EMOTIONAL TRUTH

Now published in Proceedings of the Aristotelian Society Suppl.  Vol.  76:247-263  as the first half of a symposium with Adam Morton (July 2002) 
 
 

ABSTRACT: Taking literally the concept of emotional truth requires breaking the monopoly on truth of beliefs-like states. To this end, I look to perceptions for a model of non-propositional states that might be true or false, and to desires for a model of propositional attitudes the norm of which is other than the semantic satisfaction of their propositional object. Those models inspire a conception of generic truth, which can admit of degrees for analog representations such as emotions; belief-like states, by contrast, are digital representations. I argue that the gravest problem-objectivity-is not insurmountable.

I  Generic Truth.
A `true likeness' is not one that is not false. When we say that Tolstoy's novels are true to life, we don't mean to claim that they are, after all, non-fiction. In these and some other domains we speak of truth, but assume we are not speaking strictly. Must this be the case for emotional truth? The phrase sometimes refers to kindred properties such as authenticity, a difficult notion worth elucidating, but about which I have little to say. I propose instead to take literally the idea of truth-valued emotions.

The concept of emotion is Janus-faced. In one direction emotions face inward, either as `perceptions referred to the soul' (in Descartes), or as perceptions of bodily states aroused by some exciting cause (in William James). In the other direction emotions face outward, suggesting that (at least some) emotions provide us with correct or incorrect representations of something in the world outside us. It is in this facing-out stance that emotions might claim to be literally true or false. In pursuit of this hypothesis, I shall offer some reasons for assigning a broader scope to the concept of truth as correspondence, and survey some of the difficulties that such an extension to emotions of the idea of literal truth may bring.

A mental state M can be said to be true or false, only if
(1) it is subject to a norm N;
(2) N is determined by M itself, yet
3) N looks for its satisfaction to some reality existing independently of M.

These are necessary but not sufficient conditions for standard ascriptions of truth-value. I shall postulate that they are sufficient to capture the core of a generic notion of truth, which might be summed up in this slogan: A story defines its truth, but whether it is true can never be part of the story. My question, then, is how this might apply to the relation between an emotion and that which, if anything, the emotion represents.

In philosophy the entities to which truth value is attributed directly are commonly held to be not mental states, but propositions. Mental states that incorporate propositions in a suitable way (I shall speak generally of `attitudes') inherit truth-value from their propositional objects. But it is unclear what propositions are. It seems safe to regard them as posits tailored to play just two roles: as objects of propositional attitudes, and as bearers of truth-value. Among the attitudes, beliefs then remain typical of those that admit of truth or falsity. They obviously satisfy the conditions just stated: a belief (1) specifies or `expresses' a proposition; (2) it thereby determines a norm, according to which it is true or false; and (3) the satisfaction of that norm is independent of the belief's existence.

A tight connection therefore seems to hold between truth and belief. For as is often noted-and as witnessed by Moore's paradox-the aim of belief is truth. Nevertheless, it would be misleading to say that truth-value belongs essentially to belief. For, as Frege made clear, we need to allow that propositions may remain unasserted. Otherwise the antecedent and consequent of any conditional would be asserted merely in virtue of figuring in the conditional, trivially short-circuiting Modus Ponens. More generally, belief's monopoly on truth might be infringed in two ways. First, some attitudes may lack propositional objects and yet also be true or false. Perceptions may provide examples. Second, one might attribute a truth-like property to other propositional attitudes, differing from beliefs in their aim. An example would be desires. In this second class of cases, one might ascribe truth-value derivatively to the attitude on the basis of the truth-value of its object. Thus one might say that my desire for oyster ice-cream is true iff I get some. But a more interesting analogy with the truth of beliefs would focus not on the semantic satisfaction of propositional objects, but on the attitude's attainment of its aim, its success. For belief, success is truth; but it lies elsewhere for other attitudes.

