1 Introduction

“Empathy” is a dazzling term. It has various connotations, both moral and epistemic.Footnote 1 Empathy might not only refer to pro-social behavior and a feeling of distress in regard to others who suffer, but also to a form of understanding of other people. It is therefore not surprising that ever since the term has been coined at the end of the 19th century by the aestheticist Robert VischerFootnote 2 empathy has been quickly linked to philosophical problems concerning other minds and intersubjectivity. This connection has seen a recent revival in the theory of mind (ToM) debate. For the last 10 years, one could notice a strong re-emerging interest in empathy. What is surprising, however, is that authors from what are typically described as opposing camps claim that empathy is central to their respective views. In particular, proponents of both simulation theory (ST) and authors arguing for the so-called phenomenological proposal (PP) (Zahavi 2011) have put the notion of empathy at the centre of their approaches as the three following quotes demonstrate:

“Empathy [is] again being taken seriously by philosophers, psychologists, and neuroscientists and is being defended in a variety of forms by so-called simulation theorists.” (Stueber 2006, p.4)

“In this section, I turn to a principal rival of theory-theory, namely, simulation (or empathy) theory.” (Goldman 2006, p.17)

“[…] the empathic approach. This approach […] argues for the existence of a specific mode of consciousness, called empathy, which is taken to allow us to experience and understand the feelings, desires, and beliefs of others in a more-or-less direct manner.” (Zahavi 2001, p. 153)

There are at least two possibilities to make sense out of this puzzling fact: On the one hand, it might be the case that proponents of ST and PP agree that empathy is crucial, but defend different mechanisms by which it is brought about: simulation versus direct perception. This way of framing the debate is implicit in e.g. de Vignemont and Jacob’s (2012) contrast between their simulation-based approach and Zahavi’s (2008) direct-perception view—and it is not implausible given the fact that both approaches allow delineating empathy from the three closely related phenomena of emotional contagion, emotional sharing and sympathy. On the other hand, however, it might also be the case that these authors merely employ the very same term, but do in fact talk about different phenomena which they consider to be crucial for understanding others. Given the long-lasting lack of a clear and agreed upon definition, the term “empathy” might even be especially prone to be used as a label for divergent research projects.

In the first part of this paper, I make a case for the latter possibility and argue that the notion of empathy is indeed used inconsistently in the ToM debate. To do so, I will focus on the simulation-based approach to empathy which has been proposed by de Vignemont and Jacob (2012) and compare it to the direct-perception view which has been defended by Zahavi (2007; 2008; 2011; 2012a; Zahavi and Overgaard 2012). I contend that the conflict between these approaches does not turn out to be a disagreement on the mechanism by which empathy is brought about. Rather, it is to be traced back to the fact that their proponents attribute two different roles to empathy. Whereas Zahavi takes empathy to be fundamental to any kind of access to other people, de Vignemont and Jacob take empathy to be a very special form of understanding others. By restricting themselves to either one of these two roles, these authors ultimately talk about different phenomena when using the term “empathy”.

The result of the first part could be used to discuss in a next step the deeper reasons motivating certain terminological decisions.Footnote 3 This would suggest, however, that one would have to decide between either using the term “empathy” to describe a basic form of access to other people or a further-reaching kind of understanding. This dichotomy is, I believe, misleading, and that is why I introduce in the second part of my paper Edith Stein’s account of empathy (1917/2008). Stein’s proposal is most interesting for the current debate precisely because she does not restrict the notion of empathy to either one of the two mentioned roles. In what could be called a process model of empathy, she combines some of the seemingly divergent intuitions about empathy that have been brought forward by defenders of ST and PP respectively. By not allocating her account to either one of these two sides, I aim at showing that both simulation-based approaches and the direct-perception view can profit from her very delicate phenomenological descriptions. I conclude that thus taking Stein’s account seriously points to one way of integrating the discussed accounts of empathy and of thereby overcoming the persistent thinking along the established camps of the ToM debate.

2 A Simulation-Based Account of Empathy

Before starting off with the first proposal—de Vignemont and Jacob’s model of empathy—one preliminary remark is due: The examined approach is just one among a rather heterogeneous group of simulation-based approaches in the ToM debate. The notion of simulation is an umbrella term which is used to refer to the fact that the psychological and / or neurological states of two people are isomorphic to a certain degree. The role this isomorphism is meant to play and the precise way it is explicated differs, however, to such a large extent that the details of any so-called simulation-based approach are crucial for its evaluation. As we will see in what follows, de Vignemont and Jacob’s model is not to be equated to either Gallese’s mirror matching mechanism (2001) nor to Goldman’s hybrid view (2006; 2011) to name just two other prominent simulation-based approaches.

