Happiness feels light and sadness feels heavy: introducing valence-related bodily sensation maps of emotions

Emotions play a central role in everyday life: they guide our behavior toward or away from positive or negative incentives, and they interact with information processing and evaluation when making decisions (Buck, 1985; Forgas, 1994; Frijda, 1987; Schwarz, 1990). Despite this relevance and a long history of research, no consensus about the definition, function, underlying physiological processes, and measurement of emotions has been reached (e.g., Ekman, 2016; Gainotti, 2020; Izard, 2013). This is partly due to the fact that emotions are multifaceted phenomena involving physiological responses across the whole body, and changes in subjective feelings, cognition, and behavior (Damasio, 2004; Levenson, 2019; Scherer, 2000).

One of the major controversies is related to the question of whether emotions are best organized in terms of dimensions or in terms of discrete categories (Barrett & Wager, 2006; Kragel & LaBar, 2016; Posner et al., 2005). Discrete models of emotions assume that a limited number of universal and innate basic emotions have been developed during the course of evolution—such as fear, anger, disgust, sadness, and happiness—which subserve adaptational strategies (Ekman, 1992, 1999). Distinct emotion models typically build on the assumption that each emotion has its own specific eliciting condition and its own specific neural, physiological, expressive, and behavioral reaction pattern (Ekman, 1992; Izard, 1991; Tomkins, 1984). Scientific evidence for this assumption is mixed and the matter remains under debate (Barrett & Wager, 2006; Cacioppo et al., 2000; Panksepp, 2007).

Instead of specific mechanisms for each emotion, dimensional models assume that all affective states can be understood as varying degrees of common neurophysiological or “psychological” dimensions. Most prominently, the two-dimensional circumplex model of affect proposed that each emotional state can be described in terms of the level of arousal/activity, and in terms of valence (pleasantness, unpleasantness) (Russel, 1980). Although additional dimensions such as attention-rejection, dominance-submissiveness, power, or intensity have also been proposed (Bradley & Lang, 1994; Mehrabian, 1996; Osgood et al., 1957; Schlosberg, 1954; Wundt, 1905), two-dimensional bipolar activation/valence models remain the most influential ones (Bliss-Moreau et al., 2020; Posner et al., 2005). For example, the arousal and valence dimension accounts for the major proportion of variance in judged similarities among emotion words, suggesting that these two dimensions are the basic properties of the cognitive representation of different emotional states (Bush, 1973; Russel, 1980).

Dimensional models typically build on the assumption that each organism possesses basic approach and avoidance tendencies, and that different emotions emerged from these basic responses in combination with the cognitive interpretation of the situation—valence determines the direction (approach vs. avoidance) and arousal determines the intensity of such basic responses (e.g., Cacioppo & Berntson, 1994; Lang et al., 1998; Russell, 2003). Recent neuroimaging studies confirmed the existence of distinct neuronal networks subserving the valence and arousal dimension of emotions (Colibazzi et al., 2010; Posner et al., 2009), but also the dimensional approach remains debated (e.g., Kragel & LaBar, 2016).

While the discrete and dimensional approaches to emotions are often contrasted against each other (e.g., Adolphs, 2017; Barrett, 2006; Izard, 2007), some authors argued that the issue about the format of representation will most likely never be resolved and that both approaches make valid contributions in understanding emotions (Ekman, 2016; Harmon-Jones et al., 2017). Such an integrative view might also advance the measurement of emotions, which often has not progressed along with theory (Harmon-Jones et al., 2016). Emotions are typically measured using physiological/behavioral and/or self-report measurements. Research driven by distinct models typically focuses on physiological/behavioral responses from facial muscles or from the autonomic nervous system (Ekman & Friesen, 1971; Ekman et al., 1983; Gomez et al., 2005; Konishi et al., 2019; Levenson, 2014; Nardelli et al., 2015; Siegel et al., 2018). Limitations of this line of research are that it is laborious, that often only one specific physiological response is measured, and that the component of subjective feeling is often neglected. Research driven by dimensional models typically focus on self-report measurements (Harmon-Jones et al., 2016). For example, one of the most widely used self-report measurements is the Positive and Negative Affect Schedule (PANAS; Watson et al., 1988), asking participants to indicate to what extent positive and negative emotional words describe their feelings. Other typical self-report measurements require to choose a graphical (e.g., faces, manikins) or numeric representation of emotions on valence and arousal (and sometimes other) dimensions (e.g., Bradley & Lang, 1994; Broekens & Brinkman, 2013; Mehrabian, 1995).