Emotions may stake a claim under both headings. Like perceptions, they sometimes lack a propositional object. And as in the case of desires, the truth of their propositional object does not define their success even when they can be said to have one. To see this clearly, recall Robert Gordon's observation that some emotion ascriptions are `factive', while others are `epistemic' (Gordon 1987, 26 ff.). The former, such as S is embarrassed that p, presuppose that the subject knows that p, while the latter, such as S fears that p, presuppose that the subject does not know whether p.[1] Obviously in the latter case, the truth-value of p does not determine the appropriateness of the emotion. Even in the factive case, however, the truth of p is not sufficient to vindicate the emotion. By contrast, the truth of p always vindicates the belief or the assertion that p.

But if semantic satisfaction does not determine the aim of an emotion, what does? To answer this we need to proceed on two fronts. We must explore the way in which a state without a propositional object might be true; and we must ask what might define the success of an emotion, in a way precisely analogous to the sense in which truth defines the success of a belief. The former quest will look for inspiration to the model of perception. The second will explore what it is for an emotion to be, in the relevant sense, successful.

Proceeding somewhat indirectly, I begin by acknowledging some of the difficulties that might threaten to scuttle the project before it gets off the ground. The most insistent difficulty, concerning the prospects for emotional objectivity, will bring me back both to the analogy of perception and to the reconstruction of a relevant notion of emotional truth as success. I shall conclude by sketching some reasons to adopt the suggestion that the specific domain of propositional truth, for which we are accustomed to reserve the literal meaning of truth, is distinguished by its digital as opposed to analog mode of representation.

II Some Logical Problems.
A basic feature of the paradigm truth-valued states-propositions, assertions, or beliefs-is that they can be negated. Furthermore, where the embedded proposition exhibits a subject-predicate structure, it can be negated in two ways, yielding contraries distinct from contradictories. Thus p is false if and only if ~p is true, while Fa is false if[not-F](a) is true. Is any such pair of standards for negation applicable to emotions?

Some named emotions are commonly felt to be polar opposites (love and hate, hope and despair, admiration and contempt, gratitude and resentment). Such pairs may plausibly be regarded as contraries, while equanimity-rather than indifference-might relate to both as their contradictory. But how are the norms of contrariety to be grounded?

Compare the case of desire. Beliefs demand consistency: if p and q are inconsistent, that inconsistency is automatically an indication that belief that p and belief that q cannot both be right. By contrast, someone might suggest that no such demand exists for consistency of desire. For two desires may aim at inconsistent states of affairs without entailing that at least one must be mistaken.

This is partly right, but harbours an important confusion. It presupposes that a single criterion of consistency is appropriate to both beliefs and desires. But that presupposition begs the question against the distinction alluded to above, by confusing the satisfaction conditions of desire with condition of success. For any two beliefs, compatibility coincides with consistency. But for two desires to be consistent, it is not necessary that their contents be jointly satisfiable, but only that their contents be jointly desirable (de Sousa 1974). So while a desire for p and a desire for q (where q implies ~p) are clearly incompatible, it does not follow that they should be regarded as inconsistent. And while this raises difficult questions about how to cash in the claim that two desires are inconsistent, it makes room for inconsistent desires without requiring that consistent desires also be for compatible objects.

Emotions are similar, but messier. The reason is that there is no single proper object of all emotions. Each emotion is linked to its own specific evaluative continuum, and so defines its own proper object, and thereby the dimension along which contrariety might be defined for that emotion.

To illustrate how the distinction between truth and satisfaction might work out for a standard emotion, consider the example of fear. This can readily be construed as having been honed by natural selection to favour the avoidance of danger. The formal object of fear-the norm defined by fear for its own appropriateness-is the Dangerous. Fear that p is satisfied iff p is true, but it is successful iff p is actually dangerous. In general, for any emotion sufficiently complex to afford the identification of a propositional object:

E(p) is satisfied iff p is true
E(p) is successful iff p actually fits E's formal object.

Where the emotion admits of a target (t) but lacks a propositional object (as in certain kinds of fear), semantic satisfaction consists in successful reference, while success still depends on whether the target fits the formal object:[2]

E(t) is satisfied iff t exists
E(t) is successful iff t actually fits E's formal object.

In all cases, the emotion's success is independent of semantic satisfaction. Fear of monsters is not semantically satisfied, but may be successful. The converse may be the case in fear of spiders.

