Moral Concepts and Emotions

MORAL CONCEPTS AND EMOTIONS Laura Schroeter and François Schroeter What roles do emotions and reason play in moral judgment? This question was at the centre of the disputes between moral rationalists and moral sentimentalists three centuries ago and it still divides contemporary philosophers. While some believe that ‘cool’ reflection is the key to morality, others claim that ‘hot’ emotions are the essential constituents of moral judgment. Over the past decade or so, however, sentimentalists have been on the offensive. Recent empirical results show – among other things – that hot emotional centers are activated during moral judgment and that emotional deficits are often associated with significant impairments in moral understanding. Many philosophers and psychologists have taken this type of results to provide important support for a broadly sentimentalist position. See for instance Blair 1995Haidt 2001Nichols 2004Prinz 2007. Prinz (chap 3.1.3) suggests that empirical data are best accounted for by a radical sentimentalist position similar to Hume’s. In his view, to believe (in a non-deferential way) that an action is morally wrong is just to have a sentiment of disapprobation toward it, and moral concepts are simply constituted by emotions (chap. 3.1.3). For a critique of the current resurgence of sentimentalism (and of the interpretation of the empirical data which have triggered this resurgence) see Joyce 2008. Our sympathies are with the opposing rationalist or intellectualist camp. In particular, we believe that reflection – the articulation of principles or rationales for our immediate responses in the process of reflective equilibrium – provides our best method of epistemic access to moral truths. But our aim here is not to articulate or defend a full-fledged rationalist position. Rather, in order to counter the new wave of sentimentalism, we’d like to take a closer look at the distinctive constituents of moral judgments, moral concepts. We’ll argue that the emotions play less central a role than sentimentalists typically assume in a characterization of our moral concepts. We start by sketching our general approach to concepts – the jazz model of meaning. We then focus on the two questions about concepts that have been of most interest to philosophers: the question of what it takes to grasp them, and the question of the determination of their reference. Do emotions figure prominently in an account of what it takes to grasp moral concepts? Do they provide a central contribution to the determination of the reference of these concepts? Focusing on these questions, we believe, can help clarify important aspects of the role of ‘hot’ emotions and ‘cool’ reflection in moral judgment. Of course, this specific focus will leave out many of the themes which have featured centrally in the traditional debates about moral rationalism. ‘Moral rationalism’ is an umbrella term covering a number of distinct theoretical theses clustered together by historical accident – they can all be associated with an author like Kant. The cluster includes (1) the psychological thesis that the faculty of reason is the source of moral judgment, (2) the metaphysical thesis that morality is constituted by the deliverances of reason alone, (3) the epistemological thesis that moral truths can be known purely a priori, and (4) the substantive thesis that moral requirements are requirements of rationality (binding all rational agents). We will make no attempt to defend this cluster here. In fact we don’t believe all its component theses are true: in particular, we reject (3) and we are uncomfortable with the faculty psychology involved in the formulation of (1). 1. The jazz model Traditional approaches to meaning favor a “resemblance” account of what it takes for different speakers to be competent with the public meaning of a term and thus to grasp the relevant concept. According to these approaches, every competent speaker must grasp the same criterion for applying a term – or, more generally, the same pattern of assumptions about its reference. For instance, a proponent of a “resemblance” account may suggest that all competent speakers implicitly assume that water is the clear, potable fluid that actually flows in rivers, fills lakes, comes down as rain, etc. Any speaker who didn’t implicitly accept this pattern of assumptions would then fail to be competent with the meaning of ‘water’ and to grasp the relevant concept. Influential contemporary champions of “resemblance” models of meaning include Christopher Peacocke 1992 and Frank Jackson 1998. We favor a radically different approach to meaning and concept possession. Instead of looking for some invariance in the cognitive role different speakers associate with a term, we think we should adopt a “connectedness’ account of competence. Two speakers share the same meaning if their uses of a word are connected in the right way – the way that makes them part of a joint linguistic practice. For more detailed expositions of our “connectedness” model of meaning, see Laura Schroeter and François Schroeter 2009Ms. An analogy may be helpful here. In debates about personal identity, some theorists hold that two person-stages are stages of the same person just in case they resemble each other in some core respect – by sharing a particular core set of psychological states, for instance, or by sharing the same underlying substrate such as a biological brain or an immaterial soul. Other theorists take personal identity to be determined by relations that connect different person-stages – for instance, a connectedness theorist might hold that two person-stages are