Surprise as an Emotion: A Response to Ortony

The link between judgments and experience

Taking another perspective on surprise, much of our work has examined responses to surprised facial expressions, which aligns more closely with emotion perception than experience. Although emotion perception, especially for the facial expressions of others, does not necessarily capture the specificity of one’s emotion experience (Barrett et al., 2007; Ortony & Turner, 1990), they are not entirely independent constructs. For instance, emotion expressions are thought of not only as symptoms of emotional states but also as important tools for communication (Adolphs & Anderson, 2018; Fridlund, 1994; Manstead, 1991; Parkinson, 1996). Thus, the signal that is communicated by a facial expression may be just as important as the emotional experience itself. In fact, some have argued that an observer can “reverse engineer” or reconstruct some meaning about the person and the nature of the emotion-eliciting event—including its valence—on the basis of the emotion that is expressed (Frijda, 1986; Hareli & Hess, 2010, 2012). ¹ Further, learning theorists argue that one’s responses to an expression are, at least in part, driven by the learned associations between the expression and outcome over time (Izard, 1971; see also Adolphs & Anderson, 2018, p. 97; Orr & Lanzetta, 1980). For example, we learn to associate an angry expression with a negative outcome and a happy expression with a positive one. Thus, our appraisals of a surprised facial expression provide clues about both the nonverbal message conveyed in the expression and, to some extent, the experience of the messenger.

Additional evidence for the link between judgments and experience comes from affective neuroscience research—which highlights a shared neural basis for both the perception and experience of emotions—suggesting that there is some underlying response in the perceiver that mimics the expresser (Kober et al., 2008; Lindquist et al., 2012). Taken together, although distinguishing these aspects of affective processing is an important endeavor for scientific research, both are often considered to be the sine qua non of emotion (Adolphs, 2017). For this reason, we offer evidence that supports the notion that surprise is valenced, drawing from our own work on perception of surprised facial expressions. We first address Ortony’s assertion that surprise is neutral and then turn to characterizing the valence of surprise.

Surprise is not valence-free

Ortony (2022) stated that “surprise is always the result of the registration of a discrepancy between what is encountered and some reference point” (p. 54) and that “all that is needed for valence-neutral surprise is that the surprising event be of no subjective importance to the person evaluating it” (p. 55). In response to the first point, we would argue that this definition is more descriptive of the term “unexpectedness,” which is instead a cognitive state that registers a discrepancy between what is encountered and some reference point (see also Roseman, 1996).

This brings us to Ortony’s second point about subjective importance. To that end, we argue that the experience of surprise—or any emotion—relies on an experience of subjective importance (see Ely et al., 2015; Ortony, 1988). Indeed, an attempt at reducing subjective importance (i.e., psychological distancing) is an effective strategy for regulating one’s subjective (J. I. Davis et al., 2011; Denny & Ochsner, 2014; Kross et al., 2005) and physiological experiences of emotion (Ayduk & Kross, 2010; Ochsner et al., 2012).

In other words, whereas Ortony proposed unexpectedness as a surrogate for surprise, we propose that unexpectedness (i.e., expectancy violation) is necessary but not sufficient for surprise (see more on this below). This distinction implies that a prediction error carrying no subjective importance might better be referred to as unexpectedness (and we agree with Ortony that this might not carry any emotional value), whereas surprise represents an emotion that is associated with an experience of unexpectedness that carries subjective importance. And although not all experiences of unexpectedness would register as subjectively important (e.g., trivia), we argue that experiences that elicit a surprise hold greater subjective importance. Therefore, surprise can be defined as an emotional experience that is the result of an unexpected event of subjective importance.

Surprise is different from neutral

Another approach to exploring the valence of surprise is to compare the surprised expression to the neutral one, given that facial expressions are important signals about the underlying emotional experience (Fernández-Dols & Russell, 2017; Fridlund, 1994; Frijda, 1986). Indeed, surprised expressions elicit greater electrodermal activity—an objective measure of physiological arousal (Neta et al., 2009)—and greater self-reported arousal ratings (Mattek et al., 2017) compared with neutral, and even angry and happy, expressions. Interestingly, arousal is typically positively correlated with valence intensity, in which an increase in arousal is associated with greater positive or negative valence (Brainerd, 2018; Mattek et al., 2017). In contrast to surprised faces, images with a neutral valence (e.g., neutral facial expressions or scenes) do not produce an arousal response in the viewer.