Elaborating her previously argued positions (e.g. de Vignemont and Singer 2006; de Vignemont 2010), de Vignemont focuses in a paper co-authored with Jacob on pain as her paradigmatic example. The overall aim is to distinguish two kinds of vicarious pain, empathic pain and contagious pain, on the basis of both neuropsychological evidence and conceptual reasoning. One major part of the theoretical work consists in the development of a general definition of empathy for which de Vignemont and Jacob propose the following five conditions:

  1. i.

    The empathizer E is in some affective state s*.

  2. ii.

    E’s affective state s* stands in some similarity relation to the target T’s affective state s.

  3. iii.

    E’s being in state s* is caused by T’s being in state s.

  4. iv.

    E’s being in s* makes E aware that her being in s* is caused by T’s being in s.

  5. v.

    E must care about T’s affective life.

By means of these conditions de Vignemont and Jacob intend to distinguish empathy from four related although different psychological phenomena: The first condition is meant to differentiate the empathizer’s experience from the mere belief that another person is in a certain mental state (standard mindreading). The second condition serves to distinguish empathy from sympathy. A sympathizer also reacts affectively in regard to another person—she might, for example, feel sorry for another person being in pain—, but her emotions are different from the ones the target has. With their third condition, de Vignemont and Jacob are able to distinguish empathy from coincidental emotional sharing, i.e. a situation in which two people are, independently of each other, in similar affective states due to a common cause. Conditions four and five, finally, are the decisive ones to discriminate empathy from emotional contagion. A person who experiences e.g. contagious pain does not need to be aware of the fact that her experience has been caused by another person’s pain. In fact, she does not need to even know that another person is in pain at that moment—she might simply “catch” the emotion and experience pain as well. Furthermore, de Vignemont and Jacob assume that empathy is, in contrast to emotional contagion, not an automatic reaction. It presupposes in their view a certain pro-attitude in regards of others and can be modulated by a variety of top-down factors.Footnote 4

As the second and the fourth condition will turn out to be the crucial ones in our context, let me elaborate on them in further detail. The second condition has two important implications: First, it restricts de Vignemont and Jacob’s definition of empathy to affective states only. That is, one can only empathize with another person’s emotions, not with her beliefs or intentions. Second, this condition—which has also been called the isomorphism or interpersonal similarity condition of empathy—makes their proposal a simulation-based approach. The differences between these approaches lie, as already indicated, in the details such that one has to additionally ask: How and in what sense do the empathizer’s and the target’s mental states resemble each other? What precisely is meant by the notion of simulation? Unfortunately, de Vignemont and Jacob do not provide any additional specification of their second condition in the context of the general definition of empathy. They do make, however, certain suggestions in this regard in relation to their leading example, empathic pain: Besides the somewhat obvious point that standard pain and empathic pain differ with respect to their causes—whereas standard pain is caused by an injury, empathic pain is caused by either the perception or the imagination of someone else’s pain—de Vignemont and Jacob mention three other aspects: First, they endorse the view that empathic pain is the result of a process of so-called enactment-imagination, i.e. a process in which one creates a certain mental state “by running off-line one’s own pain systems” (2012, p. 298). This implies that the brain systems which are activated in standard and in empathic pain overlap. Second, they qualify this implication by claiming that empathic pain is restricted to the activation of an affective pain component only whereas standard pain paradigmatically involves the activation of both an affective and a sensory-motor component. This again means that the empathizer’s pain is experienced as being unpleasant, but not in relation to a precise bodily location and without the concomitant preparation for motor action which is typical of standard pain. Finally, de Vignemont and Jacob state that empathic pain is other-directed. This is meant to indicate that empathic pain has a certain representational content, and that the subject whose mental state is represented is not oneself, but another person. Putting these features together yields the following characterization of empathic pain: Experiencing empathic pain is the result of imagining another person’s pain. This act of imagination involves the activation of parts of one’s own pain system. Therefore, experiencing empathic pain hurts—one painfully represents another person’s pain. If and how this characterization can be generalized to other cases of empathic experiences will be further discussed in the next section.

The fourth condition is de Vignemont and Jacob’s strategy against certain counterexamples which could be brought forward against the simpler formulation “iv. E is aware of T’s being in s”. What is important for what follows, however, is that both formulations imply that the empathizer already has a certain understanding of the target’s mental state when she empathizes. The affective state which the empathizer experiences is not the means by which she becomes aware of her target’s mental life in the first place. Rather, empathy in de Vignemont and Jacob’s sense already presupposes knowledge about the other person and her mental state. Empathy is thus conceived as a special form of knowing what another person experiences, as a way of thoroughly developing one’s understanding by recreatively imagining what this person “goes through”—a view which de Vignemont also characterizes as follows: “Empathy provides access to other people’s mental states from a first-person perspective” (2010, p. 290). Turned around, this means that empathy is far from being the default response to another person’s affective experiences. We might, for example, be aware that another person is happy, sad or in pain, but, given that we do not really care, we do not empathize with her. Precisely this idea finds its expression in the fifth condition of de Vignemont and Jacob’s definition. This understanding of empathy as a very special kind of access to someone else’s affective life is one feature which allows distinguishing this proposal from the simulation-based accounts given by Gallese (2001) and Goldman (2006, 2011). Both Gallese and Goldman take empathy to be one—and in their view the most important—form of mind-reading tout court, i.e. a way of initially becoming aware of someone else’s experiences. Jacob (2011, p. 525), by contrast, has explicitly drawn a distinction between simulation-based approaches to mind-reading and simulation-based approaches to empathy to make it clear that he takes simulation to be essential for the latter, but not for the former.

