Empathy is a contested and ambiguous term, associated with the individual experience of feeling something about another person. If we understand empathy in semantically standard English as “a person's controlled conscious reaction to the emotional experience of another person” (Gladkova Reference Gladkova2010, 280), it is something that requires conscious effort – control. The subjective character of this experience, its character as a form of “feeling,” and the immunity of private feelings to outside checks, makes it a poor candidate as a source of knowledge. Nor is empathy in this strong sense, or its related senses, a particularly stable object of analysis. Empathy is claimed to vary culturally (Aaker and Williams Reference Aaker and Williams1998; Shimuzi 2000). Empathy and empathy related terms, such as sympathy, are semantically different in different languages (Gladkova Reference Gladkova2010). The term itself, and its German source term, Einfühling, are relatively recent creations.

The discussion of empathy has changed radically in recent years as a result of the discovery of mirror neurons and subsequent discoveries of their extensive and special roles in various forms of human interaction. If we try to understand empathy in relation to mirror neurons, things get even more complicated and contentious. On the one hand, there is a range of phenomena that are associated with the mirror neuron system that fits with empathy, including a wide range of measurable neural “mirror” responses to the pain of others, to the movements of others, and to such things as the focus of attention of another on an object. On the other hand, the relation between mirror neurons and conscious thought is in general not understood, and the relation between mirror neurons and empathy in the sense of simulation is itself contentious. As Shaun Gallagher explains in his contribution to this issue, it seems clear that mirror neurons play a supporting role in simulation, which plays a supporting role in conscious empathic thinking, but also that mirror neuron system activation, simulation, and empathy are three different processes, and that mirror neurons alone are insufficient to account for empathy in the most fully elaborated sense of the term.

In this article I will be concerned with the very low level form of “empathy,” the weak sense, which does appear to be closely associated with mirror neurons, does not involve a high level (or perhaps any level) of consciousness, but which is nevertheless significant in relation to some important, indeed fundamental, philosophical issues. My concern will not be with the mirror neuron evidence itself, nor with either simulation or higher level controlled or conscious empathy, but rather with the class of primal understandings of others that occur at a most basic, largely unarticulable, and preconscious level.

Primal Understanding

The philosophical literature often appeals to a primal kind of understanding of others or of ideas, an understanding at a level so basic that it barely seems like understanding at all, and certainly not like “interpretation.” In a much quoted comment on Wittgenstein's concept of rule-following, Rush Rhees said “We see that we understand one another, without noticing whether our reactions tally or not. Because we agree in our reactions, it is possible for me to tell you something, and it is possible for you to teach me something” (conversation quoted by Winch [Reference Winch1958] 1990, 85). The point of these qualifications about “noticing whether our reactions tally” is to avoid the thought that agreeing in our reactions is determined by an inference from the behavior of the other person. The agreement in reactions is either there or it is not: a reaction is direct, not something inferred from something else using some inductive procedure or theory, even an implicit theory. The “agreement” itself may not even need to be recognized. To “see” that we understand each other may simply amount to never having occasion for questioning that we do.

Terms like “agree” in the quotation from Rhees are metaphorical or analogical – agree is a concept imported from law, used here to do double-duty as an analogical description of an act and its normative sequelae; the thought being that agreements once made are to be kept, so that to agree is to accept a rule. There is always a question with such terms, of what is doing the work of explanation and description and what is mere baggage. “Agreement,” as Rhees uses it, like empathy in the strong sense, is one of those middle terms that needs to be separated from its baggage of metaphor, and then reconstructed. In this case, reconstruction would include distinguishing it from empathy in the strong sense, and asking what else might the underlying class of phenomena to which agreement belongs include? As it happens, Wittgenstein, whom Rhees was explicating, makes a parallel distinction in his discussion of rules, in connection with the term interpretation, where he stresses that to follow a rule is not to interpret a rule (Wittgenstein [Reference Wittgenstein, Anscombe, Hacker and Schulte1953] 2009, 86 §198).Footnote ¹

“Following” is a term that points to something that is part of interpretation, but is primary, whereas interpretation, which usually is understood to involve some sort of additional thought, is after the fact. It is thus a term that captures the distinction, without adding some of the baggage associated with these other terms. It points to a class of related terms, which share the characteristic of being primal, tacit, and with conscious or explicit analogues. Some of these have been discussed in the neuroscience literature, such as the phenomenon of joint attention. This is a very rudimentary form of following the thought of another. It does not require “consciousness,” as empathy in the strong sense does, but might be done consciously and intentionally, at which point it becomes a form of empathic thinking.Footnote ² Another closely related case would be ostensive definition. “Definition” is normally explicit, a relation of words with words. Ostensive definition, pointing at something to explain what it means, requires following the intention of another to use pointing to define something and also following the direction of the thought toward the object.