stages of the same person just in case the later one “quasi-remembers” the earlier one, or just in case the two stages are causally connected in the right way. In advocating a version of the connectedness approach to personal identity, Derek Parfit 1984 has argued that the psychological connections between person-stages are what ultimately matters in our commonsense judgments about personal identity. In a similar spirit, we suggest that what ultimately matters in our commonsense judgments about sameness of meaning is whether two uses of a term are connected up in the right way – the way that integrates them into a communal linguistic practice. The key advantage of connectedness models of personal identity over resemblance models is that they don’t posit personal essences: being a particular person is not the same as having a particular set of properties. Similarly, connectedness accounts of meaning avoid any commitment to analytic truths: being competent with a concept is not the same as accepting a particular set of attitudes or assumptions. Given the difficulty of coming up with putative analytic truths, this is an important advantage. What sorts of connections are required for two speakers to share the same meaning and for their uses of a term to be part of a joint linguistic practice? A second analogy provides a nice illustration of our answer. To achieve a coordinated musical performance, the members of a classical string quartet must first decide upon a common score that is to guide each member’s performance. Their joint performance depends on this resemblance among the players. In contrast, the members of a jazz quartet need no fixed template to guide their performance. Instead, each member may be committed to building on whatever musical themes other members of the group try out, seeking a continuation that makes musical sense of the whole performance so far. Each jazz player trusts that the others will try to take everyone’s contributions on board and incorporate them into a coherent musical structure. The jazz musicians’ coordinating intentions are what keeps the joint musical performance on track, developing interesting themes rather than degenerating into a welter of random notes. We suggest that a similar structure of coordinating intentions can explain what it takes for speakers to be competent with the public meaning of a term and thus to share the same concept. Like a jazz player, each English speaker has an implicit intention to use terms like ‘water’ or ‘morally right’ in a way that is in sync with the linguistic practice of the group as a whole. And it is because of these interlocking coordinating intentions that there can be such a thing as a joint linguistic practice in the first place. The idea that coordinating intentions are a crucial component of competence is less problematic than it may seem at first. It should be uncontroversial that competent speakers intend to use words in a way that coordinates with others in their community – we do not take ourselves to be linguistic islands. Since English speakers take themselves to be speaking a shared language, they assume the standard for the correct use of words does not depend solely on their own personal linguistic practice – the standards of correct use depend on the linguistic practice of English speakers in general. According to the jazz model, different speakers can hold their use accountable to a common standard, without settling in advance on the precise content of that standard. The idea that our coordinating intentions help forge a common linguistic practice also has a great deal of plausibility. Linguistic practices are not independent entities which would survive no matter what our linguistic behaviors and intentions were. They are partly constituted by speakers’ intentions to use their words in a way that coheres with that of others who have a similar intention. A speaker with no such intention will go off on a tangent and a community of such speakers will fail to coordinate their use of words in a joint practice. According to the jazz model, then, having a coordinating intention is the core requirement for competence with the standard English meaning of natural kind terms like ‘water’ and evaluative terms like ‘morally right’. However, a coordinating intention by itself is not sufficient for linguistic competence. A monolingual Frenchman will not acquire competence with the English word ‘right’ by simply intending to coordinate with English speakers’ practice with the term. In order to count as competent, a speaker must have enough substantive understanding of what it is the term represents to use it responsibly in his own reasoning. Moreover, this substantive understanding must be sufficiently congruent with other speakers’ if it’s to warrant a common interpretation. Clearly, Mrs. Malaprop doesn’t count as competent with the public meaning of ‘meretricious’ if her substantive understanding of the term resembles the understanding most speakers associate with the term ‘meritorious’. On our model, the speakers of a community must satisfy two conditions to count as competent with the public meaning of an evaluative term: Coordinating intentions: The individual speaker must have a coordinating intention to use the term in a way that makes best sense of the communal practice; Rough Congruence: The individual speaker’s initial understanding of the term must not diverge so radically from that of others in the community as to undermine that coordinating intention. Notice that whereas “resemblance” approaches to competence require that there be a specific pattern of substantive assumptions that every speaker must accept about