These well-established self-report measurements come along with some limitations. For example, the PANAS has been criticized (among others) for not measuring positive and negative affect directly but rather positive and negative activation (e.g., alert, active, attentive) (Harmon-Jones et al., 2016). This can lead to misleading results when negatively valenced activation is involved, such as an increase of the positive affect score for anger (Harmon-Jones et al., 2009). Such limitations have partly been overcome by assessing emotional states on a two-dimensional valence/arousal affect grid (e.g., Russel, 1980). Nevertheless, it remains a challenge for participants to express subjective feelings of arousal and valence by means of a position on a rather abstract scale, and for researchers and therapists to accurately classify emotions based on these spatial points (Harmon-Jones et al., 2016). Various other self-report measurements have been developed with a specific focus on a single emotion category (e.g., Haidt et al., 1994; Lyubomirsky & Lepper, 1999; Smith et al., 1999; Spielberger, 2010; Tracy & Robins, 2007) or on the clinical context (e.g., McNair et al., 1971; Zuckerman et al., 1965). Although more comprehensive self-report measurements have been developed in the past (e.g., Harmon-Jones et al., 2016; Plutchik, 2001), there is room for further measurements that take aspects of both the dimensional and discrete models of emotion into account, apply for a wide-range of emotions and are easily applicable.

One such novel, innovative tool for the assessment of emotions is “Bodily sensation mapping” (Nummenmaa et al., 2014). Instead of reporting feelings on a scale, participants draw their feelings in the format of bodily sensations (i.e., felt changes in bodily activity in response to emotions) in a whole-body silhouette. These drawings result in bodily sensation maps (BSMs), showing distinct activation patterns across the whole body for different emotional states (Nummenmaa et al., 2014). The same emotion-specific BSM patterns were found across cultures and also across different emotion induction techniques such as self-displacement by emotion words, and emotion induction by emotional videos (Nummenmaa et al., 2014; Volynets et al., 2019).

Bodily sensation mapping combines important elements from both the dimensional and discrete approach to emotions. It is basically a dimensional approach, since participants can indicate the degree of sensation on the activation dimension (e.g., larger and more intense painting reflects more widespread and intense sensations of bodily activation), and it allows doing this in a body-specific, topographical way. Bodily sensation mapping therefore accounts for the fact that emotions are associated with a variety of emotion-specific physiological changes and are consequently “felt” across the entire body (e.g., Barrett et al., 2007; Damasio & Carvalho, 2013; James, 1884; Kim & André, 2008). Given that BSMs rely on self-report, it remains open to what extent BSM patterns truly reflect physiological changes. Nummenmaa et al. (2014) argued that BSMs may not necessarily reflect conscious awareness of a specific physiological change but rather net sensations from different physiological systems (skeletomuscular and visceral sensations, effects of autonomic nervous system). In line with this idea, BSM patterns correspond with the major changes in physiological functions associated with the different emotions (Izard, 1991), such as an increased activity in the upper chest area for most basic emotions, likely corresponding to changes in breathing and heart rate (Levenson, 2003), or increased activity in the upper limbs for approach-oriented emotions such as anger and happiness (Nummenmaa et al., 2014). Moreover, it has been shown that individuals with higher interoceptive accuracy created more intense and topographically more specific BSM patterns following exposure to emotional stimuli (Jung et al., 2017), supporting the view that BSMs may reflect interoceptively experienced physiological changes.

Regardless of the exact mechanisms underlying BSMs, they have important theoretical and practical implications. On the theoretical level, the fact that the reported bodily sensations are topographically distinct suggests that specific bodily sensations are a crucial part of emotion knowledge (Nummenmaa et al., 2014). In line with the embodied view of cognition and emotion, changes in bodily states constitute the body’s registering of how it is being altered by its engagement with the environment, and we feel an emotional state whenever we become aware of those changes in our bodily state (e.g., Foglia & Wilson, 2013; Hufendiek, 2015; Johnson, 2015). The distinct topographical representations of bodily sensations may play an important role for becoming aware about one’s emotional state and for the emergence of one’s emotional feelings (Nummenmaa et al., 2014), which eventually provide meaning to a situation (Johnson, 2015).