Evidenz and Empathy

Some early twentieth century writers, notably Weber, put the understanding of the actions of other people and the understanding of abstract ideas under the same heading, and used the same epistemic language to characterize the goal of understanding in both cases. In the methodological introduction to the text that was published as Economy and Society (Weber [Reference Weber1922] 1978), he indicated that both the understanding of ideal objects, which for him included abstractions, abstract values, and the like, as well as the understanding of intentional, or, as he called it, meaningful action, sought the same epistemic end, which he called Evidenz.

Weber gives the example of a man chopping wood. We have, he said, direct observational understanding of the subjective meaning of the act – simultaneously for the man and for us, the observer. We do not, as Weber describes this, infer the subjective meaning of the act by deriving it from its elements – the axe, the moving arm, the wood chips flying, and so on. Nor do we infer the content of something invisible, an intention, from these visible elements. If asked for a justification we could invent some such reasoning,Footnote ³ but it would not be what we had actually done. Weber does not say that these identifications cannot be in error, or, indeed, make any other epistemic claim about them. He simply says that “The highest degree of rational understanding is attained in cases involving the meanings of logically or mathematically related propositions; their meaning may be immediately and unambiguously intelligible” and that the situation with empathy, which here he associates with emotion, is parallel: “Empathic or appreciative accuracy is attained when, through sympathetic participation, we can adequately grasp the emotional context in which the action took place” (Weber [Reference Weber1922] 1978, 5).Footnote ⁴

Evidenz thus means “immediate and unambiguous intelligibility.” The term Evidenz has a complex history, which is primarily associated with the notion of self-evidence. In Weber's own time it was associated with Brentano, who traced it from Descartes through many other figures, including John Stuart Mill and some of his contemporaries, such as Christoph von Sigwart. For Brentano as well as Weber, the term refers to a kind of understanding – Brentano used the Kantian term “judgment” – that is non-inferential and ungroundable. Bretano, however, added something striking to the notion of self-evidence that altered it from the Cartesian notion. For Brentano, the two core elements of the idea of Evidenz are these: that which is evident must also be evident to others, and to be evident is also to be something which cannot be further analyzed, justified, or given criteria for.

The same idea appears in many other guises in the analytic tradition. Wittgenstein's distinction between saying and showing, for example, fits the notion of Evidenz: what is shown has the dual features that it cannot be analyzed further and must be “shown” for everyone. Evidenz is philosophically problematic because of this feature of unanalyzability: that which is evident can never be further established, but neither can it be gotten rid of and replaced.Footnote ⁵ Brentano was concerned with the philosophical work that the concept could do, and specifically with one task – providing a solution, in the form of a radical alternative, to the Kantian and neo-Kantian problem of a priori truth. As his student Oskar Kraus explained his argument, it was this:

Brentano's radical idea, in short, is that the standard philosophical strategy of inquiring into the “presuppositions” of what is evident, results, not in some deeper form of a priori knowledge, but in a vicious circle, because our quest for ultimate knowledge can only end in something we take to be evident. Better to acknowledge the ultimacy of that which is already evident. Previous accounts of self-evidence, Descartes’ error, Brentano argued, were to give criteria for calling something evident. This had fatal consequences: it produced a regress in which it was necessary, but at the same time impossible, to produce an adequate justification for the criteria for “evidence.” The trick is to get an explication of self-evidence or being evident that does not itself require justification.