the reference of a term, our “connectedness” approach involves no such requirement. Thus, our model has the flexibility to accommodate the very wide range of evaluative disagreements we tolerate among competent speakers. As many ethicists have noted, morality is a very contested topic. According to the jazz model, different speakers can share the same concept and pick out the same property even though their substantive criteria for applying ‘morally right’ diverge in many respects. All they need is to have the relevant coordinating intention and sufficient overlap in their initial understanding. This brings us to our second question, the question of reference determination. We have sketched our approach to conceptual competence; it remains to be seen how a common reference can be assigned to all competent speakers’ uses of a word. We embrace a broadly interpretivist account of reference determination: what your words and the concepts they express refer to is determined by the best rationalizing interpretation of your representational practice. For well-known interpretivist approaches to reference determination, see Davidson 19801984Dennett 1987Lewis 1974. So, for instance, your word ‘water’ will refer to H2O if that interpretation vindicates the most important aspects of your total practice with the term – where this practice includes your various criteria for identifying instances, your epistemic hunches about how to correct those criteria, your various practical and theoretical interests in classifying things as ‘water’, and so on. The correct assignment of reference is the one that makes best sense of the totality of your interest and practice with the term. It is important to note that our interpretivist approach to reference-determination is non-reductive. The account appeals to commonsense methods of rationalizing interpretation – invoking our commonsense reflective judgments about what it takes to “vindicate” what’s most important in a practice or to “make best sense” of that practice. We do not seek to provide an algorithm for these commonsense methods of interpretation. Nor do we seek to reduce the reference relation to some naturalistically respectable causal or nomic relation. On our view, a naturalistic account of reference is correct only insofar as it accords with our best commonsense interpretive judgments. We doubt that any naturalistic account will be correct across the board, capturing the reference of names, natural kind terms, artifact terms, mathematical terms, moral terms and so on. We’ll come back to the question of reference determination below in section 3 in our discussion of the reference of moral terms. For now, we’d like to emphasize two important characteristics of our account. First, rationalizing interpretation is a holistic and non-foundationalist method. As a consequence, there is no pattern of assumptions the speaker may have about the reference that is guaranteed to be correct. Any particular assumption may be overturned in the light of the subject’s whole historical practice with the term. It’s also worth emphasizing that rationalizing interpretation takes into account facts about the subject’s social, physical, and historical environment. These factors compound the faillibility of the subject’s current assumptions about the reference. Many people, for instance, take homosexuality to be a central and obvious case of immorality. However, if the ban on homosexuality fails to cohere with what’s most important in our practice with moral terms, the assumption that homosexuality is a central case of immorality – however intuitively obvious may seem to the subject – will be overturned in an ascription of reference to moral terms. This coherentist approach to reference determination contrasts strongly with the strong foundationalism implicit in neo-descriptivist accounts like those of Jackson (1998) and Peacocke (1992). Second, the fact that different speakers are connected by reciprocal coordinating intentions can help ensure that their use of a term will be co-referential. Because of their coordinating intention, individual speakers are committed to using their terms in a way that makes best sense of the linguistic practice of the group as a whole – not just in a way that makes best sense of their own idiosyncratic practice with these terms. This means that the input into the interpretation of a word like ‘water’ will be the same for all competent members of the group: it will be the practice of the group as a whole with that word which is relevant to determining what individual speakers refer to. Thus, when speakers are connected by coordinating intentions, rationalizing interpretation will assign the same reference to their use of a term, even if they diverge in their substantive assumptions about the reference of that term. 2. Competence With a rough sketch of our general approach to concepts in hand, we can now turn to our first question. Do emotions figure prominently in an account of what it takes to grasp moral concepts? Let’s start with a few basic points which should be largely uncontroversial. It should be clear that the emotions play a crucial role in a genealogical account of many of our most important evaluative concepts, including our moral concepts. Fear, for instance, is a mechanism for detecting potential harms, which we share with both higher and lower animals. Obviously, our concept of danger has its evolutionary roots in the fear mechanism. Animal fear is a rough and ready way to detect certain types of perceptually salient and evolutionarily significant risks in the environment. Our conceptually articulated thoughts about danger allow us to perform