The classical view of emotions as natural kinds has largely dominated our scientific understanding of emotions, whether in terms of crafting experimental designs, interpretation of results, or its implementation in practical settings. The term ‘natural kinds’ expresses the idea that basic emotions are biologically-given, universal, and inherited as products of evolutionary adaptation for their adaptive value (Panksepp, 2007). Specifically, each basic emotion is hypothesised to be autonomously triggered by a specific causal mechanism in the brain that produces a signature response unique to that emotion (Izard, 2007). However, this established perspective of emotions has seen increasing disagreement in recent decades. Critics have debated that this understanding is fundamentally flawed or, at the very least, incomplete (Russell, 1994; Barrett, 2006). In this essay, I seek to argue that the idea that ’emotions are natural kinds’ is well-supported, but insufficient in itself, in that, such an isolated perspective does not comprehensively account for all evidence presented in the literature – which often results in misinterpretation or exclusion of important data. I will attempt to do so by first clarifying the terms, followed by evaluating the evidence based on three defining characteristics that would categorise emotions as natural kinds: (1) similar observable properties, (2) specific causal mechanisms, and (3) unique functional properties.
The imprecise use of the word ’emotion’ among researchers has resulted in unnecessary disagreement concerning the nature of emotions (Barrett, 2006; Izard, 2007). Opponents of the natural-kind view often incorrectly reference emotional schemas or complex emotions (e.g. amae in Japan; gezellig in Holland) instead of basic emotions in their analysis of the evidence, which leads to inaccurate conclusions or accusations against proponents of the view (Ekman, 1994; Izard, 2007). Unlike complex emotions or schemas, basic emotions do not involve higher-order cognition, or any interaction of emotions and perception. To prevent such confusion, this essay scopes the term ’emotion’ in the question as referring to basic emotions as traditionally assumed by the natural-kind view.
Similar Observable Properties
Based on the natural-kind view of emotion, it is assumed that that every basic emotion has a cluster of observable properties that are similar and projectable, such that they are reliably identified during every occurrence of that emotion (Izard, 2009). Each basic emotion is proposed to contain a core set of correlated properties that differentiates it from other basic emotions, while maintaining a profile that is similar to its discrete category (Izard, 1971). For example, instances of disgust are expected to share sufficiently similar correlated properties with each other (as defined by its category), while being clearly distinguishable from instances of other basic emotions, such as anger or fear.
One of the main evidence for a systematic relationship between components of emotion comes from experimental studies that demonstrate a strong positive relationship between participants’ subjective emotional experience and facial expression for each basic emotion (Ekman, Davidson & Freisen, 1990; Matsumoto & Kupperbusch, 2001). In most of these studies, participants are situated within a highly emotional context (e.g. recruiting bereaved individuals to share about their deceased spouse to participants) to elicit authentic emotional responses of higher intensity, which in turn, allow greater accuracy and validity in their measurements. However, despite these good intentions, several of such studies do not appropriately account for participants’ arousal level or valence when inducing such emotions (e.g. lack of control group). This could have inflated the size of any observed effects and their respective interpretations, thus potentially undermining the evidence for emotions as natural kinds. Moreover, the evidence aforementioned has also been criticised due to the lack of consistency. Several studies (with similar methodologies) were unable to replicate such findings, in which they revealed little or no association between participants’ emotional experience and their facial expression (Fernandez-Dols & Ruiz-Belda, 1997; Ruiz-Belda, Fernandez-Dols, Carrera & Barchard, 2003).