Brentano's account of Evidenz evolved and changed significantly from his most elaborate discussions in The Origin of Our Knowledge of Right and Wrong (Brentano [Reference Brentano, Kraus, Chisholm, Chisholm and Schneewind1889] 2009); his final comments, which include his definition, take the form of a set of fragments. For Brentano, unlike Weber, there were no degrees of evidence – something was evident or not. Evidenz in this sense was restricted to logical and mathematical truths, about which one had “insight” – even knowledge of one's own existence was only “evident to me,” and not evident in the general sense – evident to all (Brentano [Reference Brentano, Kraus, Chisholm, Politzer and Fischer1915] 1966c, 131–2). Brentano noted that “One usually says ‘This is evident’ and not ‘This is evident to me.’ Probably because of the faith that what is evident to me is evident to all” (Brentano [Reference Brentano, Kraus, Chisholm, Politzer and Fischer1915] 1966b, 126). The “all” he defines more precisely in terms of “the person who judges about a thing in the way in which anyone whose judgments were evident would judge about the same thing,” meaning, that anyone who was capable of making judgments with Evidenz on this thing would make the same judgment (Brentano [Reference Brentano, Kraus, Chisholm, Politzer and Fischer1915] 1966a, 122).Footnote ⁶

Because of Brentano's vacillations on the question of what judgments or insights could have Evidenz, some comment is appropriate. The point of Brentano's discussion of Evidenz is to get rid of the notion of presupposed a priori truth or framework. It would be contrary to this program to make a priori or criteria-based claims about what class of judgments could have Evidenz: it is a kind of empirical question as to whether all who understood would judge something to be evident. It would be possible to exclude classes, such as those for which illusion is possible, but that is itself a kind of empirical claim. There can be no problem of relativism for this account: if anyone else would fail to make the same judgment of evidence the fact being judged would by definition not be evident. But fallibilism with respect to what we judge others would take to be evident is appropriate.

The insistence that to have Evidenz was to be evident to all was critical to the argument: it overcame the objection, still formulated by the logical positivist Moritz Schlick, that Evidenz was just a subjective feeling (Schlick [Reference Schlick, Heath, Mulder and Van De Velde-Schlick1910] 1979, 60–1, 98). Another is the suggestion that evident to all is a criterion for evidence. If so, it has the same problems that Descartes’ and Husserl's accounts had, namely that it invites a regress: how would one know that the criterion had been met? It would seem that we would need empathic access to the minds of others to know what is evident to them, and this would require its own criteria and these could not, as Brentano frames the problem, be applied to Evidenz.

This is not a problem for Weber's use of the term, but it is for Brentano's. Brentano rejected the idea of degrees of Evidenz because he was concerned to preserve the idea that Evidenz implied truth. But what we get is something more like this: it is impossible to reject that which is evident, because there is no alternative: anyone who was capable would judge the same way. But if we drop the problematic concern with truth, and add back Weber's use of Evidenz and degrees of Evidenz as they apply to the subjective meaning of action, we can use his basic idea of “evident to all” in a different way: as an account of the kind of understanding we have of such things as subjective meaning, or of the kind of tallying of reactions in the quotation from Rhees. The understanding we have, or fail to have, is an understanding of how others understand the same thing. To have Evidenz, an understanding must also be evident to others, and to all the others who are capable of understanding. This is not a criterion, but a definition of Evidenz in terms of empathy. We can expect to achieve Evidenz in this sense most readily with the low level sort of empathy that corresponds to the tallying of reactions and ostensive definition, and to the kinds of following of the reasoning of others that are given in primitive form in the phenomenon of joint attention.

Psychologism and Regress

Self-evidence has had a bad press in twentieth-century philosophy as a result of Gottlob Frege's and Edmund Husserl's critiques of psychologism. Frege's argument, which is now an elementary piece of conventional philosophy, is that there is a difference between truth, such as mathematical and logical truth, and taking to be true – which is precisely what Evidenz is concerned with. The error of psychologism is one of confusing “true” with “taken to be true.” Being true is independent of being taken to be true. Thus, as Frege puts it,

This difference, in present terms, is related to the difference between what people do and what they ought to do – to normativity. Arithmetic, for him, is a normative science. What people believe about natural numbers (though what Brentano speaks about is not what people in fact think, but what “anyone whose judgments were evident would” think) has nothing to do with the truths about them. Even a community of arithmetic users could be wrong about doing sums, or do them incorrectly: nothing depends on what is in their heads or what they do. Evidenz, in contrast, does depend on what people think (or at least what people whose judgments were evident would think): it is therefore a psychological fact, not a fact about mathematical truth.