a similar task of risk assessment in a much more systematic and sophisticated way. Moral concepts are also rooted in our emotions. As many authors have emphasized, for instance, our thinking about what’s morally wrong seems to have its evolutionary roots in emotions such as anger and resentment, guilt, indignation or disapproval. See Gibbard 1990, chap. 7 and 14 for the locus classicus. It is also safe to assume that emotions play a crucial role in the normal process of acquisition of many of our central evaluative and moral concepts. A child will typically get an initial grip on the meaning of the word ‘danger’ by hooking this word up with the deliverances of his fear mechanism. Similar conclusions seem to hold in the case of moral words – and the concepts they express – and emotions like anger, resentment or guilt. It would be wrong, moreover, to think that the emotions are simply a ladder that facilitates the emergence and acquisition of evaluative concepts and can be safely dispensed with after that. Emotions typically continue to play an important and epistemically valuable role in the deployment of evaluative and moral concepts in competent adults. Even for reflective and theoretically sophisticated subjects, for instance, fear often provides a useful heuristic for judging whether a situation is dangerous. Reflection, after all, is typically time-consuming and effortful. In a lot of situations, it makes perfect sense to trust our fear mechanism and to take our fear responses to constitute strong prima facie evidence that a danger is looming. Similarly, it makes sense to take your pangs of guilt evidence that you did the wrong thing when you divulged a confidential report. This is not to say, of course, that virtuous epistemic agents should always uncritically leap from a fear or guilt response to a judgment about danger and moral wrongness. Often enough, quick and dirty heuristics lead to the wrong conclusions. A virtuous epistemic agent will know when reflection is called for to check more carefully whether a situation triggering an emotion of fear or guilt is really dangerous or morally wrong. Granting all these points, however, does not support the conclusion that emotions – like fear or guilt – are necessary for grasping central evaluative concepts such as the concepts of danger or moral rightness. According to the jazz model, speakers must have coordinating intentions if they are to share the same concepts. But there will be few other strict specific requirements for competence with particular concepts. As long as her understanding of a particular topic has enough overlap with that of others in her linguistic community, a speaker who has a coordinating intention will count as competent with the relevant concept. Imagine, for instance, a speaker who does not feel the emotion of fear, or even a Martian who is not equipped with any emotional mechanism at all. According to our model, this subject could still be competent with the concept of danger if he were appropriately hooked up with our linguistic practice, intended to coordinate his use of the term ‘danger’ with ours, and shared enough of our more reflective conception of what it takes to be liable to cause harm and thus dangerous. The fact that our normal ways of thinking about danger are typically infused with emotions makes no difference here. A subject can still be competent with the concept of danger if he does not feel fear, or see any important connection between danger and fear. (Here again, the point generalizes to the case of the concept of moral wrongness and the emotions typically associated with it). We think this result is intuitively plausible. Our standards of linguistic competence are permissive: we are ready to tolerate important divergences in competent speakers’ assumptions about what they are talking about. As long as we can profitably engage in discussion about what’s dangerous with a very sophisticated but emotionless Martian, there would be little point in denying that he is talking about the same subject matter as us and sharing our concept of danger. The jazz model also casts doubt on the claim that experiencing the relevant emotions is sufficient for grasping evaluative concepts like danger or morally wrong. According to radical sentimentalists like Jesse Prinz, many of our evaluative and moral concepts are simply constituted by emotional dispositions. In Prinz’s view, to possess the concept of moral wrongness is just to have a disposition to feel disapprobation when presented with particular actions. A moral judgment that an action is wrong is an occurrent emotion of disapprobation toward that action (2007, chap. 3.1.1 and 3.1.3). According to the jazz model, it makes no sense to simply reduce moral concepts and judgments to emotional responses. To begin with, possession of moral concepts requires an intention to coordinate one’s use of moral words with that of others. Like in the case of any other concept expressed by a public language term, a coordinating intention is crucial to ensure that all competent speakers pick out the very same property with their uses of public language terms and with the corresponding concepts. In addition, competent speakers must understand well enough what they are talking about when they use a particular term, and this understanding must overlap sufficiently with that of others in their community. The jazz model emphasizes that very few – if any – specific assumptions about the reference of a term will be strictly required for competence. For any particular assumption about the reference, however important it may be, we can usually imagine a speaker who rejects that assumption