Furthermore, studies testing the relationship between emotional experience and physiological response have likewise been shown to be inconsistent, with findings ranging from strong to little or no correlation. Yet, upon closer inspection, it is found that findings of weak relationships between measures are mainly confined to earlier studies (Mandler, Mandler, Kremen & Sholiton, 1961; Weinstein, Averill, Opton & Lazarus, 1968), while majority of later studies revealed moderate to strong relationships (Hubert & de Jong-Meyer, 1990; Ekman & Davidson, 1993; Ekman et al., 1990). One possible explanation for the observed pattern could be that later studies had more extensive methodologies and improved technology to better record and measure participants’ responses, which are important for determining coherence between response systems. This can be inferred from Mauss, Levenson, McCarter, Wilhelm and Gross’ (2005) study, which made use of online, second-by-second precision measurement of participants, which allowed them to assess participants’ subjective emotional experience (of amusement and sadness) in real-time without inferring with the emotion elicitation. This was not achieved by previous studies, which were largely limited to retrospective and aggregated ratings. Moreover, the study was also able to concurrently measure three properties of basic emotions (not done in previous studies). By doing so, this reduced possible error arising from the selection of an incorrect measure for each emotional response, thereby increasing the reliability of their measurements. Based on these improvements in experimental design, their study revealed moderate to strong correlations between participants’ physiological response and emotional experience, and their physiological response and facial expressions; thus, potentially supporting the idea of clusters of similar, observable properties for each basic emotion.
However, deeper consideration of the evidence would reveal that it is not conclusive that each basic emotion has a common set of observable properties, probably because it is not sufficient in itself. Rather, such discrepancies within the evidence can be better accounted for by adopting an integrated position (Matsumoto, 1989; Ekman, 1994). In particular, this perspective does not ignore the role of culture or social learning in characterising basic emotions, such as its expression and appraisal of them (Ekman, 1992). Strong evidence for this comes from cross-cultural studies of emotions. These studies support the idea that the observable properties of basic emotions are universal, but do not deny the influence of cultural differences (Ekman, 1994; Matsumoto, 1989; Russell, 1994). Among these studies, Scherer and Wallbott (1994) provides a rather comprehensive analysis. In particular, they investigated the proposed differential patterns of basic emotion properties by collecting data across 37 countries. Based on their analysis, they found highly significant main effects of response differences for seven discrete basic emotions, in terms of subjective experience, physiological responses, and expressive behaviour, with strong effect sizes. While this is indicative of basic emotions being universal, it does not imply absence of cultural influences (which seems to be assumed by several past studies since they did not attempt to measure such differences). Specifically, their analysis found that cultural differences (country-dependent) had explained a significant amount of variance in participants in terms of people’s emotional response, social and symbolic representation, and regulation. The data supports neither an extreme position of natural-kinds nor socially constructivism, but rather a perspective that accounts for both factors. These findings are further supported by Elfenbein and Ambady’s (2002) meta-analysis, which evaluated studies on basic emotion recognition (linked to expression) across cultures. By sorting native and non-native participants, they found that recognition of a non-native’s emotions was significant (and above chance), implying that basic emotions are universal – they have the same set observable properties that allow them to be identified by people even of different cultures. Notably, the analysis revealed that accuracy of recognition was greater for native participants (i.e. in-group advantage), while controlling for display rules and exposure between cultures. This additionally suggests that basic emotions are uniquely shaped by cultural factors, therefore making them more distinct to members of the same background.
Besides evidence from adults, developmental studies also serve as important evidence in determining whether basic emotions have similar observable properties. Past research provided much support for the existence of similar observable properties in infants. A study by Izard, Fantauzzo, Castle, Haynes, Rayias, and Putnam (1995) demonstrated that infants, from 2 months of age, displayed distinctive patterns of facial expressions for a range of basic emotions towards stimuli that required minimal appraisal (e.g. mother’s smile elicited infant’s expression of joy). Critically, these expressions were not only congruent to that seen in adults, but also recognisable by adults, therefore giving support to the idea that basic emotions are natural kinds. Moreover, this evidence can be further supported by consistent findings across different cultures. Results from Camras, Oster, Campos, Miyake and Bradshaw (1992) study revealed that similar facial expressions for several basic emotions were observed between Japanese and American infants from 5 months of age.