Emotional truth, then, refers not to semantic satisfaction, but to success. I follow widespread practice in saying that fear's assessment of p or t as dangerous consist in some sort of evaluation of p or t. Success is tied to the correctness of that evaluation, and I will need to say more below about how the evaluation relates to the rest of the emotional experience. But this suffices to suggest how the notion of opposition, if any, appropriate to a given emotion is internal to that emotion. And while this provides no handy criterion of emotional contrariety, it at least suggests a way in which such a concept might have application, as well as explaining why it is difficult to cash out in practice.

Another disanalogy is sometimes adduced between belief and desire and might apply a fortiori to emotions. When deliberating about what to do, there comes a moment when it is appropriate to say: Now is the time to decide. And then one does so, definitively and rationally. But deliberating about what to believe is different. For that amounts to making up one's mind about what is true, and there is always a gap between the rationality of making up one's mind about p and the truth of p. At best there can come a moment when I am justified in making up my mind. But that cannot give me a rationally sufficient ground for the truth of the proposition. At best, it can be the right moment to decide only on the rationality of behaving as if it were true.[3]

Yet this contrast too is misleading Admittedly, the pressure of time can furnish a reason to decide (that it is rational) to believe that p, but can never be evidence for p. What is rational to believe is only my best bet under current constraints. Nor is deciding to believe p equivalent to deciding to act as if p were true. But the following parallel still holds: while the pressure of time and other constraints can be a perfectly good reason for deciding (that it's rational to) do p, it can't be grounds for the proposition that p is objectively best, or even that p is what will seem best in the light of infinite consideration.

The crucial disanalogy between beliefs and other attitudes lies elsewhere. Only one of two incompatible beliefs can be true, and therefore only one can be successful. Among incompatible desires or emotions, on the other hand, no single desire or emotion need be uniquely successful.

The very idea of an objective best, however, may seem to beg the question against a prevalent view that neither emotions nor desires can be assessed in terms of anything objective at all. To this I now turn.

III The Claim of Objectivity.
On the problem of emotional objectivity, Plato made an early start on two fronts. The Philebus argued that pleasures can be false, not merely in the derivative sense of being associated with or caused by a false belief, but in itself. That claim, extended to (other) emotions, presupposes that there can be an objective correlative to pleasure or emotion that is not a mere projection. That demand is made explicit in the Euthyphro, where Plato posed the problem of whether the gods love piety because it is pious, or whether calling it pious is merely to claim the gods love it.

The meaning of `objectivity' is subordinate to the contrast with `subjectivity', and that term has at least a dozen different senses (de Sousa 1999). But a clear paradigm of objectivity can arguably be found in mathematical intuition. Imagine someone saying: I understand your statement that all triangles have three sides, but I disagree. One would be confident in objecting: Your `disagreement' suffices to show you have not understood. But unexpectedly, at the other, most subjective end of the spectrum, it seems plausible to admonish someone who doesn't share my individual tastes with a curiously similar demand for taste universalization:

(TU) If oyster ice-cream tasted to you as it does to me, you could not fail to find it delicious.

The cases seem very different, for in the case of taste we lack any independent way of supporting the counterfactual. Nevertheless, it is instructive to explore this further. Since taste exemplifies extreme subjectivity, it would not be surprising if it failed to meet the conditions for generic truth. If despite that it succeeds, on the other hand, we may assume that more complex emotions will also pass the test. If not, then to diagnose exactly how it fails might help us to discern how emotions must differ from taste if their claim to be truth-valued is to be vindicated.

If my taste for oyster ice cream (TOI) could be said to be truth-valued, the following must hold: (1) TOI must be subject to a norm of appropriate liking or aversion (the `valence'),(2) that norm of appropriate liking or aversion must somehow be defined by the character of the taste (the `quale'), but (3) the quale cannot suffice to determine the satisfaction of the norm in question.

One problem is to make sense of the quale's defining its own norm of success-the appropriateness of liking or aversion. A second problem is how to make the relation between the quale and the valence contingent: it must be possible for an inappropriate valence to be present or an appropriate one to be absent. In other words, if taste is really objective then taste universalization (TU) must fail.