But this line of argument ends badly. Frege himself has no solution to the regress problem. As he concedes,

So in a sense, Brentano's point stands – there is no getting rid of Evidenz or something like it, because even on Frege's account we must take these laws to be evident. Moreover, contrary to his expressed aim of avoiding a shift from logical to psychological considerations, even Frege must resort to psychological language in his discussion of concepts. People must somehow relate to these concepts in order to understand and use them. So he uses a term which appears repeatedly in the analytic tradition, “grasping” (Frege [Reference Frege and Furth1893] 1964, 23). This term is a surrogate for psychological terminology, but it is loaded: it makes sense only if there is something to be grasped. But it is difficult to see what grasping consists in other than the achievement of Evidenz described by Weber and Brentano. If the only difference is that grasping implies that there is something being grasped, the usage is misleading: it pre-decides the issue of the nature of concepts and conceptual experience in favor of the idea that concepts are external and object-like.

If we return to the point where this particular intellectual short cut has not gained currency, it becomes apparent that grasping, psychological normalcy, and Evidenz are closely related ideas. This closeness is exhibited in the classic discussion of regress problems, Lewis Carroll's “What the Tortoise said to Achilles.” Achilles offers a Euclidean proof. The Tortoise, however, rejects the evident character of the final step, and insists on asking what would happen “if I failed to see its truth,” asking merely that it be written down so that it can be justified by another step, D. After noting that only obtuseness would prevent anyone from doing so, Achilles does so and says

And with this we are off to an infinite regress. This little story is usually taken to show the need for rules of inference in logic, but Carroll's language of seeing is more to the point. What is needed is something that can be shown, even, in this case, to the obtuse. But explicating showing in terms of presuppositions, in this case a presupposed rule of inference, leads merely to a regress.

We do not live in a world where people fall into regresses because they continually demand further justifications of the reasons given by others. Why? The standard answer is that there is a pregiven normative structure that people in general, or users of a common language, or some other group, share, which is the basis for their rationality and for the commonality or sharing of their rationality. The reasoning here is rarely articulated, but goes something like this:

1. 1) There are no infinite regresses of this kind in actual reasoning.

2. 2) For there to be a regress of the Tortoise type the Tortoise would have to deny rational common ground – to fail to acknowledge the reasons of Achilles, indefinitely.

3. 3) To end a regress is to arrive at and acknowledge a reason that is shared.

4. 4) Reasons are normative.

5. 5) Reasons come in connected structures of reason.

6. 6) To acknowledge a reason is thus to arrive at a common normative structure.

7. 7) The regress stops when we reach this common normative structure of reasons, concepts, categories, synthetic a priori truths, or the like.

8. 8) There is, because there must be, a common normative structure of this sort.

What is in common in these views, and common to dozens of thinkers who could be added, is the idea that there is some kind of more or less autonomous shared structure external to the person which must be grasped or internalized as a precondition of mutual understanding.

Here are a few. The early Wittgenstein, in the Tractatus, claimed that “the tacit conventions on which the understanding of everyday language depends are enormously complicated” (Wittgenstein [Reference Wittgenstein, Pears and Mcguinness1921] 1961, 37, para 4.002). The structural relation is clear in this case: there is everyday language, and there is the structure on which it rests. It is ambiguous in this passage whether these “conventions” are local, or whether the structures in question are universal. John McDowell in Mind and World criticizes “this idea: that when we work at making someone else [who is ‘initially opaque’ to us] intelligible,” we are relying on a “system of concepts within which the other person thinks” meaning a local system of concepts (McDowell Reference McDowell1994, 34–35). Instead, he thinks, we rely on a deeper set of concepts we already share with the other person, “a standpoint from which we can join her in directing a shared attention at the world, without needing to break out through a boundary that encloses the system of concepts” (ibid., 35–6). Robert Brandom, who focuses on linguistic normativity and on understanding language as a structure of inferential moves, thinks that the inferential structures are local, community relative, or as he calls them, “social,” and take the form of a system of practices (Brandom Reference Brandom1994).

The Tortoise, for McDowell and Brandom, is engaged in denying the things that would end the regress. The problem is not obtuseness, but a normative failure – a failure to acknowledge things that he must acknowledge if he is, for McDowell, a rational being, bound by the norms of reason, or for Brandom, a normatively bound member of a community. The term community is important here: to the extent that they deal with “local” normativity, these accounts rely on the notion that there is a collectivity of some sort behind every normative regress-stopper, in the form of a consensus about correctness or proprieties, for example. Even the universal, Kantian structure of normative reason seems to depend on a collectivized notion of the common possession of this structure by the community of rational beings.

