As mentioned above, definitions of empathy cover both cognitive and affective empathy (Cuff et al., 2016; Singer and Lamm, 2009). It has been shown that cognitive and affective empathy partly rely on different neural circuits (Shamay-Tsoory et al., 2009). This suggests that cognitive and affective empathy are independent to some extent. However, their specific interaction and interdependencies are still debated (Cuff et al., 2016; Preston et al., 2020).

It has recently been argued that affective empathy can occur without cognitive empathy but not vice versa because affective representations in the bodily state are necessary to understand the emotional state of another person (Preston et al., 2020). Notably, affective empathy is considered the result of an initial bottom-up process wherein the perception of external stimuli, e.g., the emotional state of another person, induces a representation in the observer itself (de Waal and Preston, 2017; Preston et al., 2020). On the contrary, cognitive empathy is thought to be a top-down process in which internal stimulation leads to cognitive empathy processes. Despite this difference in the origin of stimulation (external versus internal), and the sequence of processes (bottom-up to top-down process for affective empathy and top-down to bottom-up process for cognitive empathy), both affective and cognitive empathy activate affective representations in the observer, leading to similar representations in bodily states (de Waal and Preston, 2017; Preston et al., 2020).

The distinction between cognitive and affective empathy may also be crucial when discussing situational variables affecting state empathy. In this regard, twin studies estimating genetic and environmental portions suggest that affective empathy is more heritable than cognitive empathy, as demonstrated by a recent meta-analysis (Abramson et al., 2020). From this finding, one might conclude that affective empathy is more strongly determined by genetic factors than cognitive empathy, and in turn, that affective empathy may be more stable and cognitive empathy more situated. Cognitive empathy may thus be more susceptible to the potential effects of detrimental environmental factors, but also more responsive to interventions meant to increase empathy.

Preliminary findings contradict these conclusions. For example, Wolf et al. (2015) did not find changes in cognitive empathy measures after acute stress. In contrast, however, they report an increase in affective empathy following acute stress. Other studies report beneficial effects of acute stress on emotion recognition, a subcomponent of cognitive empathy (Domes and Zimmer, 2019). The apparent inconsistency in these results can be attributed in part to differences in the selection of measurement tools for empathy. For example, Wolf et al. (2015) implemented the Multifaceted Empathy Test in its Condensed and Revised Version (MET-core-2; Dziobek et al., 2011) to measure empathy, a task that includes the identification and the affective sharing of complex emotions. Alternatively, participants in a study by Domes and Zimmer (2019) were asked to decide whether one of two basic emotions or a non-emotional condition (angry, happy, or neutral) was present on pictorial stimuli. Such methodological disparities likely contribute to the mixed findings, an explanation that will be elaborated later in the manuscript.

The issue becomes even more complex when one considers that both cognitive and affective empathy lead to changes in bodily states (de Waal and Preston, 2017). Thus, the construct of embodiment must be recognized when discussing empathy. Embodiment is a psychological construct that refers to the idea that our thoughts, emotions, and behaviors are shaped by the physical body and its interactions with the environment (Newen et al., 2018). In the context of empathy, embodiment implies that bodily physical sensations influence the way we experience and respond to others’ emotions (Niedenthal, 2007). One implication of this approach is that perceiving another person’s emotions depends on the interaction between your bodily state and the emotions the other person displays (Niedenthal, 2007). This idea also implies that perceiving an emotion is always situated, and empathy should be considered an embodied process instead of a purely cognitive one.

In sum, because empathy is conceptualized as a complex and dynamic process influenced by a variety of internal and external factors, it is imperative to establish guidelines in measurement standards for empathy research. The recent theoretical and empirical developments in empathy research lead us to conclude that trait empathy represents an individual’s overall capacity for both cognitive and affective empathy. In contrast, state empathy is the specific expression of this capacity in a given situation. In this regard, the definition of Håkansson Eklund and Summer Meranius (2021) appears incomplete and requires expansion for clarity. We propose the following addition (see italics) to this definition: The general construct of trait empathy includes four factors: understanding, feeling, and sharing another person’s emotion with a clear self-other differentiation. The expression of empathy is situated within both psychological and physiological factors that influence the empathic response.