Surprised and neutral expressions are different in another important way: When presented in a temporal sequence with clearly valenced expressions, neutral faces are subject to an anchoring effect (Russell & Fehr, 1987), whereas surprised expressions reveal a context effect (Neta et al., 2011). Specifically, when presented either alongside or in temporal succession with clear expressions, the appraisal of a neutral expression shifts in the opposing direction (i.e., anchored) such that, if the clear expression was positively valenced (e.g., happy), the neutral face is viewed more negatively, and if the anchor was negatively valenced (e.g., sad), the neutral face is viewed more positively. In contrast, in a similar paradigm, some evidence indicates that the appraisal of a surprised face shifts to be in line with the clear expressions, suggesting a contextual influence (see also H. Kim et al., 2004); surprise is viewed more positively when presented with positively valenced (i.e., happy) expressions and more negatively when presented with negatively valenced (i.e., angry) expressions. Taken together, whereas neutral faces are valence-free, surprised faces carry emotional value that is shaped by contextual influence.

Surprise is ambiguously valenced

Ortony (2022) argued that the need to describe surprise as positive or negative suggests that simply saying “surprise” is insufficient, evidencing that surprise is inherently valence-free. ² In other words, he posited that, because the valence needs to be explicitly clarified, surprise itself must be inherently neither positive nor negative. In contrast, we believe that these linguistic arguments do not point necessarily to an inherently valence-free surprise but could just as likely suggest that surprise does not have a clear valence. Indeed, the idea that surprise can be pleasant or unpleasant does not suggest that surprise is neutral/unvalenced but rather that it is ambiguous—an emotion that is more flexibly experienced in positive and negative contexts alike—setting it apart in a subtle but important way ³ from the other (basic) emotions (Mattek et al., 2017).

Put another way, we agree with Ortony (2022) that surprise does not appear to have a clear valence. However, an important distinction between his view and ours is that, rather than interpreting this to mean that surprise is neither positive nor negative, we argue that it is either positive or negative; rather, surprise is ambiguously valenced. ⁴ In fact, quite a lot of work has demonstrated that surprise is valenced. For example, using a machine-learning classifier, we have isolated facial features of surprise that transmit subtle yet reliable valence signals, which implies the expression of surprise is both generated and perceived as valenced (M. J. Kim, Mattek, et al., 2017). Moreover, because the expression is used to signal multiple meanings in different contexts—some positive and others negative—the valence appraisal is often modulated by individual differences in the perceiver that are stable (Neta et al., 2009) and generalizable (Harp et al., 2021). In other words, valence signals conveyed by surprised faces are driven by environmental (H. Kim et al., 2004; Neta et al., 2011) and cultural (S.-M. Kim, Kwon, et al., 2017) context.

Notably, context is a powerful force driving our responses to facial expressions (Barrett et al., 2011). But what happens when we see an expression, such as a surprised face, in the absence of contextual cues? If the expression had “intrinsic affective neutrality” (Ortony, 2022, p. 56), one might expect that the perceiver would experience no emotional response (see Phillips et al., 2001). In contrast, we have conducted ample research showing that surprise—when presented without contextual information—is initially negative (Neta et al., 2009, 2021; Neta & Tong, 2016; Neta & Whalen, 2010), even in individuals who ultimately interpret surprised expressions as having a more positive valence. Evidence for this initial or default negativity ⁵ does not suggest that surprise is intrinsically negative per se but rather that it holds some emotional value that skews negative—a bias that is evident in response to ambiguity even in nonhuman animals (Bateson et al., 2011)—when not accompanied with sufficient contextual cues.