On the practical level, BSMs may be a useful tool to improve the specificity and efficiency of communicating and visualizing emotional feelings. People typically have difficulty in assessing, discerning, and describing their own emotions (Becker‐Weidman & Hughes, 2008; Saarni, 1999). Instead of describing their emotional feelings verbally or transferring their feelings into an abstract value on a scale, people can just “paint” a BSM to represent how their body feels. Nummenmaa et al. (2014) proposed BSMs as “somatic markers” for emotional disorders (although we think that the term “bodily sensation markers” is more appropriate, since no somatic measurements are used). BSM is also a promising tool for emotion detection in the context of user experience. A recently developed smartphone application showed that using a BSM approach leads to better classification of the user’s emotion when compared to the classical affect grid approach in which participants provide a single (i.e., body-unspecific) arousal and valence value (García-Magariño et al., 2018).

However, before the potential of BSMs can be exploited and its theoretical and practical contributions further elaborated, a critical advancement is required. Previous BSMs are so far based only on the activity dimension (García-Magariño et al., 2018; Hietanen et al., 2016; Nakul et al., 2020; Nummenmaa et al., 2014; Sachs et al., 2019; Salazar-López et al., 2015; Torregrossa et al., 2018; Volynets et al., 2019), thereby completely neglecting the valence dimension. There is now a general agreement that both arousal/activity and valence are the most prominent underlying dimensions of every emotional experience (Colibazzi et al., 2010; Lang et al., 1993; Posner et al., 2005, 2009), and some authors emphasized that both these experiences depend on feedback from the whole body (Russell, 2003; Wiens, 2005). Thus, the previous activity-based BSM tool only covers a limited spectrum of bodily sensations. This limitation becomes evident for example when comparing the BSM for anger and pride: These two BSMs look quite similar (both show a selective activation in the upper body, see Fig. 1 in Nummenmaa et al., 2014, or Fig. 3 in this study), although there is a fundamental difference in the quality of the feeling associated with these emotions (e.g., Barrett & Bar, 2009). This example highlights that a sensation of activation is unspecific in terms of valence (it can be a pleasant or an unpleasant activation). Following the circumplex model of affect, we therefore argue that only the combination of activity and valence-related BSMs allows for a comprehensive assessment of emotions. In fact, the degree to which an experience feels pleasant or unpleasant has a strong influence on our emotional well-being and motivation. For example, experiencing loss or disappointment can be devastating, whereas experiencing success or pleasure makes us feel good and motivates us to (re)approach the rewarding situation. Moreover, chronic hypoactivity of the mesolimbic dopaminergic system that subserves pleasure and reward can lead to prevalence of negative affect and eventually result in depression or anxiety disorder (Posner et al., 2005). For these reasons, it is important to also consider the valence-related dimension in BMSs, particularly when using BSMs as “bodily sensation markers” for emotional disorders (Nummenmaa et al., 2014). The aim of the current study was therefore to (1) explore whether bodily sensations of valence are also part of emotion knowledge, so that emotion-specific BSMs of valence can be found (similar to the already established activity BSMs), and (2) to assess the added value of including both valence and activity-related BSMs in emotion research. This study therefore provides an important step in advancing the theoretical and practical value of bodily sensation mapping.

