emotion category. Neural activity should instead be thought of as a snapshot in time of a particular combination or “recipe” of networks that correspond to basic psychological ingredients.

In much the same vein, the constructionist approach I take also cautions drawing firm boundaries between mental states just because we experience them as distinct. The brain-based evidence demonstrates that core affective representations contribute to a host of other mental states including, but not limited to, emotions, thoughts, and body feelings (e.g., Oosterwijk et al., 2012). The fact that sensations from the body form a basic psychological ingredient that contributes to all mental states, begins to break down distinctions between commonsense categories such as “affect” and “attention” (Duncan & Barrett, 2007a, 2007b) or even “emotion” and “cognition,” more broadly (Duncan & Barrett, 2007a; Pessoa, 2008). The brain-based and behavioral evidence I have discussed here further underscores the idea that “cognition” and “emotion” are not distinct processes: traditionally “cognitive” processes (representations of prior experiences, language) are constitutive of emotional experiences (Lindquist & Barrett, 2008a; Lindquist, Wager, Kober et al., 2012; Oosterwijk et al., 2012).

Although mental state categories such as “anger,” “disgust,” “fear,” “memory,” “thoughts,” or even “emotion” and “cognition” might not be entities, they are important mental events that should be studied. In the emotion literature, this is where psychological constructionism makes contact with other models. Emotions clearly evolved for a purpose—even if what evolved was not necessarily a modular package, but more basic processes that combine in humans to produce flexible reactions to the world around us. Emotions are also real events (even if the kind of “real” we are talking about has social rather than biological ontology; Barrett, 2012) that, once formed, can have important downstream consequences for social behavior (DeSteno, Dasgupta, Bartlett, & Cajdric, 2004), relationships (Algoe, Gable, & Maisel, 2010), decision-making (Lerner & Tiedens, 2006), and even health (Algoe & Stanton, 2012; Catalino & Fredrickson, 2011). By telling your loved one that you are feeling an instance of pride, you are communicating something about your internal state and its relationship to the environment, and perhaps even something about the actions you might take next. The ability to construct emotion might thus be one of human beings’ most adaptive traits. Continued understanding of this process will help scientists begin to understand the basis of all human experiences, and to better understand how and when disorder in those experiences occurs.