The plausibility of an initial negativity

It is worth noting that Ortony suggested this intrinsic or initial negativity is implausible. For example, returning to the experience of surprise, he suggested that an intrinsic or initial negativity mechanism would play out such that a larger than expected tax refund registers (consciously) as negative and that negativity is then overridden and viewed positively. However, many of these real-life circumstances are embedded within contextual cues that prevent an initial negativity from being consciously registered. For example, let us say you received an unexpected present from a friend. As you open the gift wrap, chances are you will be quite surprised and pleased (so long as it was not a truly ill-intended prank). This is because your expectation—driven in large part by the context of the friend—sets you up to feel positive (i.e., sufficient contextual cues informing your expectations), and so an initial negativity is unlikely.

In this regard, an initial negativity is more likely to occur when the context does not provide you with sufficient information. To better illustrate this point, let us return to the example above—an unexpected gift from a friend. In a context without the friend present (e.g., an unexpected gift simply appears on your front porch), you might experience an initial (even subconscious) negativity: What is that box doing on my front porch? How did it get there? What could be inside? And only on opening the box to find a gift and a card indicating who sent it does an experience of positivity unfold. Likewise, returning to the example of a tax refund, an initial negativity mechanism could play out such that the moment you open the mailbox to find an unexpected letter from the Internal Revenue Service—assuming this letter registers as subjectively important—elicits an initial (even subconscious) negativity because there is not enough information to know whether this will result in a positive or negative outcome. But then, the additional information provided on opening the envelope and finding a check for a larger than expected amount would prevent any registering of negative valence.

In other real-life scenarios, an initial negativity might be more likely to be consciously registered. For example, playing with a jack-in-the-box for the first time will undoubtedly lead to an experience of surprise. Assuming no prior knowledge of how the toy works, you will first likely be startled by the unexpected and sudden appearance of a doll, followed by a sense of relief shortly after realizing what has transpired. Likewise, if people were to jump out from behind a sofa and shout “surprise!” as you entered your home, you would likely be startled initially but soon realize these intruders are your friends throwing a party for you. Thus, some element of initial negativity—whether that negativity is consciously registered or not—is indeed plausible.

Evidence of valence signals

Functional neuroimaging methods have provided useful evidence for the valence of surprise. For example, there is an extensive literature showing that the amygdala is responsive to biologically relevant stimuli, including emotional stimuli (Adolphs, 2010; Whalen, 1998). Several studies sought to answer which specific component of emotional stimuli might be eliciting amygdala activity and suggested that valence was key (Anders et al., 2008; Jin et al., 2015). It follows then that if surprise does carry valence information, this will be reflected in amygdala activity. Functional neuroimaging studies utilizing surprised facial expressions found supporting evidence for this proposition. This line of work relied on various approaches—leveraging individual differences in valence bias (H. Kim et al., 2003; Neta et al., 2013), manipulating contextual information (F. C. Davis et al., 2016; H. Kim et al., 2004), and analyzing facial features of surprise (M. J. Kim et al., 2020; M. J. Kim, Mattek, et al., 2017). These findings revealed that the amygdala showed increased activity in response to the valence information carried by surprised faces. Still other work has explored surprise in real-world contexts—such as sports spectating—and found that surprise is associated with behavioral, physiological, and neural signatures of emotion (Antony et al., 2021). The breadth of these findings lends myriad converging evidence that valence signals were gleaned from surprise (expressions and experiences alike).

Surprise meets the minimal requirements

Given the wealth of evidence suggesting that surprise is not valence-free and that it instead appears to have a valence that varies as a function of individual and contextual factors, we argue here that surprise is in fact valenced. And if it is valenced (and, according to Ortony, it is also intentional and conscious), then it meets Ortony’s minimal requirements for being classified as an emotion. In other words, surprise is an intentional, valenced, and conscious state.