A particular challenge for reaching this aim is to find an appropriate way to capture subjective valence-related sensation for specific body regions. While it is straightforward to ask for a global pleasantness–unpleasantness or positive–negative rating as in the traditional dimensional assessments of valence (e.g., on a Likert scale), we doubt that these terms describe best what people actually experience in different body parts. To illustrate, expressions such as “my chest feels negative/unpleasant” or “my head feels positive/pleasant” sound odd.¹ Following the idea that the “wisdom” of language can help to discover the representational structure of a psychological phenomenon (e.g., Davitz, 2016; Kövecses, 2003; Oatley & Johnson-laird, 1987; Shaver et al., 1987), we considered how people describe their positive and negative feelings in language. Emotional well-being—or lack thereof—ranging from happiness on the positive side to sadness and depression on the negative side of the valence spectrum is often expressed by means of the metaphor of weight (e.g., Hamdi, 2016; Hung et al., 2017; Yu, 1995). Someone who is happy feels light-hearted, whereas someone who is sad feels heavy-hearted or feels heavy in his/her chest. Similar weight-related expressions of emotions can be found across different languages and cultures (Moradi & Mashak, 2013; Wenfeng, 2008). Do these metaphorical expressions simply reflect figurative language, or do they possibly reflect an embodied grounding of valence? Striking evidence for the latter view comes from patients suffering from depression. They often describe their emotional state with a feeling of bodily heaviness (Barkow et al., 2004; Fuchs & Schlimme, 2009), and antidepressant effects as bodily lightness (or decrease in bodily heaviness, respectively) (Stocker et al., 2019; see also van Schalkwyk et al., 2018). Some examples of patients reporting antidepressant effects are: “I had less heaviness in my throat and my chest”, “The weight was gone”, “I felt happy and light”, “All the pressure in my head just starts to feel light and normal again” (Stocker et al., 2019). If pleasant feelings of lightness and unpleasant feelings of heaviness can be attributed to specific body parts by patients suffering from depression, then maybe the valence of emotions is also embodied as sensations of weight in the non-clinical population. Zhao et al. (2016) presented healthy participant a picture of a balance as priming stimulus with a heavy and a light stone on the left and right balance pans. The same illustration was then also used to present emotional words (with positive and negative emotion words instead of stones). They found that participants were faster in judging the valence of the words when weight and valence were congruent (i.e., positive-light, negative-heavy) when compared to the opposite pairing. Such an automatic coupling between weight and valence supports the view that the use of weight may not simply reflect figurative language. Specifically, Zhao et al. (2016) concluded that the metaphorical relationship between weight and valence might be derived from bodily experiences with bodily weight and psychological weight. For example, a person may feel exhausted when holding heavy things, which would reinforce a negative mood. Also, a person feels unhappy and sad when her/his psychological burden is heavy. Moreover, happiness and pleasure are associated with the release of hormones and neurotransmitters (the so-called “happy hormones”) such as serotonin, dopamine, oxytocin, noradrenaline, and endorphins (e.g., Breuning, 2016). Some of these prepare the organism to act (e.g., dopamine and noradrenaline increase blood flow of the internal organs and muscles) and decrease pain levels during physical activity (e.g., endorphins, serotonin). Such states of increased motivation and energy help to overcome physiological and psychological barriers, and weights might appear lighter in such conditions (e.g., Farrell, 1985; Tajadura-Jiménez et al., 2015). Thus, there might be an experiential basis for the link between sensations of bodily weight and valence. This argumentation is in line with conceptual metaphor theory, according to which people often refer to familiar, tangible, and concrete concepts (source domain) to understand and express abstract concepts (target domain) (Lakoff & Johnson, 1980; Lakoff & Kövecses, 1987). As stated earlier, emotions can be considered as somewhat abstract feelings that are difficult to express (Becker‐Weidman & Hughes, 2008; Saarni, 1999). The associations between bodily weight and valence may be an illustration of conceptual metaphors where the more concrete experience of bodily weight serves as source domain used to understand and describe the more abstract domain of emotional valence, and the mapping of these domains is grounded in specific bodily sensations (Damjanovic & Santiago, 2016; Johnson, 2015; Niedenthal et al., 2005a). This is also in accordance with Borghi and Binkofski (2014)’s “Words As social Tools” view, according to which both sensorimotor and linguistic experiences form the basis of abstract concepts. Given the wide and systematic use of the weight–valence association across cultures and across clinical and non-clinical populations (Moradi & Mashak, 2013; Wenfeng, 2008), we used the concept of bodily weight (lightness, heaviness) as a proxy for valence in this study. Arguably, there are also other concepts that could potentially serve this purpose, as further elaborated in the “limitations of this study” section of the discussion.

The outline of this study was as follows: We first assessed whether specific BSMs can be created based on the feeling of bodily weight (Experiment 1). To this end, participants were asked to put themselves into different emotional states as indicated by emotion words (basic emotions: anger, fear, disgust, happiness, sadness, surprise; non-basic emotions: anxiety, love, depression, contempt, pride, shame, envy),² and to mark bodily regions that change with respect to the sensation of bodily lightness or heaviness. Since valence is an inherent part of emotional feelings (Posner et al., 2005), we expected that participants will consistently associate different emotional states to specific sensations of bodily lightness and heaviness, resulting in emotion-specific “weight-BSMs”. Following the metaphorical use of weight when describing subjective feelings, we expected a systematic increase in bodily lightness for positive emotions, and an increase in bodily heaviness for negative emotions. The distinctness of the different BSMs was tested by means of a machine learning classification approach (following Nummenmaa et al., 2014).