3 Against the Simulation-Based Account

Based on resources from the phenomenological tradition, Zahavi has presented various arguments against simulation-based approaches to mind-reading and to empathy (Zahavi 2008, 2011; Zahavi and Overgaard 2012).Footnote 5 Given that not every argument suits every simulation-based account, I will focus on the extensive discussion he gave of the proposal of de Vignemont and colleagues in which the two most telling counterarguments concern conditions two and four respectively.

In order to put pressure on the isomorphism-condition, Zahavi (2011) invites us to consider the following two examples: Imagine a person who sees that her friend is sad and feels sad for her. Or imagine a person who realizes that her friend is angry and gets angry in her turn. In both cases, Zahavi claims, de Vignemont and Jacob’s five conditions are fulfilled, but neither case seems to him to be rightly called a case of empathy. Why? In both cases it seems to be a more or less arbitrary coincidence that the (alleged) empathizer reacts with a similar kind of affective state as the target. She might have also reacted with concern, with fear, with happiness or any other emotion, depending on the situation. What then, Zahavi asks, motivates the claim that the emotional reaction should be given a special name (and, possibly, a special role in understanding others) once it is similar to the affective state the target experiences? I think that there is an obvious way for de Vignemont and Jacob to dissolve Zahavi’s worries: Neither of the cases he considers should be called empathy in their sense because the affective states of the target and of the (alleged) empathizer are not sufficiently similar. Although the quality of emotion is the same, the intentional object is different for both people. The friend might, for example, be sad about a bad exam, but Zahavi’s so-called empathizer would be sad about the friend. The second case is even more obvious: Whereas the friend might e.g. be angry about the teacher who has given her a mark, Zahavi’s so-called empathizer would have the friend as her object, i.e. get angry about the friend herself (not the teacher). I concede that this strategy of distinguishing the empathizer’s and the target’s emotions by reference to their intentional objects is not explicitly mentioned in de Vignemont and Jacob’s proposal.Footnote 6 I believe, however, that this strategy conforms well to their account.Footnote 7 Empathizing in de Vignemont and Jacob’s sense means trying to imagine what a certain situation feels like for the target. Empathizing with a sad friend means imagining her sadness about the mark and, to a certain extent, feeling sad about the mark, too. Empathizing with an angry friend means imagining her anger and, to a certain extent, feeling angry about the teacher, too. One reason for Zahavi to have misunderstood de Vignemont and Jacob’s intention might lie in the fact that he took their notion of other-directedness too literally. This notion is not meant to indicate that the other person is the intentional object of the empathizer’s affective state; it rather stands for the fact that the empathizer is willing to share another person’s experience.

A second worry expressed by Zahavi and Overgaard (2012) concerns the following potential vicious circle:

  1. 1)

    De Vignemont claims that empathy should enable us to understand what others feel (e.g. de Vignemont and Singer 2006, p. 439).

  2. 2)

    The simulation-based account of empathy presupposes in condition fourFootnote 8 the understanding of the other’s feelings.

  3. 3)

    Empathy cannot do its supposed job.

However, the circularity problem can be easily dissolved if the notoriously ambiguous term “understanding” is closer scrutinized. Zahavi (2007, 2008, 2010) explicitly employs the term in such a way that it describes our initial awareness of what another person experiences. Whereas it is clear that precisely this notion of understanding is involved in the second premise of the argument, the crucial question concerns the first premise. If de Vignemont’s quotation from an earlier paper conforms to the way she presents the matter in the proposal made with Jacob, she should be read as stating that empathy allows an enhanced understanding of someone else’s feelings. If so, the counterargument is simply not conclusive because it conflates two different notions of understanding. Nevertheless, this discussion reveals that the deeper disputed issue lies in the fact that empathy is given two different roles in understanding others: Zahavi assumes that empathy is the basic form of interpersonal understanding, that it is foundational for any other kind of access to other people’s minds. De Vignemont and Jacob, by contrast, take empathy to provide a very peculiar form of understanding another person from within. But who is right? What influences one or the other terminological decision? One considerable factor is the philosophical tradition on behalf of which these authors speak, and we will have a closer look at the phenomenological tradition down below. But another, more substantial reason that motivates Zahavi’s dissatisfaction with a narrow and cognitively rich notion of empathy might be the very fact that such an account entails a lot of presuppositionsFootnote 9: De Vignemont and Jacob do not specify in condition three how the target’s being in a certain affective state causes the empathizer to be in a similar state. Even more, they do not specify in condition four how the empathizer initially becomes aware of the target’s mental state. Zahavi seems to fear that these subtle and basic questions tend to be overlooked if the role of empathy is narrowed down to a higher-order form of understanding. His own account of empathy, by contrast, is meant to be able to address precisely these questions.