However, there has also been conflicting evidence showing that infants do not display distinct patterns of observable properties. A later study by Camras, Meng, Ujiie, Dharamsi, Miyake, Oster, Wang, Cruz, Murdoch and Campos (2002) showed that nine-month-old infants of different cultures did not exhibit a distinct set of observable properties for ‘surprise’ in response to novel stimuli. By using an expectancy-violation paradigm, they found that only 30% of infants were judged to display the prototypical facial expressions or behaviours of ‘surprise’ as defined by its discrete category, in which cultural factors explained majority of the variance. This evidence could potentially support the notion that emotions are constructed by socio-cultural factors rather than being natural kinds. Yet, these results may have underestimated the level of similarity of expression of each basic emotion across cultures: (1) the study did not account for differing cultural dispositions (e.g. Asians less expressive; Gross & John, 1995), and (2) only measured for expressions of ‘surprise’, therefore ignoring the possibility that infants responded with a different basic emotion, such as fear – which could have been elicited by infants due to the uncertainty of the event, and likewise account for the observed ‘bodily stilling’ of infants. Even if the study did not underestimate their findings, it does not necessarily prove the lack of unique patterns of observable properties in infants. Rather, this seemingly contradictory evidence could actually provide a more comprehensive understanding of the nature of emotions when considered alongside Izard’s (1994) review. Based on cumulative evidence, Izard (1994) found that the distinct patterns of expression of each basic emotion only remained relatively stable in the first 9 months. This could possibly explain why Camras et al. (2002) did not find any distinct observable properties of ‘surprise’ in their infants, who were at the proposed limits of the age boundaries for consistent expression of basic emotions. When collectively considered, the evidence supports an integrated view that while each basic emotion displays similar observable properties, they can be altered by cultural influences at a later stage of early development.
Therefore, based on the various evidence discussed within this section, I argue that basic emotions are not completely natural kinds; in that, while basic emotions have distinct patterns of similar observable properties that categorise them as natural-kinds, they are also characterised by socio-cultural influences.
Specific Causal Mechanisms
For basic emotions to be natural-kinds, all instances of each basic emotion should be intrinsically derived from the same underlying cause that is independent of other processes (Izard, 1992). Evidence for such causal mechanisms comes from neuroimaging studies, which propose that basic emotions can be localised within distinct circuits or regions of the brain.
Recent studies have found discrete neural correlates for basic emotions in humans. In particular, a comprehensive and summative report is provided by Vytal and Hamann’s (2010) meta-analysis of 83 neuroimaging studies (fMRI and PET), which corrected for the limitations of meta-analyses by Murphy et al. (2003) and Phan et al. (2002). Based on their results, they found consistent and discriminable neural correlates of the basic emotions of happiness, sadness, anger, fear and disgust. Each basic emotion was found to display a consistent signature pattern of brain activation that significantly differentiated it from other basic emotions: happiness activated the rostral anterior cingulate cortex (rACC) and the right superior temporal gyrus (STG), sadness activated the middle frontal gyrus and subgenual ACC, anger activated inferior frontal gyrus (IFG) and parahippocampal gyrus, fear activated amygdala and insula, and disgust activated the IFG/anterior insula (different region from anger). The existence of such specific localizations for basic emotions is also strongly supported by animal studies – more substantiated than human studies due to limitations of testing humans (e.g. ethical issues) and recording human brain activity (e.g. difficulties of using fMRI to resolve slowly firing subcortical regions that are tightly organised; Panksepp, 2007). The comprehensive animal data reported in Panksepp (2005) has been found to correspond well with human data. Specifically, animal studies provide compelling homologous evidence for overlapping primary affective circuits for basic emotions. Such circuits have been shown to respond to unconditioned emotional stimuli, which, in turn, generate emotional experiences in animals (Panksepp, 2005; Panksepp & Watt, 2011). Based on cross-species studies, these primary emotional circuits have shown to be homologous in humans, and found in slowly-firing subcortical limbic regions that project to higher cortical regions for cognitive appraisal and representation (Panksepp & Watt, 2011).