and manages to secure competence. One deviant assumption is not usually enough to undermine linguistic competence with the public meaning of a word – provided that the rest of the speaker’s understanding of the term overlaps sufficiently with that of others with whom she intends to coordinate. By the same token, however, sharing one normal assumption will not normally suffice to secure linguistic coordination with one’s community. Competence with the meaning of our term ‘morally wrong’, for instance, cannot be secured simply by using the term to express a certain stereotypical emotional response of outrage or disapprobation. No matter how much you might want to coordinate, you won’t count as competent with the term ‘morally wrong’ if you take it to function like ‘ouch’ or ‘yucky’. At a minimum, you must take there to be a property picked out by the term, and your understanding of this property must overlap sufficiently with that of other English speakers. For instance, competent speakers normally distinguish between moral wrongs and mere violations of conventions, they have opinions about different types of moral wrongs and their relative importance, they recognize better and worse moral arguments, they implicitly take moral wrongness to supervene on non-moral properties, and so on. Moreover, we expect competent speakers to distinguish between their own emotional responses and moral wrongness: the mere fact that one feels anger or disapprobation toward a particular action does not eo ipso mean that the action is morally wrong. Thus, someone who used the term ‘morally wrong’ simply to express her emotion of outrage or disapprobation would be so radically out of step from the rest of the English speaking community as to preclude profitable discussion. We conclude that much more than having an emotional disposition is required for competence with our shared moral concepts. 3. Reference determination This section will address the most important aspect of concepts, the determination of reference. What role, if any, do the emotions play in the determination of the reference or our moral concepts? One of us has argued that in the case of central evaluative and moral concepts, standard sentimentalist approaches to the determination of reference fail . Sentimentalists claim that emotions typically play a central role in the determination of the reference of our evaluative and moral concepts. According to the most sophisticated sentimentalist proposals, what ultimately determines whether x is dangerous, for instance, is whether fear would be justified – warranted, or appropriate – toward x. More generally, sentimentalists suggest that for many important evaluative concepts our ultimate criterion for determining whether an object has an evaluative property is whether that object warrants a particular emotional response. We think that the sentimentalist attempt to reduce the question of reference to the question of the appropriateness of particular emotions fails for central evaluative and moral concepts like dangerous or morally wrong. In a nutshell, the assessment of our emotions is not what we are ultimately most interested in in our evaluative and moral judgments. Our ultimate criterion for determining whether a situation is dangerous, for instance, is whether it is liable to cause harm, not whether fear would be appropriate toward it. Even though fear still provides us with quick and dirty heuristics for evaluating dangers, even though fear is the evolutionary ancestor of our reflective thinking about danger, what we are ultimately interested in in our danger evaluations is not whether we should feel fear. The assessment of our emotion of fear is a specialized and marginal task which is ill-suited to capturing the ultimate criteria for determining the reference of an important evaluative concept like danger. The main point of the sophisticated reflective methods – assessing what causes harm and the likelihood of its occurrence – we rely on in our most careful judgments about danger is simply not to tell us what emotions are appropriate in a particular situation. We think that similar conclusions apply to sentimentalist attempts to reduce the reference of our concept of moral wrongness to the question of the appropriateness of particular emotions like guilt, resentment, or disapprobation. To prevent misunderstandings, it is important to emphasize that in our view the sentimentalist approach to the determination of reference succeeds for some minor evaluative concepts. We are not denying that our ultimate criterion for determining whether an object is fearsome of shameful, for instance, is whether fear or shame is justified toward it. Our point is that traditional sentimentalism fails in the case of our most important evaluative concepts. For more details on this point, see (Schroeter F. 2006). We don’t want to further elaborate on these conclusions – for which we have argued in detail elsewhere – and we’ll simply take them for granted in what follows. What we would like to suggest is that even if the standard sentimentalist attempts to reduce the reference of evaluative and moral concepts to a simple function of the appropriateness of particular emotions fails, there is room for a less central role of the emotions in the determination of the reference of our most important evaluative and moral concepts. We think our interpretivist approach to reference determination can help make this weaker sentimentalist case. According to our holistic model of rationalizing interpretation, the correct assignment of reference must vindicate what’s most important in the group’s total linguistic practice with a term. In