In the previous experiments (1–3), participants explicitly put themselves into different emotional states as indicated by different emotion words. It is therefore possible that semantic/linguistic associations contribute to the creation of BSMs (see “Discussion” for further elaboration of this issue). To reduce semantic influence, we aimed to replicate the pattern of results obtained in Experiment 2 for sadness induced by a sad movie scene instead of induced by the explicit emotion word in a final experiment (Experiment 4). We particularly focused on sadness, because it is a common experience in the general population and one of the main criteria of major depressive disorder (e.g., Tebeka et al., 2018).

In sum, this study explores for the first time if and how valence, as one of two fundamental aspects of emotional feelings (Posner et al., 2005), is represented with specific body topography for different emotions, and clarifies the most appropriate way (activity vs. valence) for using BSMs in emotional assessment for different emotional states.

The aim of Experiment 2 was to compare weight and activity BSMs and particularly to assess a possible benefit of using weight BSMs. To this end, the same individuals created both weight and activity BSMs, and classification performance of LDA was compared for either weight or activity BSMs alone, and also for the combined weight and activity BSMs. Importantly, subjective ratings were collected to assess the appropriateness of both concepts (weight, activity) for each emotion. In a first step, participants rated after each BSM how easy/difficult they found it to create the BSMs. This rating will not only reveal how easy/difficult it is to draw BSMs for the different emotions, but it will also provide some hints about which concept (weight vs. activity) best captures each emotion. However, a difficulty rating alone might not be sufficient to capture more fine-grained differences in the appropriateness of the two concepts. For example, it can be equally easy to map sensations induced by pride as bodily lightness and activation on the chest and head, because the location and type of sensations (i.e., lightness in case of weight BSMs, and activation in case of activity BSMs) is easy to identify in both cases. Despite equal difficulty, it is still possible that one of the concepts better represents what a person actually feels (weight or activity). We believe that differences between the two concepts become most apparent when participants are forced to explicitly compare the two concepts against each other. In a second step, we therefore also asked participants to indicate the clarity with which they experience and express their bodily sensations by both concepts on a scale with lightness/heaviness on one end and activation/deactivation at the other end. Apparently, such a direct comparison is only possible after participants have displaced themselves into the different emotional states and completed the weight and activity BSMs. Therefore, the clarity rating was asked only after all BSMs have been completed.

In contrast to Experiment 1 where data were collected online, all data for Experiment 2 were collected in a laboratory setting.

As a limitation of this study—as in all studies relying on subjective reports—we do not know to what extent BSMs reflect the actual sensations that participants had when putting themselves into the different emotional states, or whether also other strategies were recruited, such as activation of conceptual knowledge about the emotion and the bodily responses associated with it (i.e., sadness feels heavy in the chest). However, even if the BSMs were not entirely the result of bodily sensations induced by the experimental manipulations, it shows at least that activity and valence-related bodily sensations are both part of the conceptual knowledge of the emotion. Moreover, it is unlikely that specific bodily sensations would be part of conceptual knowledge without systematic experiences linking the concepts together (e.g., sadness and feeling of heaviness in the chest). According to the embodied view of cognition, the activation of conceptual knowledge about emotions involves simulation of bodily states and the reactivation of somatic responses (e.g., Barsalou, 2008; Dijkerman & Lenggenhager, 2018; Niedenthal et al., 2005b, 2009). It is therefore not clear whether it is at all possible to disentangle the different processes that contributed to the BSMs (e.g., direct experience of induced sensations, activation of conceptual knowledge, activation of a previous experience where this emotion was felt in a real-life situation). Importantly, the weight and activity BSMs for sadness induced by an emotion word from Experiment 1 and 2 closely resembled that of Experiment 4 in which sadness was induced by a sad movie scene. This suggests that bodily sensations induced by “reactivated” (or self-triggered) emotions are similar to those immediately induced by the stimulus. However, the preference for weight over activity BSMs was less pronounced when sadness was induced by a movie (Experiment 4) than by the emotion word (Experiment 2). This difference might be interpreted in favor of an influence of conceptual knowledge activation in response to the emotion word, but this difference could also be due to other methodological differences between these two experiments (e.g., different sample characteristics, group vs. single test situation, computerized vs. paper–pencil approach). In this sense, the methodological differences between Experiment 4 and 2 make it difficult to compare the results between the two emotion induction methods directly. Nummenmaa et al. (2014) also induced emotions by emotional movie scenes or guided emotional imagery, and they found the same BSM patterns as with emotion words, suggesting that semantics and stereotypes play only a minor role for BSMs. Future studies could assess the link between BSM patterns, interoceptive awareness/accuracy, and emotional responsiveness more systematically to further disentangle the sensory and cognitive mechanisms contributing to BSMs.