Past research on empathy has not only been challenged by methodological issues but also theoretical ones, specifically in differentiating empathy from its related constructs (Cuff et al., 2016). A central factor in discriminating empathy from other constructs is self-other differentiation (de Vignemont and Singer, 2006). In (both trait and state) empathy, the observer is aware that the affective state of interest originates in another person and not in oneself. In contrast, when experiencing emotional contagion for example, the observer acknowledges an affective state but is unable to determine where it originated. Thus, the observer believes that the observed affective state comes from themself and fails to differentiate between sources of origin (Cuff et al., 2016).

It is also important to disentangle state empathy from pro-social behavior. Pro-social behavior is broadly defined as any action or behavior that promotes welfare in others (Pfattheicher et al., 2022). State empathy is the situated expression of empathy that can, but does not have to, be expressed in behavior. The question then arises of how state empathy and prosocial behavior are connected.

It has been suggested that (state) empathy constitutes the basis for showing prosocial behavior (Stevens and Taber, 2021). However, empathy and prosocial behavior do not seem to act on each other directly. Instead, it has been proposed that empathy resulting in prosocial behavior is mediated by compassion (Stevens and Taber, 2021). Compassion is defined as the ability to understand when another person is suffering and feeling emotionally connected to that person (Strauss et al., 2016). Furthermore, the observer can understand the common ground of this emotion, tolerate the (potentially negative) emotions that result in themself, and finally act or develop the motivation to act (Strauss et al., 2016). Stevens and Taber (2021) postulate that while compassion leads to prosocial behavior, empathy does not do this automatically. Regardless, empathy is considered an essential component of the emergence of compassionate feelings and thus prosocial behavior (Lim and DeSteno, 2016).

Two more distinctions are important to note. First and foremost, the difference between emotion recognition and empathy plays a central role in empathy research. Emotion recognition is a cognitive ability that enables the recognition of basic and complex emotions in others (Domes and Zimmer, 2019; Gamsakhurdashvili et al., 2021a). It is generally considered to be one facet of cognitive empathy (Cuff et al., 2016; von Dawans et al., 2021). Beyond this, cognitive empathy includes further processes such as mentalizing (von Dawans et al., 2021). Basic emotions typically cover happiness, sadness, fear, anger, surprise, and disgust, and are recognized across cultures (Fridenson-Hayo et al., 2016). In contrast, complex emotions are culturally dependent and rely on the context in which they occur (Fridenson-Hayo et al., 2016). Second, for complex emotions, it is worthwhile making a distinction between empathy for pain vs. empathy for other complex emotions. Empathy for pain has been researched extensively in behavioral paradigms (Gonzalez-Liencres et al., 2016; Lamm et al., 2007; Tomova et al., 2017), but only a few studies consider state empathy measures that target more complex, situated emotions as assessed by the MET-core-2 (e.g., highly satisfied, relaxed, or jubilant as positive emotions, and terrified, frustrated or desperately unhappy as negative emotions) while measuring both, cognitive and affective empathy (Drimalla et al., 2019; Dziobek et al., 2011; Dziobek et al., 2008; Gamsakhurdashvili et al., 2021a; Gamsakhurdashvili et al., 2021b; Wolf et al., 2015). Findings from previous studies targeting empathy for pain have been generalized to the wider construct of empathy (Lamm et al., 2007; Tomova et al., 2017). However, Timmers et al. (2018) showed that there are unique neural correlates of empathy for pain and that because a distinction between empathy for pain and empathy for other emotions exists, any generalizations should be made with caution (Timmers et al., 2018). These different components reveal that empathy is a nuanced concept and that measurement tools must be adjusted to the specific aspect of the research question. To the extent that it is possible, this review focuses on measurements of empathy for complex emotions.

In self-report measures, participants are asked to evaluate their empathic abilities or situational expression by choosing a response based on how much they agree or disagree with an item (e.g., “I really get involved with the feelings of the characters in a novel.”; Davis, 1983). Hall and Schwartz (2019) have recently gathered an overview of the most frequently used self-report empathy measurement instruments. Similarly, Yu and Kirk (2009) have listed available measurement tools. Based on these two reviews, we have compiled an overview of self-report measures of empathy in Table 1. Note that to refrain from a subjective bias when selecting the list of measurement tools included in this review, we decided to base our list solely on the systematic review by Yu and Kirk (2009) and the quantitative review by Hall and Schwartz (2019). Thereby, the list is not exhaustive but aims to summarize the most common approaches. It is important to acknowledge, that by using this approach to select measurement tools, some tools that are also used in research are not included in this review [e.g., the Emotional Contagion Scale by Doherty (1997) and the Emotional Empathy Scale by Mehrabian and Epstein (1972)]. To facilitate the choice of an appropriate measurement tool, it is essential to know which tools address which facets of empathy (cognitive versus affective versus both), and if they distinguish between trait and state empathy.