Structure and Empathy

The standard answer to the regress problem with which I began this paper appeals to an autonomous structure of reasons, meanings, score-keeping systems, premises, or concepts that serve as regress-stoppers because they are the things that Achilles and the Tortoise share. Sentences, judgments, and the like are true, intelligible, or valid by virtue of the existence of a structure that underlies and warrants them: they depend on something hidden. We communicate by virtue of our shared possession of the tacit structure that makes meanings and the like the same for each of us. In the language of the neo-Kantians of Brentano's time, a language that is very much still with us, this shared hidden structure was made of “presuppositions,” which could be revealed by transcendental reasoning. The other view is that “nothing is hidden.” This was also Brentano's view: that Evidenz did not depend on any off-stage or prior structure. Things with Evidenz are regress stoppers directly, because they are universally accepted as evident, and I can know that they are, or have faith that they would be, on the basis of my empathic powers.

We can contrast what we might call “empathist” to “structuralist” accounts as follows. The empathist thinks nothing is hidden in the sense that empathy works on externals. For the “structuralist,” agreement always is a result of some hidden fact, which can perhaps be made explicit, such as Brandom's score-keeping systems and proprieties. The structuralist account depends on transcendental arguments, arguments that the structure in question is the unique condition for the possibility of that which is being explained. The empathist account depends on something more akin to inference to the best explanation, consilience with other kinds of facts, such as the fact that we can empathize, and a preference for explanations that do not invoke problematic abstract entities. Because the structuralist's arguments are transcendental arguments, based on the necessity of accepting the structuralist conditions for the possibility of a given outcome, such as communication, the existence of a genuine alternative explanation is a problem. The alternative account is not merely a rival; it is a refutation, since it shows that the structuralist conditions are not “necessary.”

Structuralist accounts do not lend themselves to the consideration of alternative explanations. As we have seen in the very brief list of examples I gave earlier, there is no agreement on what the hidden structure is, what it is supposed to explain, or what it is supposed to do, other than to serve as a regress-stopper. So the project of giving an alternative account of what it is that the structuralist purports to explain is stymied at the start. This is no accident. Structuralist arguments are typically transcendental arguments or arguments that have properties in common with transcendental arguments, and the structure of transcendental arguments is typically circular, with the circularity concealed. They depend on characterizing the subject matter to be explained in such a way that alternative explanations are ruled out not by some empirical feature of the things to be explained but by the definitions of the terms used in the preferred descriptions of the theorist. The descriptions are not the kind of descriptions familiar from science, which themselves evolve in the course of the development of experiment and theory but capture supposedly essential features of the subject matter, whether it is thought, language, concepts, or whatever, which a theory is then supposed to account for. Not surprisingly, the only theory that can account for a particular set of features described in a particular way is one which respects the ontology of the description itself. If the description is implicitly “normative,” for example, the theory will have to be as well, in order to account for it. The typical conclusion, that language, thought, concepts, and so on are “normative,” merely reaffirms the commitments implicit in the description. Accepting the principle that only something normative can account for something normative precludes any alternative explanation.

If we take these kinds of explanations as a group, however, the strengths of this kind of reasoning becomes a weakness. The arguments work too well, in the sense that they generate a myriad of hidden structures. This simply reflects the exceedingly various potential regresses to be found in ordinary reasoning: over meanings, over the justification of logical claims, over normative claims of various kinds, over things like legality, over claims about the empirical world, and so forth and so on. Each of these seems to require a different kind of structural explanation, with a different source of normativity. The alternative to appealing to structures in this ad hoc way is to give some sort of account of the relations between these structures. Brandom suggests that all normativity is reducible to “the social practice idiom” (Brandom Reference Brandom1979, 190 n.7); McDowell takes a more traditionally Kantian approach and argues for universal reason. Another approach holds that local norms are nested within more universal norms. But the fact that normativists have sought to eliminate this problem by claiming that all of them have the same source is indicative of the suspicion that there is a problem here.

Given this list of issues, and the specific character of the arguments advanced for structuralist normativist accounts, what would an alternative empathist account need to establish in order to provide enough of an alternative to improve on normativist structuralist accounts based on transcendental arguments, and thus eliminate the claim to necessity on which transcendental arguments depend? Because of the peculiarity of normativist descriptions of phenomena, the fact that they are descriptions already loaded with normative content so that only normative explanations can count as explanations, we need to step back and consider the phenomenon itself rather than the phenomenon as pre-characterized by the normativist.