Suppose we think of all actual experience, on the model of a mathematical domain, as located in a multi-dimensional space encompassing all possible experiences, had by any possible conscious beings. Only some very limited ranges of experience are available to any specific kind of conscious being. (Human experience of colour, for example, can provide only partial insight into the experiences available to tetrachromatic animals.) At any point in that space, the valence of a specific experience is one of the qualitative dimensions of experience. It would then follow trivially that two experiences could not be qualitatively identical while being opposed in valence. On this picture, is there any prospect of prying apart the valence of an experience and its quale?

There are two possibilities. Either valence is a component of complex qualitative experience, or it is supervenient on other qualia. Understood in the first way, valence could always in principle be dissociated from the concomitant qualia. If my experience of oyster ice-cream consists in qualia [A,B,C liking], and yours in [A,B C, aversion], then we are not having the same experience. This would not preclude the required contingency. But what could be the measure of appropriateness between [A, B, C] and one or another valence?

It is tempting to appeal to Human Nature to set a standard of correctness. Given any quale, an evaluative response that falls foul of the norm will lack appropriateness, and on that basis we can call it perverted, abnormal, or false. The problem with human nature, however, is that if it refers to a set of interesting properties true of all and only humans, and robust enough to support normative standards, then there probably is no such thing (de Sousa 2000). Still, the suggestion is worth setting on ice for partial recuperation and reconstruction, as I shall suggest in a moment.

Now consider the second possibility, that valence supervenes on other qualia. If supervenience is understood deterministically, it will preclude the required contingency of the relation between quale and valence. But the laws governing that relation of supervenience might be stochastic, allowing two or more alternative outcomes. That would restore contingency, and providing one valence can be made out to be more appropriate than the other it would then satisfy the conditions for generic truth after all.

This last requirement remains very far-fetched in the case of taste. But it is much more likely to be met in the case of those emotions that are plausibly characterized as perceptions of value. Take, for example, the classic thought experiment in Mencius: you see a child about to fall into a well, and your apprehension of the situation immediately moves you, and you want to save the child. In this instance, what is apprehended is the need to intervene. Or better it is the nature of the total situation, in which the need to intervene roughly sums up the supervenient valence. Yet it is not impossible to witness the scene without being moved thus. Anyone whose experience lacks the appropriate valence, however, may be said to have an objectively false emotion.

This way of describing the situation avoids simple projectionism, insofar as what I perceive is not merely the shadow of my own response, but something about the character of a situation as a whole in the context not only of my own singular responses but of the feelings and interests of others. The choices to which I am led are products of a multi-dimensional landscape of valuess constituting a larger axiological whole. I call this view `axiological holism.' It stipulates that we do not apprehend value in discrete units. but only in the light of a complex of factors that transcend individual experience. No single range of facts suffices for the success of an emotional response. Biological facts will speak to its origins and may thereby assign it a proper function in the sense of Millikan (1989), but they will not determine its relation to currently relevant norms. Social norms, in turn, are every bit as likely to be irredeemably nasty as biological ones. (To endorse social norms as the touchstone of normativity would be to condemn all social reformers.) Individual biography sets up paradigm scenarios in terms of which each individual understands the world, but this defines only a narrow sense of fit between a current response and a present situation. That fit cannot be identified with value in any comprehensive sense, still less determine what is morally right. (D'Arms and Jacobson 2000).

All of these factors-biological, social, or personal, and more-may properly be confronted with one another in the hope of arriving at something like reflective equilibrium. That holistic equilibrium is as close as we can come to reconstructing a notion of normative human nature. And perhaps it is close enough. In this way, we may find some emotional responses mistaken, just as Macbeth found (`Art thou not, fatal vision, sensible to feeling as to sight?') that a perception can fail the test of corroboration by different sensory channels. Vision provides distal information about our surroundings; yet visual illusions occur. Similarly emotions in general constitute apprehensions of axiological reality; yet not every emotion is equally to be trusted. We tell which is right and which is wrong much as we test the veracity of perceptual information: by appealing to corroborating evidence. Something like the method of reflective equilibrium is commonplace in science as well as in ethics; what is not often noticed is that the items that need to come to equilibrium are typically emotional responses. The search for reflective equilibrium plays an important role not just in moral deliberation but also where the issue is purely epistemic, where, as Christopher Hookway (1998) has argued, emotions such as the feeling of plausibility or doubt play a crucial role. Without such emotions, even the most comprehensively rational argument may remain powerless to move us.