The discriminating role of subjective importance

Ortony made the “simplifying assumption that judgments about the unexpectedness of a stimulus and reported surprise are perfectly correlated” to “treat unexpectedness as a surrogate for surprise” (p. 56). This can be a useful assumption in some circumstances, but where it fails, it falls short of a convincing argument against the status of surprise as an emotion. For instance, there are a range of experiences that arise from unexpectedness that might also be described as surprising (e.g., discovering that one’s tax refund is greater—or smaller—than expected, receiving an unexpected gift, finding intruders in one’s house, hearing unexpected news about the death of a loved one, seeing Will Smith slap Chris Rock at the Academy Awards). However, there are also a range of experiences that arise from the cognitive state of unexpectedness that neither holds subjective importance nor elicits an emotional response (e.g., learning that the underdog won a playoff game against the favored team in a professional sport that you do not follow, discovering the identity of the individual that invented matches). Importantly, we disagree that these events could be described as surprising. Rather, we argue that what Ortony (2022) referred to as a valence-neutral surprise (i.e., a response to a surprising event that has no subjective importance) may be closer to a cognitive state than an emotion. In other words, unexpectedness is a broader umbrella term that includes surprise but is not synonymous with it; as noted above, unexpectedness is necessary but not sufficient for surprise.

To elaborate, let us consider the work cited by Ortony that reported increased corrugator activity in response to trivia that was previously rated more surprising. Ortony (2022) suggested, perhaps rightfully, that this corrugator response could reflect “puzzlement or even concentration.” But trivia, although it varies as a function of unexpectedness, may not elicit a true emotion of surprise (i.e., with subjective importance). For example, there may be plenty of people for whom the fact that “the invention of matches is attributed to Johnny Walker” registers as unexpected but of no real consequence or interest. This, then, does not meet the criteria for surprise; and thus, trivia may not serve as the best approach for exploring the valence—either through subjective self-report or objective physiological measures—of surprise. Another study, leveraging perceptions of surprise as a useful indicator about the experience of surprise, found that corrugator responses to surprised facial expressions were valenced to a degree that was similar to angry and happy faces (Neta et al., 2009). And this corrugator response to facial expressions—during a task in which participants were merely asked to categorize the faces as positive or negative—is not likely attributable to a state of puzzlement or concentration.

What kind of emotion is surprise?

Ortony (2022) cited work that argued that surprise was not an emotion given that it combines with other emotions (i.e., they “are not single emotions”; see Oatley & Johnson-Laird, 1987). Although this prior work vacillated on the classification of surprise as a basic emotion, we argue instead that the combination with other emotions does not make surprise any less of an emotion. Surprise is perhaps unique in its time course; the emotion experience, and even expression, can be rather fleeting. ⁶ As information is further revealed and processed, surprise may naturally transition into happiness (e.g., realizing someone jumping out from behind your sofa was a friend throwing a surprise birthday party), fear (e.g., realizing someone jumping out from behind your sofa was a home invader trying to rob you), anger or sadness (e.g., realizing someone jumping out from behind your sofa was pulling a mean-hearted prank on you), or other emotions with clearer valence (Mattek et al., 2017). Here, in all three examples, your initial reaction would likely be an emotional response (surprise) because you would not have expected someone to jump out from behind your sofa (i.e., unexpectedness that is subjectively important). And this emotional response would likely be characterized by an initial negativity, particularly if you have no prior information about such an event. But this shorter time course does not detract from the discrete emotion of surprise (for further discussion on the persistence of emotions, i.e., the extent to which it is fleeting or lingers, see Adolphs & Anderson, 2018, p. 71). And if surprise is allowed to be an emotion that evolves on a shorter timescale, it can be an emotion that interacts with other emotions. In other words, it is not the case that the “intensity of . . . surprise . . . is modulated by itself” (Ortony, 2022, p. 56) per se but rather that surprise can modulate the intensity of other emotions occurring in close succession (e.g., happiness, fear, anger, sadness).

Overall, among the basic six emotions, plenty of evidence supports the notion that surprise has a unique characteristic—valence ambiguity—and may operate on a different timescale from other (basic) emotions. We argue here that these characteristics should be embraced as a part of surprise and emotion more broadly, not overlooked because of its uniqueness. And perhaps most importantly, if surprise meets the requirements to be classified as an emotion, it very well could be a basic emotion.