assigning a reference to the term ‘morally wrong’, for instance, the interpreter will take into account the whole range of attitudes and cognitive or motivational dispositions different competent speakers associate with that term in the community. Clearly, the interpreter will take into account the multiple connections between our moral judgments and the emotions: for instance, the fact that strong emotions influence our moral judgments, that guilt provides prima facie evidence that one has done something wrong, that guilt, resentment, and indignation are the evolutionary ancestors of our more sophisticated thinking about moral matters, etc. She will also take as input into interpretation other cognitive and motivational aspects of the linguistic practice associated with our moral terms: the catalogue of rules of thumb competent subjects rely on in their classification of actions as morally wrong, their potentially idiosyncratic hunches about the relative importance of different moral violations (in case of a conflict between keeping a promise and helping a friend, when should we follow one requirement rather than the other?), their implicit assumptions about what’s a good moral argument or a good procedure to correct one’s moral judgments, their commitments to the authority of moral considerations, the motivational role these considerations play in ordinary subjects, etc. The property assigned as what the term ‘morally wrong’ picks out will be the property which makes best sense of this total mix of assumptions and cognitive / motivational dispositions. Within this interpretivist framework, the criticism we have formulated at the beginning of this section toward standard sentimentalist proposals can be reformulated as follows: the task of assessing particular emotions (guilt, resentment, indignation, or disapproval) is not what’s most important in the total linguistic practice we associate with moral terms. The modest concession we now would like to make to sentimentalists is that the emotions nevertheless play a sufficient role in the practice associated with moral terms to “leave their mark” on the property which makes best sense of the linguistic practice associated with those terms. Even though the reference of moral terms cannot be reduced to what makes particular emotions warranted, we need to mention the emotions in a full characterization of the property picked out by these terms. To make our point, it may be helpful to look at the simpler example of the evaluative term ‘dangerous’. Clearly not everything that involves some tendency to cause some harm counts as dangerous: both the tendency and the harm must be significant enough for a situation to qualify as dangerous. Now, how is the threshold settled which determines whether a situation is likely enough to cause enough harm in order to count as dangerous? A full answer to this question may be quite complex. Perhaps some subtle contextual factors are involved in fixing the cut-off point for what’s dangerous, so that the same situation can be slightly above the cut-off point in one conversational context and slightly below it in another. For ease of presentation, we will ignore such complications here. It seems unlikely that any objective joints of nature can provide us with a compelling cut-off point to distinguish the dangerous from the merely annoying. On the face of it, rankings of greater and lesser risk and rankings of greater and lesser harm seem to form a smooth curve, with no particular point on that curve emerging as a non-arbitrary candidate for a threshold. It is much more plausible to suppose that contingent facts about our human psychology, and in particular about our fear mechanism, help determine the cut-off point for the dangerous. Our fear mechanism itself privileges a particular threshold on the risk/harm curve: fear is only triggered in situations that involve a certain degree of risk and a certain degree of harm. This threshold has the advantage of being relatively stable across humans, creating a species-wide salience space for assessing certain types of risks. In the absence of a compelling objective threshold for risk assessment, this sort of shared human sensibility could play an important role in helping us converge on a conventional cut-off point for categorizing situations as dangerous. To be dangerous, then, would be to pose roughly the level of risk that could normally engage our fear response in paradigm cases. This is not to say, of course, that fear is our ultimate criterion for determining whether a particular situation is dangerous. The basic subject matter picked out by our term ‘dangerous’, we suggest, cannot be characterized simply in terms of our fear response. Roughly, for something to be dangerous is just for it to be likely to harm some of a subject’s basic capabilities. We have sophisticated reflective methods for assessing whether a situation is genuinely likely to cause harm, and these methods – not our fear responses – are our ultimate court of appeal when it comes to discriminating between real and apparent dangers. See (F. Schroeter 2006) for an extended defense of this point. Still, our concept of the dangerous is categorical, not comparative. So even if our reflective methods provide us – without the help of fear – with a complete ranking of possible risks and possible harms, we still need to set a cut-off point on the risk/harm curve in order to determine the extension and thus the reference of the concept dangerous. Our suggestion is that fear may play an important role in settling that cut-off point. We believe that a similar point can be made about our moral concepts. There is a point