Equilibrium, it may be objected, establishes only coherence. And it is an excessively weak theory of truth that is satisfied with coherence. Compare perception again. Each sensory channel provides a specific mode of information. But primary qualities are apprehended through different sensory channels. Multi-modal access is the warrant of objective reality. What then is the analogue of multi-modal access for emotions?

IV  The scope of our emotional access to value: a musical analogy.
A helpful analogy is suggested by fascinating paper by Dmitri Tymoczko (2000) on Milton Babbit and John Cage. Tymoczko describes both composers as philosophers who, in their different ways, questioned the relevance of beauty or pleasure to aesthetic appreciation. Babbitt, in particular, aims to break the link between value and the ordinary listener's emotional response to music:

Advanced music, to the extent that it reflects the knowledge and originality of the informed composer, scarcely can be expected to appear more intelligible than [higher mathematics] to the person whose musical education usually has been even less extensive than his background in other fields. (Babbit, quoted in Tymoczko)

Does that amount to an expansion of the scope of music, or to a reductioad absurdum of Babbit's methods? There is, perhaps, a faintly discernible third possibility: namely that the domain of music is not actually circumscribed by our capacity to hear patterns, nor by the emotional responses typically evoked by the acoustic domain. Whatever the merits of this view of music, it provides a model for an alternative perspective on the objective correlates-the potential truth-makers-of emotions in general. Recall Tymoczko's objection to regarding as music what cannot be appreciated by the ear, however well trained: If `the relationships are out there, in the objective world, but we cannot apprehend them', they must then cease to count as music. But precisely, Babbit might respond, what matters is that the relationships are objectively there. And if they are continuous with patterns that can be heard as well as apprehended in other ways, that boosts their claim to be regarded as objective. So much follows from the principle of multi-modality as a touchstone of objectivity.

The retort would have some force; and the thought generalizes to non-musical emotions. If the values apprehended by emotions are objective, we might expect that they are not exhausted by actual emotional responses. This needn't commit one to the existence of a Platonic world of values which our emotions apprehend only dimly; but it does evoke the possibility that, just as Babbit's music might be appreciated on paper, in the spirit in which a mathematician apprehends a proof, even by those whose auditory capacities are not up to hearing them, so one might, by a non-emotional process of ratiocination, apprehend values inaccessible to the emotional capacities of people at some given stage of personal, social or biological development.

On this view, something like emotional experimentation may, by analogy to musical experimentation, enlarge the domain of values to which we have access. But while the domain of values is not independent of the facts about conscious beings, it is neither simply projected from, nor ever exhausted by, the actual repertoire of human emotions-any more than all possible thoughts can be exhausted by the repertoire of actual humans thoughts past, present, or to come.

V A Test Case: Huckleberry Finn.
Several philosophers have discussed Huck Finn's decision to give up on morality and `take up wickedness' by 'stealing Jim out of slavery'. Everyone agrees that it is Huck's emotions, as opposed to his explicit moral principles, which produce the true answer. In Huck, the two faces of emotion merge: his authentic emotion is also the true one. It corresponds to objective values which he apprehends, despite his conviction that he is doing wrong.