4 The Direct-Perception View of Empathy

In order to be able to understand Zahavi’s positive account of empathy, it is necessary to have a closer look at his background, that is to say the phenomenological tradition with Husserl as its main point of reference.Footnote 10 Two very general observations are relevant for the context of this paper: First, although some phenomenologists in the circle around Husserl concede that the term “empathy”, taken in the strict sense of the word, refers to the understanding of affective states only, these philosophers have a much broader conception in mind. As Stein has put it

“The comprehension of foreign mental states [Erlebnisse]—be they sensations, feelings, or what not—is a unified, typical, even though diversely differentiated modification of consciousness and requires a uniform name” (1917/2008, p. 78, my translationFootnote 11),

they assume to be dealing with a single, unified capacity for understanding someone else’s mental states, and this capacity, they stipulate, is called empathy.Footnote 12

Second, the proposals which have been given by scholars such as Geiger, Stein, Scheler and Husserl are to be understood as more or less critical reactions to both the argument from analogy and to suggestions earlier made by Theodor Lipps. Among other things, Lipps (1907) argued that we come to understand the mental states that another person expresses by the following three-step model: First, we rely on our inborn tendency to imitate their expressions which, second, recreates a similar kind of experience in us which, third, we project onto the other person. Lipps explicitly states that his account implies that what we know of each other are in fact just our own experiences, projected onto the other person. Although later phenomenologists share Lipps’ intuition that empathy is—contra defenders of the argument from analogy—an experience,Footnote 13 many of their critical comments focus on one general theme: Lipps’ account seems to them to be too much self-focused. It does not do justice to the fact that empathy is—as they assume—the capacity to really understand another person in her very own way of being. The experiential character of empathy and the attempt to explain how to get access to another person’s mind thus seem to me to be the two core characteristics which scholars from the group around Husserl are trying to capture in their respective accounts.

As we will see in what follows, their strategies differ to a substantial degree. In particular, I intend to show here and in the following section that Zahavi’s direct-perception view—which tends to be received as “the” phenomenological account nowadays—is very much influenced by the work of both Scheler and Husserl (see esp. Zahavi 2008), but neglects certain important developments we find in the work of Stein. To do so, I first try to give a systematic overview of the proposal Zahavi defends on behalf of the phenomenological tradition before pointing to relevant differences to Stein’s account of empathy in the next section. The basic intuition by which Zahavi starts off—and which allows his proposal to be called a direct-perception view of empathy—is that empathy is to be conceived in analogy to perception. Perception seems to be the ideal model because it helps to understand how empathy could have the two just-mentioned features phenomenologists are so eager to accommodate: There is no denying that perceiving something is an experience which—at least if we follow our naïve realism—allows us to “reach out” to the external world and to gain access to something beyond ourselves. Whereas Scheler takes the analogy in his perceptual theory of other minds literally, Zahavi seems to follow both Husserl and Stein in being more cautious: Empathy must at least be a special form of perception because mental states are obviously different entities than external objects and events. Zahavi thus turns to Husserl’s idea of appresentation and to his suggestion that the perception of mental states might be more or less similar to the perception of the hidden aspects of an external object. As we are able to co-perceive this hidden profile together with the aspects which are in plain view we can also say that we are able to co-perceive the mental states of others together with their natural expressions. Of course, this analogy is not perfect either. The hidden aspects of an object can come into view if we e.g. move around the object, whereas mental states cannot because there is no way of—literally or metaphorically speaking—getting beyond the expression. Contrary to external objects, mental states have aspects which are only accessible to the person who has them and are in principle inaccessible to another person.