In short, we need to make the question more like a question about rival scientific theories. If the empathist can establish that an account, any reasonable account, can be given of the main issues listed above, and the account can be linked reasonably closely to a body of scientific fact – something that is excluded by definition by standard normativist accounts, such as Brandom's, which insist on the principle that only the normative can explain the normative – the empathist wins, at least in the temporary, prospective, fallibilist way that scientific accounts “win.” The point applies equally to the argument that naturalistic accounts are excluded by definition – the existence of such accounts shows the definitions to be circular. The descriptions themselves become an issue: hanging on to a favored description when there is no empirical difference in the face of a better alternative account is special pleading. The same result is reached if the account given by the empathist fits with a broader range of fact than that covered by the structuralist account. As we will see, however, we can go beyond this: to use the empathist argument to account for some oddities in the alternative theories, oddities which, once explained, also discredit these alternatives.

Is Empathy too Weak a Reed?

Is empathy up to this challenge? The usual issues with empathy have to do with the epistemic flaws of empathy – the ease with which we err in our empathic projections onto others. But these issues are issues primarily with empathy in the strong sense, the higher forms of empathy. They take a different shape in the case of following the thought of another person, and also in relation to Evidenz. If we ask the question of whether our teacher means what we mean when we write “2 + 2 = 4” on the blackboard, and then ask if what is evident to us is also evident to the teacher, we do not need deep insight into the mind of the teacher. Moreover, we have a great deal of supplementary support for our empathic sense that what is evident to her is the same as what is evident to us, and that indeed she sees it as evident. We have taken the tests, been corrected, and have observed other students being corrected. We have gone through exercises with apples and oranges that help convince us that the expression is evident, and so forth. While none of these things is a criterion for evidence, and none on its own would warrant our acceptance of the expression as evident, or justify our faith that anyone else would see it as evident, there is nevertheless little room for error with respect to this empathic projection. We can safely dismiss any actual case of a person failing to see that this is evident as a case of error, misunderstanding, insanity, or something similar.

This account, then, explains what the rivals explain, and explains deviations in the same way. Indeed, it is striking that Frege indulges in psychological language, as when he says that if there were people who denied the laws of logic, we would say “we have here a hitherto unknown type of madness” (Frege [Reference Frege and Furth1893] 1964, 14); that Saul Kripke comments that – “they might lead me to go insane, even to behave according to a quus-like rule” (Kripke Reference Kripke1982, 27). What is different between the empathist and the structuralist is that the empathist does not need to rely on notions like community or other collective facts in addition to psychological facts.

The same goes for grasping a concept or a rule. The normative structuralists also have to deal with the fact that some people fail to grasp. And in accounting for this they must appeal to some set of psychological facts. The empathist must as well. But the core of the empathist's account is the thought that learning something human or intellectual will have an empathic element. It is a demystification of something that remains mysterious in structural accounts. For the empathist, learning basic arithmetic is a matter not of absorbing, through some unknown mechanism of transmittal, a system of tacit premises or presuppositions that would enable the student to take 2 + 2 = 4 as evident, but of understanding the teacher and other students and coming to see 2 + 2 = 4 in a way that would empathically universalize – such that anyone else who could take it as evident would take it as evident. This is an awkward formulation, but the thought is simple. One aspect of learning is figuring out what the teacher wants, which is a problem of empathy. The empathist's point is that this, together with the other kinds of input the student gets, both empathic (for example, through seeing other students learn) and empirical (by counting apples and oranges, for example), is sufficient to account for the evident character of the equation.

How does this relate to the problem of the collective character of structure? For the empathist, nothing need be hidden. The empathic projections in question in this and other cases works on the surface – on what the teacher says and does, for example. What is added to the surface is added by empathic projection. Nothing more is necessary. The sense of Evidenz – the faith in the universalizability of the evident – may, in specific cases, as we will see shortly in connection with such things as cultural differences or past historical agents, require supplementation. But in the ordinary case there is nothing other than our empathy by virtue of which this projection is possible. This is simply to say that empathy is a plausible candidate for the task of providing what needs to be added to feedback to account for the patterns of actual conduct that the structure hypothesis purports to account for, namely regress stopping.