But there is some dispute about how to best describe the case. McIntyre (1990) argues that the main lesson of the story is that if akrasia is defined as a conflict between what one believes to be one's best reasons, all things considered and the real reason on the basis of which one acts, an akratic action may be entirely rational, because one may be mistaken about one's own best reasons. She contends that Jonathan Bennett (1974) wrongly sees Huck as irrational because he characterizes `Huck's dilemma as one in which general moral principles and reasons conflict with "unreasoned emotional pulls".' (p. 381, quoting Bennett p. 127). Rather, she suggests, Huck need not be viewed as irrational even while he is akratic. For what he does is inconsistent not with his actual values but only with what he falsely believes to be his values (McIntyre, p. 386). Her point is not merely that some objective reasons might exist to justify the akratic action, but that the so-called akratic might, after all, be doing the right thing from her own point of view. Rationality in action is `evaluative consistency,' and that could be attained even if the action were akratic in the sense just defined:

Evaluative consistency may exist, for example, in view of the fact that if the agent had had more time to reflect, she would have changed her mind about what the best thing to do would be. Thus she would have been saved from akrasia not by changing her behaviour but by changing her evaluation of it. (ibid).

First, it appears to be neither necessary nor sufficient that the various emotions participating in the weighing in search of equilibrium be conscious.

Second, despite the fact that standards of contrariety for emotions are, as we have seen, obscure, it is principally emotions themselves, and not propositions, which are weighed against one another in the quest for reflective equilibrium.

Third, in the case of Huck Finn, the veracity of an emotion is hard to disentangle from its authenticity. We touch here on what I called at the outset the inward-looking face of emotion. The sense of emotional truth I have sought to articulate is one which posits a correspondence between the emotion, characterized by a specific formal object, and some property of the human-inhabited world. But the values apprehended by emotions depend in part on who we are. They are no less objective for that; but what reflects my own individual nature-what makes for my emotional authenticity-therefore comes to seem, after all, potentially relevant to the objective world of value.

VI  Species of Truth: Emotional and Propositional.
I began by advocating an extension of our notion of truth as correspondence, based on a core intuition that can be summed up in the slogan: A story defines its truth, but whether it is true can never be part of the story. (A corollary notoriously dooms the ontological argument: Whether a thing exists cannot be part of its nature.) This required, in effect, the satisfaction of three conditions, which I conclude that at least some emotions are able to meet.

(1) Emotions are subject to a norm defined by their formal objects: what I fear must be dangerous; she of whom I am jealous must figure in a certain sort of triangle; what angers me must be a wrong.

(2) The norm in question is determined by the emotion itself. This is often manifest in the fact that there is an air of tautology about the characterization of the formal object: he whom I love must be lovable; what I regret must be regrettable.

(3) But the appearance of tautology is misleading, because the attainment of success for emotions-the actual fit between the object or target of the emotion and its formal object-depends on a vast holistic network of factors which transcends my actual response.

If emotions are properly said to be truth-valued in a generic sense, then the narrower class of truth-bearers traditionally targeted by philosophy-propositions or belief-like attitudes-no longer need to be regarded as the paradigm truth-valued attitudes. They form only a subclass of truth-valued states, a special case. What then is the differentia that sets them apart?

My hunch about the answer is this: Belief is digital; the representations involved in Truth's broader domain are analog. A digital representation is necessarily part of a system of representation, and can function only once all possible signals are assigned to a finite set of discrete symbols. An analog representation, by contrast, admits of varying degrees of precision and an indefinitely large set of possible symbols.

This hypothesis suggests another way in which emotion resembles perception. For while we can sometimes perceive that some proposition holds, in other cases the content of direct perception seems typically to be analog (Peacocke 1986). Furthermore it allays three worries raised by the notion of literal truth for emotions.

1. Generic truth legitimizes talk of more or less, by incorporating analog correspondence, which can be more or less exact. But as traditionally conceived, truth admits of no gradations. A proposition is either true or false: tertium non datur. While this is rejected by advocates of ontological vagueness or fuzzy logic, it can be seen to apply at most only to a species of truth-bearer. Within a narrower domain of digitized representation, there are no degrees. So we get the kind of on/off truth we associate with well-defined propositions.

2. Digitality is not necessarily conventional, as shown by the example of the digital system embodied in DNA. But insofar as digital representation exists only in the context of a system of discrete values suited to indefinite copying, most digital systems are likely to be conventional. This should lead us to expect that the typical examples of truth in the sense precluding degrees will be bound to the conventional medium of language.