At this point, Zahavi seems to be undecided as what to do with this more or less ideal analogy between empathy and perception. On the one hand, he heavily relies on the analogy when arguing against simulation-based approaches to mind-reading. Defenders of the isomorphism-condition, Zahavi reckons, assume that simulation is necessary “in order to make the leap from the perceptual input […] to the output, which is the ascription of mental states to the others.” (Zahavi 2012a, p. 247). This assumption is wrong, he claims, because it is unnecessary: Once it is accepted that we are in fact able to perceive mental states, there is no need to resort to any kind of simulation, be it imitation or imagination. While this argument clearly involves the claim that we have a perceptual access to someone else’s mental states, Zahavi repeatedly resorts to a weaker claim when confronted with the above-mentioned worries that mental states differ in important respects from external objects and events:

“When claiming that we are able to experience others, and as a consequence do not exclusively have to rely on and employ inferences, imitations or projections, this is not meant to entail that we can experience the other in precisely the same way as she herself does, nor that the other’s consciousness is accessible to us in precisely the same way as our own is. Second- (and third-) person access to psychological states do differ from first-person access. But we shouldn’t make the mistake of restricting and equating experiential access with first-person access. It is possible to experience minds in more than one way.” (2010, p. 295, emphasis in original see also 2012a)

As long as it is granted, Zahavi seems to think, that we can have an experiential access to someone else’s mind, nothing in particular hangs on the claim that this access is perceptual. The essential point is that it is experiential although nevertheless different from the access we have to our own mental states. Zahavi presents this line of reasoning as the minimal common ground between the proposals of Scheler, Husserl, and Stein, and I think he is right in so doing. But it seems to me that this is no more than stating the intuition which all these scholars share: They think that empathy is an experience. However, the crucial question is: How is the idea of an experiential access to other minds, the idea of empathy being an experience to be understood if not along the lines of a direct perceptual access? It is precisely at this point that Stein’s account of empathy provides additional insights.

5 … is not “the” Phenomenological Proposal

Edith Stein was a pupil of Husserl’s who had, after having read his “Logische Untersuchungen”, joined him in 1913 in Göttingen to pursue her PhD. Having taken classes both with Husserl and Scheler, Stein quickly decided to work on empathy—a term which Husserl had already used in reference to Lipps’ work—which she considered an important, but underdeveloped topic in Husserl’s thinking (see e.g. Stein 1985/2002, p. 218f.). Having reconstructed the history of this idea from Herder to the beginnings of the 20th century, Stein (1917/2008) developed her own systematic account of empathy by delineating it not only from Mill’s argument of analogy and Lipps’ imitation-based account, but also from the proposals given by Max Scheler and Hugo Münsterberg. In so doing, she aimed at giving a detailed phenomenological description of the phenomenon at issue, and explicitly denied to be interested in the potential mechanisms by which it is brought about (p. 14). Husserl commented on her work at different phases but considered it in the end not as a pupil’s work, but as an independent contribution.Footnote 14 His appreciation not only resulted in a PhD with distinction, but also in a job as his assistant in Freiburg from 1916 to 1918.

Stein’s strategy in developing her proposal is to compare the phenomenon which she takes to be empathy to various other kinds of intentional acts. Taking up the discussion as to whether empathy is a perceptual act, Stein considers this idea in the very first section of the systematic part of her work, and she concludes:

“Thus empathy as the comprehension of the mental state in itself does not have the character of outer perception.” (p. 15)

Stein maintains this claim for the very same reasons which we have already considered in the previous section: In perception, spatiotemporal objects and events are—as she calls it—given primordially (originär) to us, i.e. in their fullest presence. As someone else’s mental states have aspects which are in principle inaccessible to us, they cannot be given primordially to us.Footnote 15 To make things a bit more complicated, however, she immediately adds in her author’s copy:

“The complex act, by contrast, which co-comprehends the expressed mental state with the bodily expression must arguably be called outer perception. The primordially given expression ‘appresents’—as Husserl likes to say—the mental state which exists in the here and now as ‘something that is given along’.” (p. 15)

The addendum is far from being the only one in which Stein added some qualifications to her original rather strict dismissal of empathy as being perceptual (see e.g. also p. 19 and 20). One can only speculate about her motivation to do so, but it seems plausible to me that Stein tried to take her supervisor’s views into account and to point to a possible way of combining her and Husserl’s ideas.Footnote 16 One way to make sense out of these remarks is to assume that Stein distinguishes between two aspects both of which form part of the phenomenon of empathy: The act of becoming aware of someone else’s mental state on the basis of its expression and the understanding of the mental state in itself. Stein does not provide any further explanation as to what precisely is meant by the latter, but it is clear that she must have an enhanced form of understanding in mind which goes beyond the initial awareness. She thus works with a notion of empathy which is broad on two counts: She does not restrict it with respect to the kind of mental state that might be empathized with, and she does not restrict it with respect to its role of either providing the initial awareness or an enhanced understanding of someone else’s mental state.