The next question is whether empathy is enough of an answer, or the right kind of answer. Structure accounts claim to produce objectivity: the structure is, by definition, the same for everyone. Empathy is not objective, but intrinsically subjective. The traditional problems with empathy involve reliability: we often err in our empathic projections, though primarily with empathy in the strong sense. So empathy alone is not an objective regress stopper. But this is only part of the story. As we have noted, much of the time our empathic errors are corrected by feedback: if we think we understand the arithmetic teacher, but do not, we are liable to be unable to account for what needs to be accounted for. The question is whether empathy, or low level empathy, plus feedback, is enough, or to put it differently, whether there is anything more to be had. In the case of “+ 2” it is doubtful that anything more is needed. To the extent that there are actual deviants who are unable to match their use of “+ 2” to that of their teachers and fellow students it is likely that there is a more serious and pervasive problem of understanding other people or performing mathematical tasks. Intersubjectivity of the sort produced by mutual comprehension based on empathy plus feedback produces as much “objectivity” in this case as there is to be had, a point to which we will return. And universal intersubjective agreement – literal rather than analogical agreement behind the surface agreement – is as much as there is to objectivity. But more than one intelligible inference may fit the facts, and the facts may underdetermine our judgments, and allow for more than one intelligible alternative in a given situation.

If we are considering these accounts as rival theories, as we would rival scientific theories, limiting the issues to the role of empathic universalization in these philosophically central cases is misleading. Empathic universalization is not restricted to the evident, the obvious, or the truths of logic, and the implications of the notion go beyond these issues. Empathy may, in fact, account for a great deal of actual conduct, social interaction, and so forth. This was a point made by Weber and later by Alfred Schutz. As with Evidenz itself, empathy is impossible to eliminate as a part of social interaction and the explanation of social interaction. This becomes clear when we consider failures of empathy, cases where our capacity to understand is challenged, or in which we cannot initially understand, empathically or rationally, what the behavior in question means. If we are to explain a Hindu ritual of purity, for example, which initially seems strange, even unintelligible to us, such as the treatment of untouchables, we attempt to give an account of it which in some sense normalizes it, and normalizing it amounts to making it intelligible. But making it intelligible in this context amounts to bringing to bear facts which allow us to describe it in a way that allows us to empathize. The idea of defilement by touch is already intelligible to us as a part of our own thinking. We know, and find intelligible, the kind of thinking that goes into the notion of contagion by touch already – from such diverse sources as the laying on of hands of the apostolic succession to the child's game of tag. So our problem of making the Hindu practice intelligible in the end relies on us filling in the rest of the context in such a way that we can match it up with something intelligible to us.

Intelligibility here bottoms out not in a theory of rationality, but in an actual point of empathic contact. Contagion thinking seems to be universal, and thus universally intelligible. But contagion inferences are not “necessary”: we can follow them without accepting them. The inferential part of this thinking is ungroundable, final, and so forth. Perhaps it is rooted in the primal mirrorable experience of the non-mechanical effects of touch, which provides a template for the notion that touching produces non-mechanical transformation. But there is no reason for us to think that this kind of inference will always be correct as a matter of empirical fact. And indeed more generally, empirical fact falls into a different category from following the thought of another, shared attention, and so forth. The traditional Cartesian issues of illusion, for example, cannot be resolved by appeals to evidence, for by definition another person, with knowledge of the illusory effect, would not judge the illusory effect as evident. This is why Brentano distinguished “evident to me” from Evidenz as such, which he associated with “insight.” As with Weber's discussion of indirect understanding of action, based on inferences, we need to confirm contagion inferences, if they are inferences about actual causal relations, for example about disease. Whether they predict or not is a question separate from the question of whether we can follow the thinking. Following is empathic, final in the sense that there is no further explanation of it. Contagion reasoning itself is a matter of Evidenz, in the sense that there is no additional justification of it to be had. But of course it is not evident in the sense that it is necessary or necessarily “true.”