In addition, Vieten et al. (2024) recently published a very concise and comprehensive overview of both, empathy and compassion measurement tools. While these authors have focused on including both constructs, the present review solely focuses on the work by Hall and Schwartz (2019) and Yu and Kirk (2009) and thereby on empathy measurement tools and the distinction of trait and state empathy.

Performance measures, in contrast to self-report measures, require participants to make a (behavioral) forced choice between different alternatives with one of the alternatives being the correctly identified emotion. This opens up the possibility of measuring empathy based on responses as participants do not subjectively have to rate their empathic skills. In the Reading the Mind in the Eyes Test, participants are instructed to identify the emotional state of the protagonist by evaluating the expression of a pair of eyes on a picture and choosing one of four alternatives (RMET; Baron-Cohen et al., 2001). Other frequently used tests are the Pictorial Empathy Test (PET; Lindeman et al., 2016) and the Multifaceted Empathy Test, both in its original version (MET; Dziobek et al., 2008) and a Condensed and Revised Version (MET-core-2; Drimalla et al., 2019; Dziobek et al., 2011). In line with the questions raised in the previous chapter on self-report measures, we provide a systematic categorization of empathy measurements. The tools are (1) categorized based on whether they assess cognitive empathy, affective empathy, both constructs, or neither, (2) further classified as a self-report measure or performance measure, and (3) grouped on whether they specifically address trait empathy, state empathy, both constructs, or neither. In addition to the tools addressed by Yu and Kirk (2009) and Hall and Schwartz (2019), we also included the MET (Dziobek et al., 2008), the MET-core-2 (Drimalla et al., 2019; Dziobek et al., 2011), the BES (Jolliffe and Farrington, 2006), the PET (Lindeman et al., 2016) and the RMET (Baron-Cohen et al., 2001) in Figure 1 based on the frequency of their use in research. Where possible, we have included a citation of the sentence upon which we based our decision in Appendix Table A. We based the categorization not on how these measures have been used in the past, but solely on what is stated in their respective manuals. This caused certain challenges; for example, we categorized the RMET (Baron-Cohen et al., 2001) as “not addressed” even though it has been used as a state measure in most studies. Similarly, the IRI (Davis, 1983) is frequently used as a trait measure of empathy but the original manuscript did not state specifically if this tool is used to measure trait or state empathy. As we believe this to be the most objective approach, we ask that researchers address the distinction between trait and state empathy in their manuals. Lastly, we want to point out that some information displayed in the table was adapted and extended based on the work by Yu and Kirk (2009).

As depicted in Figure 1, a high number of self-report measures cover trait empathy. In contrast, to our knowledge, only one self-report measure has been developed to approach state empathy, namely the BLRI (Barrett-Lennard, 1962), and no self-report measure specifically targets both, trait and state empathy in one tool. While Batson et al. (1987) mentioned in their manual of the Batson scale that it is advisable to establish a clear conceptual distinction between variations in empathic emotion experienced in specific situations, and that the construct gauged by self-report indicates a more general concept, the authors did not include two separate measures for trait and state empathy.

Figure 1 outlines that most measurement tools fail to define whether the self-report measure targets trait or state empathy. Although this can be indirectly inferred from the wording used in some of the items of the questionnaires, we want to emphasize that the authors should address the issue specifically to evolve and further define the correct use of the respective measurement tool. Correspondingly, performance measures were classified as a trait measurement tool, a state measurement tool, or did not address the distinction at all. We consider performance measures an appropriate means for assessing situated empathy that is influenced by contextual factors, e.g., the testing environment. The extent to which a performance measure can assess trait empathy via a repeated-measure design deserves further investigation.