But how, precisely, does Stein characterize empathy? How does she combine these different aspects of empathy without making it look like an arbitrary mixture of varying intuitions? Fundamental to Stein’s account is the insight that empathy is to be conceived in analogy to acts such as memorizing, expecting, and imagining something.Footnote 17 In contrast to perception, these acts do not present their objects as being primordially given, but rather re-present (vergegenwärtigen) or reproduce (reproduzieren) them in one way or another. Narrowing the phenomenon further down, Stein claims that empathy is somewhat similar to acts in which one’s own mental states are remembered, expected or imagined,Footnote 18 and she goes on to compare these acts to one another: Memory and expectation differ from imagination and empathy generally with respect to their temporal structure. Whereas there is a clear temporal distance between the moment at which we remember or expect something and the moment at which the remembered or expected event takes place, we do not find a similar delay in imagination and empathy. Rather, whenever we imagine something or empathize with someone, the phantasm or the empathized mental state is experienced as being present. Imagining one’s own mental states can, however, be distinguished from empathy on the basis of two further features: First, whereas an imagined mental state is experienced as unreal, as a mere possibility (e.g. “If I was in situation XYZ, I would be angry.”), an empathized mental state is posited as existent. The other’s mental state is not “merely imagined, without real life” as Stein writes on page 20, but is experienced as being real. Second, whereas imagining our own mental state involves our own self, empathy involves another subject. In short, then, empathy is for Stein an act which represents someone else’s mental state non-primordially, but as now existing. The following Table 1 summarizes again the aspects which Stein takes into account in order to have reached this conclusionFootnote 19.

Table 1 Stein’s distinction between different kinds of mental acts
Full size table

These differences notwithstanding, Stein also highlights a fundamental similarity between all the acts which represent a mental state as being non-primordially given: Expecting, remembering and imagining our own mental state as well as empathizing with someone else’s mental state ideally proceed in three steps or modalities of accomplishment:

  1. 1.

    “The emergence of the mental state,

  2. 2.

    The fulfilling explication,

  3. 3.

    The comprehensive objectification of the explained mental state.” (p.19)

Applied to the case of empathy, this means that we

  1. 1.

    Are vaguely aware of someone else’s mental state,

  2. 2.

    Follow a tendency to be drawn into this state, and

  3. 3.

    Objectify the mental state in an “apperceptive grip” (p. 17) of it.Footnote 20

What is particularly interesting for the context of this paper is the way Stein describes the second phase. If we follow our tendency to fully explicate or understand someone else’s mental state, it is, Stein claims, no longer given as an object to us. We are—as she states metaphorically—guided by the other person into her perspective onto the world and focus, as she does, on her intended object. We are drawn into the other’s mental state, we re-live it. To a certain extent, Stein thus endorses in this second phase the isomorphism-condition. Consequently—and in contrast to the way Zahavi and Overgaard (2012) suggest—Stein does not criticize Lipps for arguing for an inner resonance between the empathizer and her target. Rather, she thinks that that their respective accounts overlap at this point. Of course, the isomorphism between empathizer and target can be explicated in very different ways, and some of her critical remarks on Lipps’ account concern precisely his way of elaborating it. For example, she disagrees with Lipps on what the ideal or fulfilled form of empathy (volle Einfühlung) consists in because Lipps claimed that empathy is fully developed if and when the empathizer experiences the very same mental state as the target (i.e. if, for example, the empathizer is herself happy about the event about which the target is happy). Translated into Stein’s framework, this means that the mental state in question should not only be given primordially to the target, but also to the empathizer. Stein concedes that this might happen, but she does not consider it as a case of empathy any more. In her view, empathy always presents a mental state as non-primordially given, even in the second phase. We are always well aware of the fact that the mental state we share is not our own, that it “does not issue live from my ‘I’” as Stein writes (p. 20), but that it is the representation of someone else’s state which is given in its fullest only to the other person. Connectedly, Lipps thinks that the empathizer and the target merge in full empathy in such a way that the empathizer does not distinguish any more between her own and the other’s mental state. By contrast, Stein denies that this identification is a structural feature of empathy.

Besides the isomorphism, another feature of the second phase is worth to be highlighted. Whereas Stein concedes—although again in a later addendum in her author’s copy on which she does not elaborate any further—that the first and the third phase could be described as forms of perceiving someone else’s mental state in Husserl’s sense, she is quite clear that the second phase is fundamentally different. Why? In the second phase, the empathizer does not any more explore the target’s mental state “from the outside” as the analogy to perception would suggest, but rather joins into the target’s take on the world. It thus seems to be precisely this phase to which Stein referred when she initially denied that empathy as the understanding of the mental state in itself had the character of outer perception. Stein does not content herself with the (more or less ideal) analogy to perception and Husserl’s ideas of appresentation precisely because she takes this enhanced or higher-order form of empathy (höhere Vollzugsstufe der Einfühlung) into account.

6 Stein and the Contemporary Debate

After this introduction on Stein’s account of empathy, two main differences to Zahavi’s direct-perception view should be obvious: First, Stein does not restrict the notion of empathy to what Zahavi calls the basic understanding of another person. She presents a three-step model of empathy which is meant to also yield a higher-order form of understanding. I think that it is precisely this further reaching role of empathy which has been neglected in today’s phenomenological proposals. Second, Stein is not only well aware of the drawbacks of the analogy between empathy and perception, but she contributes to the debate by making a positive alternative proposal: Empathy is in her view better understood on the basis of an analogy to acts which present mental states as being non-primordially given. In particular, she claims that the case of imagining one’s own experiences is an act which shares many—though not all—features with the case of empathizing with someone else’s experiences. Both acts involve a moment in which we re-live an imagined or empathized mental state before we summarize the gained insights in an apperceptive grip.