at which minor offenses become too trivial to qualify as morally wrong and should be considered morally indifferent instead. Whatever one’s theoretical commitments about the precise nature of moral properties – whether one is a consequentialist, a deontologist or a virtue theorist – it seems unlikely that the underlying structure of the moral subject matter will provide a salient cut-off point for distinguishing genuinely moral wrongs from trivial offenses. The gravity of offenses seems to involve a smooth curve with no non-arbitrary cut-off point separating the serious from the trivial. In the absence of any clear joint of nature, contingent facts about our psychology may help to fix a threshold for our moral categories. It is only when an offense reaches a certain level of gravity that it will trigger emotions like resentment, indignation, or guilt. Given how important these emotions have been and still are in shaping our moral practices, it seems plausible that the thresholds relevant to triggering these emotions may play a role in settling the cut-off points for counting as falling into the extension of our moral concepts– just as we suggested they do for the dangerous. In this way, our emotional sensibilities may provide the scaffolding for a common cut-off point for our moral categories. If emotions do play this role – and we suspect they do – then we must reject the radical rationalist position that moral concepts and moral judgments can be fully explained without ever mentioning the emotions. In helping determine thresholds for category membership, the emotions associated with moral judgments help determine the precise extension of our moral concepts. It follows that just which properties our moral concepts pick out depends in part on our shared emotional sensibilities. In this somewhat indirect way, we suggest, emotions may “leave their trace” on the properties assigned as the reference of moral concepts though rationalizing interpretation. 4. Why we are not sentimentalists Rejecting radical rationalism about moral concepts, however, does not put us in the sentimentalist camp. There are three types of claim the sentimentalist can make when she says that emotions are crucial to characterizing moral concepts: she could be making a claim about the reference of moral concepts, about the reference-fixing conditions associated with these concepts, or about their competence conditions. We’d like to show that, as proponents of the jazz model of meaning, the modest concession we have made to sentimentalists at the referential or metaphysical level does not commit us to anything the sentimentalist would regard as characteristic of their position. Consider reference first. Sentimentalists typically claim that emotions figure centrally in the best metaphysical characterization of the nature of moral properties. For instance, a simple sentimentalist might claim that for something to be morally wrong is for it to be disposed to trigger an emotional response of, say, moral outrage in normal observers in normal conditions. And a more sophisticated sentimentalist might claim that for something to be morally wrong is for such an emotional response to be rational (or fitting, or apt). We reject such sentimentalist analyses of moral properties. We believe the fundamental nature of moral properties is explained independently of the emotions, through a substantive moral theory. For instance, if rule utilitarianism turns out to be the best normative theory, then that theory also provides the best metaphysical characterization of the nature of moral properties. However, such theories do not settle just how serious an offense must be in order to count as morally wrong. We have suggested that emotions may help determine a standard threshold between the morally significant and the trivial. Although this ancillary role is important to settling conventional boundaries of the category picked out, it does not reflect the underlying metaphysical character of moral properties. While emotions do “leave a trace” in delimiting moral properties, they do not figure in an analysis of the metaphysical nature of those properties. The second role sentimentalists could assign to the emotions is semantic: emotions are crucial to fixing the reference of moral concepts. In principle, a sentimentalist analysis could play this reference-fixing role even if that analysis did not capture the essential nature of the moral property picked out. You can refer to skanky beer by relying on the reference-fixing criterion, “whatever chemical kind induces this yucky taste”, without supposing that the yucky taste will figure in your ultimate metaphysical characterization of the stuff in question. Similarly, we might fix the reference of the concept ‘morally wrong’ as whatever theoretically coherent property makes moral outrage appropriate, without thinking that moral outrage will figure in our ultimate account of the nature of that property. So are we agreeing with the sentimentalist claim that emotions play an essential role in fixing the reference of moral terms? We are. The problem, however, is that the significance of this claim is dramatically deflated on the jazz model of meaning. For the reference-fixing claim to be an interesting claim, one must accept a “resemblance” model of meaning. According to the “resemblance” model, (i) there is a core set of assumptions, motivations or emotions which all competent speakers associate with a term; and (ii) the correct assignment of reference is fixed by that subset of the subject’s current way of understanding the term. On this model, not all aspects of the subject’s understanding of a term are