One of the few studies to have addressed the difference between trait and state empathy was conducted by Zhao et al. (2021). They used both the IRI (Davis, 1983) and the EQ (Baron-Cohen and Wheelwright, 2004) as measures for trait empathy, and an adapted version of a task-based empathy measure by Neumann et al. (2013) to assess state empathy. Similarly, van der Graaff et al. (2016) measured trait empathy using the IRI (Davis, 1983) and state empathy using an adapted task-based empathy assessment that included watching emotionally loaded film clips, a subjective rating, and identification of the emotion. The selected videos for the state empathy task included four different clips representing either happiness or sadness and were taken from Dutch documentary films (van der Graaff et al., 2016). In addition, prior to the task as well as in between emotional video clips, participants watched fragments of an aquatic video fostering relaxation (van der Graaff et al., 2016). Notably, both studies used a self-report measure for trait empathy and a performance measure for state empathy; a division that we favor as well.

Given the classification, tools such as the MET (Dziobek et al., 2008) or the RMET (Baron-Cohen et al., 2001) have further limitations. For example, the extent to which existing performance measures of empathy can distinguish simple emotion recognition from cognitive empathy has been questioned (Preston et al., 2020). Both the RMET and the MET ask participants to recognize the emotional state of a target on pictorial stimuli but fail to include higher-level processing steps (Preston et al., 2020).

At this point, it is important to note that in addition to self-report and performance measures, neural activity might serve as a further category to assess empathy. To the best of our knowledge, only one study has specifically targeted the distinction between trait and/or state empathy from a neuroimaging perspective. Zhao et al. (2022) identified a neural network representing state empathy that included the bilateral middle cingulate cortex, the bilateral supplementary motor area, the left inferior frontal gyrus, and the anterior insula. The authors could link intrinsic brain activity in these regions to trait empathy measures conducted using the IRI (Davis, 1983). Yet, as far as we know, no neural network representing trait empathy distinctively from state empathy has been identified. We decided not to include neuroimaging measures as a third category in this review as studies approaching empathy from this perspective usually combine their measures with a self-report and/or a task-based approach to measure empathy.

It is important to apply caution if considering most measures as indicators of trait empathy. Performance and self-report results may be subject to the influence of situational or contextual factors during data collection. Therefore, it is crucial to interpret such measurements with care and account for potential sources of variability. In sum, Figure 1 shows that indications, of whether existing tools measure trait and/or state empathy are lacking, a problem that may partially explain inconsistencies in past empathy research and that can easily be addressed. Our above categorization enables researchers to select appropriate tools for their research questions and study designs, and facilitates the comparison of findings across studies.

In addition to differentiating between trait and state empathy, we further summarized whether respective measurement tools address the cognitive and/or affective domains of empathy. Older empathy measurement instruments in particular tended not to address this difference. This might be due in part to more recent theoretical developments since the differentiation between cognitive and affective empathy may have been established after the development of older tools.

Taken together, the question remains whether or not, and to what extent, existing measurement tools should be adapted. We argue that separating trait versus state empathy does not necessarily lead to new measurement tools. Instead, two advances should be made: (1) authors should clearly state if their experiment targets trait and/or state empathy, choose the appropriate tool, and argue why the tool is suited for the trait and state dimension; and (2) when performing empathy experiments, authors should gather information on situated influences that potentially modulate state empathy and include them as co-variances in the analyses. For this to happen, a consensus needs to be reached on which individual factors modify state empathy. This topic will be discussed in the next chapter.

As discussed earlier, current theoretical approaches consider empathy to be a capacity with expressions that are situated within and influenced by contextual factors (Hall and Schwartz, 2019). Both affective and cognitive empathy are assumed to include a bottom-up process as well as a top-down control mechanism (de Waal and Preston, 2017; Hall and Schwartz, 2019; Preston and de Waal, 2002; Singer and Lamm, 2009). Empathy for pain is the main area considered in terms of neural correlates of empathy. In this context, studies show that neural networks related to empathy processes (mainly insula and anterior cingulate cortex, ACC) are activated even when participants are not explicitly asked to emphasize. This is understood as the initial bottom-up process that cannot be controlled by individuals (Singer and Lamm, 2009). Only later can the empathy process be regulated by top-down control mechanisms. Although a clear distinction between bottom-up and top-down processes is challenging to determine empirically, a theoretical distinction between the two holds value by providing testable predictions and hypotheses for future studies. Furthermore, several studies illustrate that the empathic reaction after stimulus onset alters over time, indicating that there are fast, intuitive, and slow deliberative processes influencing the final expression of empathy (for an overview of examples for both processes in more detail see Singer and Lamm, 2009).