However, data from neuroimaging studies has also shown mixed findings, and therefore are not entirely conclusive in their overall interpretation. It has been argued that such ambiguity, especially in data from humans (possibly due to limitations in neuroimaging humans), suggests that emotions should be categorised as dimensional rather than discrete (Russell, 1994; Barrett, 2006). Though a plausible interpretation, it may not provide a comprehensive account of the evidence. This is because such an interpretation implies that theories of discrete and dimensional emotions are mutually exclusive. This has been refuted by Panksepp (2007), who asserts that there is strong neural evidence to suggest that both theories can co-exist. In a later review, Panksepp and Watt (2011) present that the neural evidence can be stratified into different levels of analysis, in which dimensional emotions refer to higher-level concepts of affective states stemming from cortical regions, while discrete basic emotions refer to the rudimentary emotions resulting from low-level primitive systems based in sub-cortical regions. In line with this perspective, such ambiguity should therefore not undermine the substantial neural evidence that supports the existence of distinct causal mechanisms for basic emotions.
Hence, though further research may be warranted for greater clarification, it can be argued that the current neural evidence supports the idea that basic emotions are natural-kinds.
Unique functional properties
Lastly, basic emotions can be categorised as natural-kinds on the basis of having unlearned and unique motivational and regulatory properties. These properties are argued to be innate on the basis that they serve adaptive functions necessary for basic survival and well-being (Izard, 2009). This is supported by the fact that they only require minimal appraisal of a stimulus, and are not activated as a consequence of other perceptual processes (Izard, 2007). Specifically, these unique properties have been shown to differently motivate our response systems for cognition and action, which in turn regulate our basic emotion via action tendencies (Izard, Ackerman, Schoff & Fine, 2000; Merker, 2007). Even from birth, basic emotions have shown to perform important social and communicative functions (Izard, 2009). Infants with emotional disorders were found to display problems with their conduct at a later age, similar to that of psychopaths (Conduct Problems Prevention Research Group, 1999). Crucially, the fact that each basic emotion serves sufficiently distinct functions that cannot be learned (yet to be demonstrated by any study), provides the strongest evidence for the natural-kinds view. For example, basic emotions of ‘disgust’ inhibits approach and motivates avoidance (Davey, 2011), while ‘interest’ directs attention and motivates exploration (Silvia, 2006).
However, although basic emotions are found to be have functional properties that are unlearned and unique, research has shown that they can be altered by social factors. Unconventional evidence for this comes from studies on infants who were raised in the absence of social interaction (i.e. institutional rearing in Eastern Europe). Several longitudinal studies have collectively found that such infants suffer from severe deficits in emotional expression and regulation (Fries & Pollak, 2004; Smyke et al., 2007). This is supported by an event-related potential (ERP) study by Parker and Nelson (2005), which tested institutionalised and non-institutionalised Romanian infants. They found that institutionalised infants were significantly poorer in matching emotional expressions with their appropriate social contexts, and displayed greatly attenuated ERP amplitudes related to processing emotional stimuli, such as potential threats and facial expressions. These studies, although extreme, possibly reveal the nature of the proposed functional properties of basic emotions; that they can be altered, and even lost, when placed in an environment that does not cultivate such motivational or regulatory capacities. This potentially suggests that, while functional properties of basic emotions are innate, social factors are necessary for their development. This is further supported by the fact that the extent of exposure to a certain social environment can affect the development of such functional properties. Specifically, several studies found that the length of institutionalisation was a strong predictor of subsequent negative emotional outcomes (Castle, Groothues, Bredenkamp, Beckett, O’Connor, & Rutter, 1999; Fisher, Ames, Chisholm & Savoie, 1999).
Therefore, the evidence suggests that emotions are not completely natural kinds; on the basis that, while unique functional properties of basic emotions are innate, they are also determined by environmental factors.
In conclusion, I argue the evidence does not completely agree with the idea that “emotions are natural kinds”. In evaluating the evidence, basic emotions are found to meet the three defining criteria of a natural kind, namely having (1) similar observable properties, (2) specific causal mechanisms, and (3) unique functional properties. Yet, while this is true, it does not sufficiently define the nature of basic emotions, in that, it does not account for all the evidence presented in research. Instead, a comprehensive understanding of basic emotions would require an integrated perspective; one that accepts nature’s role in providing inherently adaptive functions that are necessary for basic survival and interaction, while not denying the socio-cultural influences that mediate the characteristics of basic emotions. In moving ahead, theorists should aim to adopt such an integrated position, and direct future research towards collectively developing a coherent and comprehensive model of emotions