When Empathy Goes Wrong

The obvious objection to appeals to empathy of any kind is this: empathy is not a reliable guide to anything; the reasoning is this: our empathic attributions often go wrong. Empathy produces results with Evidenz only in simple cases, such as Weber's example of a man chopping wood (Weber [1968] 1978, 8). And even these empathic imputations, to the extent that they rest on empirical facts, are fallible. So the assumption underlying empathic universalization, on which this whole account is based, namely that empathy is an autonomous source of knowledge, is simply false. In what I have said here I have tried to separate empathy and empathic universalization from truth and objectivity by arguing that while empathic universalization leading to intersubjective agreement is enough for truth and objectivity, empathic universalization is a broader phenomenon, and it applies to anything that another person can “follow” in the thinking of another, or, in this sense, understand. This formulation has the effect of inverting the relationship between Evidenz and truth that informed the earlier discussion: the evident is not a problematic subtype of truth, as it is in the earlier picture. It is a feature of mutual reasoning, which may or may not lead to truth (whatever that might mean in the contexts of logic and mathematics that concerned Brentano), that better matches the idea of intelligibility. The true is a subtype of the intelligible, and the intelligible is that which can be empathically understood. Evidenz is a property of the subtype of the intelligible for which there is no alternative which one could accept. This leads to a different picture of empathic universalization – not as a criterion for the acceptability of “certain” or evident statements, but as a commonplace, and essential part of ordinary human interaction.

The picture we get out of these considerations is this: potential regresses are everywhere, and they end. But they do not end in analogical facts, such as agreements or assumptions. Nothing is hidden. They end at the point that people – or Tortoises – find that they can follow each other and that the step in their reasoning is evident. Evidenz, that is, ending a regress at a point that the parties can each follow, is indispensable. Seeming to be evident and being evident are, in this setting, the same thing. Being evident means being evident to everyone – subject to the usual limitations of “everyone” being sane, capable of recognizing what others would take to be evident, and so forth. But being evident in this sense is not necessarily permanent and thus is not a guarantee of truth in the sense of “there are no alternatives.” Alternatives which we did not previously imagine can be invented, and consequently what seems evident can cease to be evident, as Euclidean space ceased to be evident, in the sense that it ceased to be without alternatives. But this does not mean that it ceased to be immediately intelligible, or that we no longer understand Euclidean geometry. Nor would the invention of an alternative form of counting mean that we no longer understood 2 + 2 = 4. The only evidence we have of successful empathy is indirect – through the success of our interactions. But it is not reducible to these interactions, or to any empirical evidence. As Weber says, we “understand what a person is doing . . . on the basis of the facts of the situation, as experience has accustomed us to interpret them” (ibid., 5). Our experiences, and thus our capacity to empathize, vary individually. And what and how we experience is itself partly the product of this variable capacity to interpret. The fact that we can empathize is not derived from anything and needs no justification. It is a natural fact. To deny it is to deny one of the causal conditions of learning such things as language, communicating, and much more.

Is this picture better than the normative structuralist's picture? It certainly explains more phenomena. But it also sheds light on a peculiarity of the normative structuralist's case. The treatment of Kripke's discussion of rule-following relied heavily on the claim that there was a distinction between community acceptance of what counted as following a rule in the community and the correct application of the community's rule. There is a parallel issue in Husserl. In what is usually taken as his attack on Brentano, he discusses insight with evidence, and argues that there is a need for a modal category of the concept of insight with evidence, namely, a concept of possible insights with evidence, which in turn must be grounded in “rational consciousness in general” (Husserl [Reference Husserl and Gibson1931] 1962, 354). The oddity of both of these cases is that there are no examples of a community which is in complete error about the correct application of its rules or of a possible insight with evidence that is not actualized. The reason, in the latter case, is that once one gave such an example, and explained it so someone would accept it as evident, it would already be an actual insight with evidence. The very process of making it evident to others, in short, would make it evident in the sense of actual empathy. Something parallel happens with the case of community acceptance. Imagine that there was a community which had been taught a rule of arithmetic, such as the +2 rule, incorrectly, and that this had been learned in the ordinary way, that is to say that the thoughts of the teacher had been followed and it was understood empathically what she meant by “+2,” and that this understanding was generally accepted in the community. In what sense would this be a case of community error? What would the error be? If they understood the teacher, the teacher's usage was intelligible, and they did as the teacher did, they would simply be following a different rule which they call “+2.” This would be something like the case of non-Euclidean geometry, which makes the axioms of Euclid understandable but no longer evident. Any case of actual error, in contrast, would be a case in which the rule had somehow been misunderstood at the beginning. But if the beginning is with a teacher and the teacher's thought had been correctly followed, there is no error. And there is no way to get error by the entire community if it involves teaching together with empathic following of the thought of those who are teaching. The case for normative structuralism is that there is something left over to be explained. But there is nothing.