Empirical studies on empathy for more complex emotions are scarce. Despite this, it can be expected that complex emotional mechanisms work similarly. Building on the approach taken by Thompson et al. (2019), we conceptualize empathy as a process by adding the trait versus state perspective. We propose the following idea: when an empathic reaction is initiated (for example by an external stimulus like a crying family member), the trait empathy measure equals the general ability and functions as a stable multiplicator that determines the magnitude (steepness) of the initial, intuitive bottom-up process of the empathic reaction. In addition to trait empathy, various situational factors influence the intensity of the initial, intuitive bottom-up empathic reaction. For example, Morel et al. (2012) report contextual influences on face perception as early as 60 ms. Considering the short time frame, this influence can be attributed to the bottom-up process. Concerning top-down control mechanisms, different emotion regulation strategies reportedly either increase or decrease state empathy (Jauniaux et al., 2020; Thompson et al., 2019). Furthermore, similar to the bottom-up process, contextual factors can influence the top-down process; these situational factors may lead either to an increase or decrease in the initiated intensity of an empathic reaction. Consequently, the result of this process is the situated expression of trait empathy, measured in most studies investigating empathy (state empathy). We consider this approach to work equally for both cognitive and affective empathy processes. For a visualization, see Figure 2.

The presented idea implies that there is an inherent difference between measuring the empathic reaction shortly after the stimulus onset or after a longer time. The first measure reflects mainly the bottom-up process consisting of the general trait capacity while the latter represents the outcome of both the bottom-up process and the top-down regulation. Thus, the crucial variable for measuring task-based empathy is the time frame for indicating an individual’s empathic response. In a recent study, a 5 s time pressure was identified as a valid method to induce intuitive thinking in decision-making paradigms (Isler and Yilmaz, 2023). Although response time is usually recorded in empathy studies, it is rarely interpreted, and its impact is generally underestimated. Measurement standards for short-term empathic measures (for example <5 s) and long-term empathic measures (for example >10 s) are needed; for a comparison see Isler and Yilmaz (2023). Both approaches are important for understanding the different aspects of empathy. However, researchers must decide beforehand which aspect they intend to measure and design their experiment accordingly. We believe that the lack of time control contributes to the large variance between results in studies measuring empathy, and that mitigating this shortcoming could be a simple and cost-efficient way to improve empathy research.

The presented theoretical framework and its implications are in line with the following results: in two studies, individuals with self-reported high trait empathy were compared to individuals with low trait empathy as measured with the EQ (Baron-Cohen and Wheelwright, 2004; Rameson et al., 2012). Participants were instructed to either look at photos and empathize with the people in the photo, or look at photos while remembering an eight-digit number (passive condition but with a high cognitive load). Remarkably, participants with high levels of trait empathy exhibited heightened expressions of empathy under cognitive load compared to participants characterized by lower trait empathy (Rameson et al., 2012). This suggests that under conditions of high cognitive load, individuals with high-trait empathy report higher intensities of empathic reactions than individuals with low-trait empathy. Considering our approach, the steepness of the level of intensity that reflects trait empathy determines the level of intensity of the state empathy reaction. Although the assumption regarding time control remains untested, its validity could be determined if future studies implement two groups in task-based empathy measures like the MET-core-2. The first group would be forced to respond within 5 s of stimulus onset while the second group could not respond until 10 s after stimulus onset. This would ensure that the empathetic reaction of the first group was primarily based on the bottom-up process while the second group’s response was based on a mixture of the bottom-up and top-down processes. If the responses between groups varied significantly, the results would suggest that the distinction between the two processes is valuable and should be made in future research.

It is critical to note that trait and state empathy may not be entirely independent of one another, contrary to what might be inferred from the model presented above. Determining the relationship between trait and state processes—whether they are interdependent or distinct—requires experimental intervention studies. Crucially, such studies must employ comprehensive measurement tools, as previously emphasized, that assess both trait and state components of empathy.