relevant to fixing its reference, for that would make it impossible for meanings to remain stable through changes in belief or sensibility. So when a proponent of “resemblance” model says that a certain emotional disposition helps to fix the reference of the term, she is making an interesting and substantive claim: not all concepts will include emotions in their reference-fixing core. But things look very different on the jazz model of meaning. According to the jazz model, reference is fixed through holistic rationalizing interpretation. So we need to take all aspects of the subject’s current and past understanding of a term into account in assigning a reference. (And if the subject intends to coordinate with a linguistic community, we need to take the community’s practice into account as well.) So it’s hardly surprising or informative on the jazz model to point out that we need to take human emotional sensibilities into account in assigning a reference to moral concepts: you need to take everything into account. When assigning an interpretation to ‘American’, for instance, the interpreter will take into account all aspects of the linguistic practice associated with that term, including superficial stereotypical beliefs like ‘Americans like ketchup’, and ‘American have limited command of foreign languages’. Of course, even if they are included as input into the process of rationalizing interpretation, such beliefs won’t “leave their mark” on the ultimate metaphysical characterization of property picked out by the word ‘American’: they are too peripheral to our overall practice with the term. Nor will they “leave their mark” on the account of what is required to be competent with the concept expressed by the term. People’s emotional reactions toward Americans will of course also be part of the input into interpretation. As proponents of the jazz model, therefore, we are happy to accept that emotions play a reference-fixing role for the concept ‘morally wrong’, in the sense of being part of the input into rationalizing interpretation. The problem is that this isn’t saying anything very impressive: every aspect of understanding and use of a term plays this reference-fixing role. On the other hand, if we understand reference-fixing in the substantive way presupposed by the “resemblance” model of meaning, we must deny that emotions plays this privileged reference-fixing role. Indeed, we deny that anything at all plays the reference-fixing role posited by the “resemblance” model. The third and final role sentimentalists might take emotions to play in an account of moral concepts concerns conditions for competence. Sentimentalists typically hold that in order to be competent with the standard concept expressed by ‘morally wrong’, one must have the relevant emotional sensibilities. That’s not to say that an emotionless Martian could never coherently use moral terms to refer to moral properties. It’s just that they wouldn’t be competent with our standard concept, for they wouldn’t understand moral terms in the standard way – instead, their understanding would be parasitic or deferential in the way a blind person’s understanding of phenomenal concepts, or a scientifically ignorant person’s understanding of theoretical terms like ‘quark’ are deferential to others’ understanding. Once again, the background assumption here is the “resemblance” model of meaning: the sentimentalist assumes that genuine competence with the standard concept requires that one share a particular set of emotional, motivational or cognitive dispositions. But as we explained in section 2, the jazz model of meaning is not committed to there being a core understanding which is necessary and sufficient for conceptual competence: all that’s required for competence is a coordinating intention and a rough family resemblance with others’ understanding of a term. So on the jazz model, an emotionless Martian could be fully competent with our standard moral concepts. Still, there is a weaker sense in which the Martian’s use of the term may be deferential. If normal emotional sensibilities leave their mark on the property picked out by moral concepts in the way we have suggested, then those who lack that sensibility will need to rely on others’ emotional reactions in order to identify precisely which property is picked out by their term. The emotionless Martian will to that extent be epistemically deferential to others’ emotional responses, even if she is fully competent with the concept. But it’s important to see that, on the jazz model, the Martian is not the only one who needs to take others’ reactions into account. A coordinating intention requires one to take others’ substantive understanding into account in determining precisely what one’s own concepts pick out. So even those with normal emotional sensibilities will need to take others’ emotional dispositions into account in their best epistemic efforts – they are not justified in simply treating their own emotional sensibilities as the final arbiter in determining the threshold for moral significance. On the jazz model, everyone is committed to a kind of epistemic deference to the group, by taking the substantive understanding prevalent in the group as a whole as input into rationalizing interpretation. To sum up: as proponents of the jazz model of meaning, we are not committed to any core sentimentalist thesis about moral concepts. Although the emotions may play a minor role in delimiting the boundaries of moral properties, they play no distinctive role in a semantic account of our most central moral concepts. References 16 Australasian Association of Philosophy Meetings Melbourne, 9 July 2008