3. Digital representation is essential to secure fidelity in multiple reproductions of a stable `message'. Insofar as they constitute a medium of social interaction, emotions tend to cluster into a limited repertoire of distinct entities, functioning as justifications and motivation for behaviour, regimented and `digitized' as a system of limited significant types (de Sousa 1997). But in the rich variety of their experienced reality, the significance of emotions is not limited to their role in influencing behaviour and social interaction. Regarded as experiences representing something outside themselves, their variety instantiates a limitless continuum, and they have no need to take on the digital character of propositions.

4. That standard truth-bearers are digital representations helps to explain the grain of truth in the often expressed anxiety about the distortion of reality introduced by abstractions. Abstraction is, by definition, a process of pruning details, of ignoring certain distinctions and aspects of reality. Since all thought requires abstraction, all thought is risky. None escapes the danger that the most important aspect of reality for present purposes is precisely that which our abstractions have left out. A vague aspiration to the `whole truth', which no utterance can contain, lies behind Nietzsche's charge that (propositional) `truths are illusions which we have forgotten are illusions'. (Nietzsche 1993, 43). The whole truth is an impossible ideal: but it usefully evokes an analog conception of representation. It has exactly the absurdity of a map on a scale of one inch to the inch, in which every nanometer is faithfully represented to scale. If emotions are, instead, conceived as analog representations of an axiological landscape, it may come to seem natural that they should admit of variable degrees of definition, instantiating a concept of accuracy that merges with generic truth.

FOOTNOTES

1. What Gordon actually says is that emotions are factive or epistemic. But that cannot be quite right, for several reasons that don't bear on the present point. (See de Sousa 1991).

2. This skirts around a current debate about whether there can be non-conceptual contents of perception. Some hold, while others deny, that perceptual content may be non-conceptual. Even in Pittsburgh, however, where all content is conceptual and every concept owes its identity to the inferences it licenses, it doesn't follow that every perception must boast a propositional object.

3. This point was made by David Owens in comments on a paper by Gary Watson, at a Montreal conference on Akrasia, May 2001. Since I draw on Owens' oral presentation and conversation, I can't be sure that he would endorse my formulation.

REFERENCES

Bennett, J., 1974, `The Conscience of Huckleberry Finn', Philosophy 49, 123-34.
D'Arms, J., and Jacobson, D., 2000, `The Moralistic Fallacy: on the "Appropriateness" of Emotion', Philosophy and Phenomenological Research 61, 65-90.
de Sousa, R., 1974, `The Good and the True', Mind 83, 534-51.
de Sousa, R., 1991, Review of Gordon 1987, Noûs 25, 373-376.
de Sousa, R., 1997, `Love Undigitized', in Love analyzed, R. Lamb, ed., 189-207. (Boulder: Westview).
de Sousa, R., 1999, `Twelve Varieties of Subjectivity', paper presented to the Sixth International Colloquium on Cognitive Science, San Sebastian, Spain. Available at http://sousa.artsci.utoronto.ca/subjectivity.html.
de Sousa, R., 2000, `Learning to Be Natural', in Being Humans, N. Roughley, ed., 287-307, 313-316. (Berlin, New York: de Gruyter).
Gordon, R. M., 1987, The Structure of Emotions: Investigations in Cognitive Philosophy. (Cambridge: Cambridge University Press.)
Hookway, C., 1998, `Doubt: Affective States and the Regulation of Inquiry', Canadian Journal of Philosophy, Suppl. vol. 24, 203-26.
McIntyre, A., 1990, `Is Akratic Action Always Irrational?', in Identity, Character, and Morality, eds . Rorty and O. Flanagan, (Cambridge, MA: MIT Press), 379-400.
Millikan, R., 1989 `In Defense of Proper Functions', Philosophy of Science 56, 288-302.
Nietzsche, F. 1993, `On Truth and Lies in a Nonmoral Sense', in Nietzsche: Selections, R. Schacht, ed. (New York: Macmillan).
Peacocke, C. 1986, `Analogue content', Aristotelian Society Proceedings Suppl. vol. 6,1-18.
Tymoczko, D. 2000, `The sound of philosophy', Boston Review Oct/Nov 2000. Citations are to the unpaginated on-line version at http://bostonreview.mit.edu/BR25.5/tymoczko.html.