Going one step further, I take the second phase to be the main reason to believe that Stein shares certain intuitions with today’s proponents of simulation-based approaches, in particular with de Vignemont and Jacob. Like Stein, these authors are after an enhanced understanding of another person which provides some sort of an inside view. The shifting focus which Stein describes in phase two fits with de Vignemont and Jacob’s understanding of the isomorphism-condition as applied to the sharing of affective states. As I have argued in sections two and three, empathizing with another person means for them not (or at least not only) to be focused on the very person, but to adopt and share her point of view onto a certain intentional object. And like Stein, they try to elucidate this aspect by reference to imagination, but not without emphasizing one decisive difference: Compared to imagining one’s own mental state, empathy is clearly special in the sense that it is not one’s own mental state that is represented, but someone else’s. They thus all agree (contra Lipps) that empathy involves the constant awareness of the distinction between self and other.

Despite this substantive overlap, one should, however, be cautious not to overlook important differences between the two proposals. Whereas de Vignemont and Jacob claim that empathy is a form of imagination—enactment-imagination—, Stein states more cautiously that empathy is analogous to imagination and also points to crucial dissimilarities. Empathy is, according to her descriptions, less active than imagination. Stein uses terms such as “being drawn into” and “being guided by” in order to make it clear that an empathized mental state is—in contrast to a phantasm—not experienced as something we come up with. It is not experienced as being the result of one’s own considerations, but rather as something that is objectively given. Stein concedes that an empathizer might go wrong in the way she explicates someone else’s mental state. But this consideration only leads her to underline the importance of perception: Perceiving the other person in her way of expressing a mental state—be it in gestures, words, or whatever—not only forms the first step of the process of empathy, but it also helps to confirm, to modify and possibly correct the way the process develops (p. 103). Stein thus explicitly states that perception is a necessary (though not sufficient) condition of empathy (p. 105f.). De Vignemont and Jacob, by contrast, leave it deliberately open as to how the empathizer becomes aware of the target’s mental state in the first place (2012, p. 302). Any piece of information about another person, or so it seems, can lead one to actively initiate the rather self-contained process of enactment-imagination. Perceptual information is not given any prominent role here. Consequently, de Vignemont and Jacob’s proposal allows for the possibility of empathizing with a person one has been told about (e.g. a fictional character) whereas Stein’s account does not.

This point is related to another, more subtle difference: De Vignemont and Jacob are willing to consider the resulting mental state on the part of the empathizer as an experience if it has a certain associated phenomenology, i.e. if it feels a certain way to be in the mental state in question. This is most obvious for their leading example, vicarious pain, but undoubtedly holds for all vicarious affective states. Given their restriction of empathy to the understanding of affective states only, they can thus perfectly consent to the claim that empathy generally (and in contrast to standard mindreading) is experiential, i.e. that it provides an experiential access to someone else’s mind.Footnote 21 As already indicated, Stein and the phenomenological tradition generally are very eager to defend this claim, too, but—and this is the crucial point—for different reasons. Stein provides an indication of her differing usage of the term “experience” in the following quote:

“He [Lipps] stresses the objectivity or the “demanding” character and thus expresses what we mean by designating it [empathy] as a kind of experiential act.” (p. 21)

She calls empathy an experiential act because it is the most concrete (anschaulich) and vivid (lebendig) evidence we have about someone else’s mental state. Empathy reveals step by step in varying nuances its object, the other’s mental state, and it does so by being directed by perception. As de Vignemont and Jacob, Stein contrasts this kind of access to the mere knowledge of someone else’s mental state. The decisive difference, however, is not the fact that empathizing feels a certain way. Rather, empathy differs from knowledge because it a) presupposes perception, and b) refers to an act in the course of which the empathizer explores her target’s mind in the above described manner. Given such a divergent conception, Stein has no compelling reason to deny that also other mental states besides the affective ones can be empathized with.Footnote 22