In this study, we used the concept of bodily weight as a proxy for valence. Besides bodily weight, other metaphors have also been related to emotions, such as body temperature (warm vs. cold), pressure, comfort/discomfort, tension, or enhancement (e.g., Barbosa Escobar et al., 2021; Bergman et al., 2015; Borhani et al., 2017; Kövecses, 1990; Rolls et al., 2008; Waggoner, 2010; Wilkowski et al., 2009). Thus, weight is only one form in which valence can be grounded in the body, and our choice of using bodily weight limits the generalizability of our findings. Future research should consider alternative ways to capture valence by BSMs to better understand which emotion is best captured by which concept, and whether similar BSMs can be found for different valence-related concepts. Regarding body temperature, it should be noted, however, that both positive and negative emotions (e.g., love, anger) have been associated with warmth (e.g., Kövecses, 2003; Lakoff & Kövecses, 1987; Wilkowski et al., 2009), suggesting that body temperature might not be a straightforward proxy for valence across different emotions.

Due to (1) the sample size, (2) the fact that our sample was not balanced for age and gender, and (3) the fact that our study did only include healthy participants, we were not able to further investigate individual differences and clinical signatures in weight BSMs. For the activity BSMs, it has already been shown that there are some moderate differences in BSMs between males and females (Volynets et al., 2019), between children and adults (Hietanen et al., 2016), and also some striking differences between non-clinical and clinical populations (depression: Lyons et al., 2021; autism: Palser et al., 2021; schizophrenia: Torregrossa et al., 2018). Future studies should extend this important line of research about individual differences and particularly about clinical implications also to valence-related BSMs.

Furthermore, differences between weight and activity BSMs were found only for clarity, but not for difficulty ratings. A possible explanation is that clarity might be more related to the question whether the concept (weight vs. activity) captures the quality of bodily sensations, while difficulty might be more related to the question how easy/difficult it is to allocate one’s bodily sensations onto the BSMs. Alternatively, it is possible that the direct comparison between the weight and activity concepts (in a forced-choice manner), as it was the case for the clarity ratings, is more sensitive to differences than when each concept is rated independently from each other, as it was the case for the difficulty rating. Due to the confound between the type of question (difficulty vs. clarity) and the experimental procedure (rating after each trial vs. comparative rating at the end), we are not able to disentangle these two possible sources with the present data.

Finally, in the human classification task (Experiment 3), participants allocated BSMs to emotion words in a forced-choice setting. This approach might have induced specific task strategies, as discussed in the Results and discussion section of Experiment 3. Such problems can be overcome by presenting each BSM separately and requiring participants to rate their level of agreement for each emotion. Such an approach might be used in future studies to better assess the certainty of classifications and more generally the extent to which emotions are represented as distinct categorical maps or as continuous/overlapping maps.

In this study, we established for the first time valence-related BSMs that systematically capture bodily sensations associated with different emotions in a somatotopic format (lightness for positive emotions, heaviness for negative emotions). We also showed that emotions can be best recognized by humans and computers when classification is based on both activity and valence-related BSMs compared to either type of BSM alone, suggesting that both types of bodily sensations reflect distinct part of emotion knowledge. This finding is in line with the circumplex model of affect, according to which valence and arousal are two fundamental and distinct dimensions of every emotional experience (Colibazzi et al., 2010; Lang et al., 1993; Posner et al., 2005, 2009). Our study highlights that BSMs provide a fine-grained assessment of emotions by taking specific bodily sensations of valence and arousal into account, and therefore extend the traditional dimensional approaches in which typically only one global value (e.g., activity or valence rating on a scale) is considered. Participants found it rather easy to express their bodily sensations by means of BSMs (except for envy and contempt), suggesting that BSMs might be a powerful tool in assessing and communicating emotional feelings. Importantly, our results suggest that valence-related BSMs might be particularly important for the assessment of sadness and depression, possibly capturing more accurately what people feel when they are sad.