Finally, it is important to have a basic understanding of the factors that increase or decrease the empathetic response when planning an experiment to reduce the risk of opposing factors falsely attenuating the effects of one another. Thus, the following chapters briefly outline increasing and decreasing factors that have been identified in previous studies.

This chapter examines internal/intraindividual and external situational factors that govern the presence of state empathy and serve as increasing factors.

Many of the aforementioned factors have the potential not only to heighten state empathy but also to diminish it. Consequently, in the following section, we summarize experimental conditions employed to reduce state empathy. Results provide valuable insights into the dynamics of state empathy and contribute to the development of more effective empathy measurement tools and interventions.

Considering the complex nature of empathy and the multitude and versatility of factors influencing empathic responses described in the previous chapters, one can easily conclude that measuring empathy poses challenges. To mitigate them, this review developed simple and cost/time-efficient ways to enhance the validity of empathy measurements: (1) Measurement choice: as outlined in Chapter 2, studies investigating empathy frequently use empathy measurements without considering the inherent differences of the accessed concepts. The choice diagram depicted in Figure 1 provides a tool that can be used to decide what measurement should be used in a study to optimally quantify the specific aspect of empathy under investigation. (2) Time control: as outlined in Chapter 3, the time frame within which participants must indicate their empathic response can cause significant differences in results. To overcome this limitation, we recommend a simplified distinction between bottom-up and top-down processes in measuring empathy. To assess a bottom-up process, studies should integrate a forced response time for the empathic response, triggering a fast, intuitive reaction. To assess the general empathic response (consisting of bottom-up and top-down processes) studies should implement a time period during which participants cannot respond, ensuring that the top-down process has time to occur. (3) Confounding factors: as outlined in Chapter 4, the measurement of empathy is sensitive to several factors that can increase or decrease the empathic response. Since there is a risk that factors act in opposition to each other (meaning that factors that both increase and decrease empathic responses are present) causing null effects, it is important to control for the confounding effects at least to some degree. To facilitate the selection of appropriate study designs and measurements, Table 2 provides researchers with an overview of how different factors can influence empathy. We believe that the consideration of the three aforementioned elements (measurement choice, time control, and confounding factors) will enhance the validity and generalizability of empathy research results.

One interesting approach that more explicitly integrates biopsychological indicators would be to establish standardized test batteries that allow for the measurement of both cognitive and affective empathy through self-report (for trait empathy), performance-measures (for state empathy), and CNS-correlates (such as fMRI- or EEG-measures). One such battery is the so-called EmpaToM task (Kanske et al., 2015) for use in fMRI. The EmpaToM, which assesses cognitive and affective empathy and compassion, was shown to discriminate between the three corresponding types of neural pathways crucial for understanding others within the same functional-imaging task (initially developed in German; Kanske et al., 2015). Concretely, the neuronal correlates of affective empathy, ToM (theory of mind, i.e., cognitive empathy), and compassion are assessed while using dynamic video sequences of neutral and negative valence (related to suffering). The EmpaToM is now also available in English (Lantos et al., 2023). We would add that it is important to extend its scope by examining complex emotions of both positive and negative (and neutral) valence rather than focusing only on negative emotions (and neutral controls).

This review aimed to outline the benefits of differentiating between trait and state empathy on a theoretical and methodological level. In addition, we challenged current methodological approaches for measuring empathy. We elaborated on the theoretical aspects of trait and state empathy and discussed both increasing and decreasing factors of state empathy. Finally, we highlighted three factors that should be taken into consideration when designing future empathy studies.

We are aware that the current review comes with limitations. It is noteworthy that more aspects than those listed in this review shape empathy. For example, Weisz and Zaki (2017) summarize that expectations of the emphasizer due to their gender play a pivotal role, but only when the gender-related expectation is made conscious. In addition, and as stated above, other studies report that the valence of emotions interacts with the female’s menstrual cycle stage and therefore specifically affects empathy (Gamsakhurdashvili et al., 2021b; Wolf et al., 2015). While this paper’s selection of individual factors is literature-driven, it is not conceptualized as systematic due to the lack of previous approaches taken to address the topic. Despite these limitations, the review opens the possibility for future researchers to assess the distinction between trait and state empathy. Future research should aim to clarify inconsistencies in methodological approaches used to measure empathy.