Two further differences are worth to be highlighted: First, de Vignemont calls the higher-order form of understanding which empathy is meant to provide the understanding of another person from the first-person perspective. I believe that Stein would have a certain amount of reservation concerning this very notion, given that she is eager to point out that the empathizer’s experience has, even in the second phase, a different quality than the target’s: While the experience of e.g. joy over a passed exam is primordially given to the target, it is non-primordially given to the empathizer; the empathizer is happy by being guided by the target’s joy, but not out of herself. The notion of a first-person perspective seems, by contrast, to conflate precisely this distinction. It suggests that empathizing with someone else’s mental states means sharing the target’s state as if it was one’s own, i.e. to be happy about the exam out of oneself, too. Second, Stein’s model of empathy yields in my view not only a phenomenologically more detailed description of the second phase, but it encompasses two other phases, too. Whereas de Vignemont and Jacob make it sound as if the notion of empathy is to be restricted to the moment in which the empathizer imagines her target’s experience, Stein has the broader context in mind. Being vaguely aware of the other’s mental state in the first step is for Stein not something different from empathy, but part of its process. Stein’s model thus provides a place for what we, following Zahavi, might call the understanding of another person at the most basic level. Whereas de Vignemont and Jacob have to concede vis-à-vis Zahavi that they ultimately have a different phenomenon in view, Stein can easily accommodate Zahavi’s understanding of the notion of empathy by pointing out that empathy involves in her view the basic understanding of others, but ideally comprises more than that.

7 Conclusion

The notion of empathy has seen a recent revival in the theory of mind debate. Both defenders of ST and of PP have endorsed the notion and presented respective accounts. In the first part of this paper I have contrasted two of them, one from each side. What appears to be a disagreement on the mechanisms by which empathy comes about, is, I claimed, ultimately a discussion as to which phenomenon deserves the label “empathy”. This discussion is, as I have tried to show, influenced by certain theoretical preconceptions. In particular, the authors on which I focused—Zahavi on the one hand and de Vignemont and Jacob on the other—disagree on the kind of understanding empathy is said to provide and on how to interpret the notion of other-directedness. The aim of the second part of this paper was then to show that taking Stein’s approach to empathy seriously sheds a new light on this seemingly clear contrast. There are three lessons to be learnt from her work: First, Zahavi’s direct-perception view is far from being the only one which has been considered and defended in the phenomenological tradition. Second, the isomorphism-condition which Zahavi attacks so vehemently has actually been endorsed by phenomenologists, too. Given that this condition can be explicated in very different ways, a closer look at Stein’s analysis finally reveals that it overlaps to a certain extent with the suggestions made by de Vignemont and Jacob. However, instead of aligning Stein’s account to today’s simulation-based approaches, I tried to make a case for taking it as an independent contribution from whose fine-grained phenomenological descriptions both sides of the actual debate can profit.

In the light of these considerations, I shall conclude by pointing out that Stein’s process model of empathy can be seen as a potential way of integrating the divergent intuitions from which Zahavi and de Vignemont and Jacob started off. Following her suggestions, empathy does not need to be conceived as providing either a basic or a more enhanced form of understanding others. This dichotomy is, I believe, a misleading consequence of the way modern accounts of empathy have been located in the broader ToM debate. Rather, the empathizer might proceed from one level to the other. Accordingly, it does not need to be the case that the empathizer is either focused on the other person herself or on the other person’s intentional object, but she might focus on both of them, one after the other. Accepting this idea allows us to furthermore distinguish between aspects of empathy for which the analogy to perception is revealing and aspects for which it is clearly not. If the emphasis lies—as in de Vignemont and Jacob’s account—on the second phase, empathy seems to be fundamentally different from perception because the target’s mental state is no longer given as an object to the empathizer. Rather, the empathizer comes to join in, to share the target’s mental life. As there is a somewhat similar shift of focus in the act of imagining one’s own mental state, there are better analogies to work with besides the perceptual one. If, by contrast, the focus lies on the first phase, perception seems to be an obvious starting point, especially if one assumes—as Stein does alongside with Zahavi—that the target must be physically present: A perceiver as well as an empathizer tries to apprehend something that is there in the here and now. Both start off from what is primordially given, and both have to proceed to aspects of their intended objects which are appresented (if one wants to use this notion in both contexts). The crucial difference, however, concerns the way in which this first apprehension is fulfilled or explicated: Whereas the perceiver can e.g. walk around to perceive the initially hidden profiles of an external object, the empathizer has to proceed to phase two to be able to fully understand someone else’s mental state.

Should we call such an integrative account a “direct-perception and simulation”-account as I have stated provocatively in the title of this paper? No. The intention of this paper was not to show that two of today’s accounts of empathy can be combined without hesitation. Stein’s phenomenological descriptions of the process of empathy are neither appropriately captured by the notion of “direct perception” nor by the notion of “simulation”. Perception does play a role, indeed. And the isomorphism-condition is accepted, indeed. But neither Zahavi’s nor de Vignemont and Jacob’s way of explicating them does justice to the phenomenon which Stein has in mind. Furthermore, the proposed integration should not make us lose sight of the further reaching issues on which the discussed authors disagree. In section five of this paper, I mention two of them, namely:

  • Is the perception of the target a necessary condition for being able to empathize with her?

  • In what sense is empathy to be called an experience?

Both questions seem to me to be of central importance for any account of empathy and thus open up fruitful areas to further research.