Corps de l’article

1. INTRODUCTION

The proposals to be outlined here depart from the idea that we do not have full introspective access to what our value claims express: we do not know whether it is a belief or a desire or some other attitude of ours that they express. Some people would disagree. They would claim that they do know what value claims express, either because they take themselves to have the relevant introspective knowledge or because, although they lack of this, they believe they are justified in asserting that value claims express, say, beliefs or desires. Of course, most of us know that when we make a value judgment we express a claim to the effect that something has a value. Although metaethics has clearly offered a number of incompatible, competing theories to explain this simple fact, there is no consensus on which of these theories is the correct or even the most plausible one. So, there are people who believe that, in fact, we are not justified in believing that we fully know what we express. This paper is about people like this—people who therefore adhere to the following idea:

(i) We are rationally justified in believing that we do not fully know what we express in our value claims (i.e., claims to the effect that something bears a value).

Clearly (i) can be questioned. For instance, we might agree that philosophers in general and metaethicists in particular do not see eye to eye on the cognitivism/noncognitivism issue, but deny that they are the experts on metaethical issues; they do not in fact tell us much about what it is rational or irrational to believe about what we express in our value claims. Such a skeptical approach to what metaethics has accomplished would, in my view, be a gross mistake. But be that as it may, I will assume that the question “What do people who endorse (i) express in their value claims?” remains an interesting one, given how many people endorse (i) and continue to make value claims.

It should be stressed that those who actually believe (i)—and I believe this includes people who hold metaethical views, but who have not necessarily studied metaethics—might think there are stronger reasons to believe, for instance, that value claims primarily express beliefs. I certainly do so. However, whatever particular metaethical view we adopt, in light of (i) we acknowledge there are good arguments against that view that we cannot at present satisfactorily meet. So, although we might still think that our view is, on balance, the best available, we recognize that unless these counterarguments are not overturned, the core question “What do value judgments express?” remains open. People who believe something to this effect still make value claims: de facto, being in this state of mind does not prevent people from thinking that something is good or bad, or horrible or great, or better or worse. But it is the tentative hypothesis of this paper that these people’s belief in (i) is reflected in what they express in making value claims.[1]

This paper, then, explores the viability of a novel metaethical proposal about what value claims made by the people just described convey—a proposal that promises to reflect the peculiar fact that these people do not quite seem to stand behind what they claim in natural-language terms. We should therefore look for attitudes that are in some way attended by a reservation. I shall refer to the proposal that I think explains people’s rational justification for making value claims whose full import remains obscure to themselves, and which I want to explore here as On-Conditionalism (henceforth, On-C).

It is worth saying at the outset that not everyone belongs to the group of people described above, and therefore On-C cannot be generalized to cover what all people expresses in their value claims. I trust the reader will keep this in mind. On-C is a proposal about what people who believe in (i) express (whether for the right or wrong kinds of reasons). It is these people I have in mind when I talk below about what “we” express.

In its initial form, On-C claims:

(On-C) All value claims are expressive of the cognitive state that p on condition that q.

Here is what On-C might tell us about a specific claim:

(On-Ca) “x is good,” asserted by a person in a situation S, expresses that x is good on condition that goodness exists.

I shall refer to the attitude, or attitudinal state, expressed by a value claim according to On-C as a conditional belief. And by “value claim” I have in mind a linguistic assertion—e.g., the uttering of a sentence, or the entertaining of a thought, containing at least one thin or thick value term (see Williams 1985, pp. 140–143). Here are two examples (henceforth I will not give any details about the speaker’s situation and so I will omit “asserted by a person in a situation S”):

(1) “Pleasure is good.”

(2) “Rescuing the child was a courageous thing to do.”

The gist of On-C is that there is no agreement (among people or among metaethicists in general) on what we express when we utter such sentences as (1) or (2). On-C suggests that, if we believe this, we should accept that when we make sincere value claims we are expressing conditional beliefs.

The expressions “value claim” and “value assertion” suffer from a well-known ambiguity. They might refer to some sentence, utterance, or conscious thought in a natural language employed by the speaker, or thinker, or they might refer to that which is expressed by the sentence, utterance, or thought in question. In what follows, unless otherwise indicated, I shall have the first sense in mind. Quite what utterances (etc.) express is debatable. Do they express mental attitudes or the contents of these attitudes? Generally speaking, I find the second of these suggestions more plausible than the first. However, since in the present paper I am more interested in the attitude than in its content, I will sometimes speak as if value claims express conditional beliefs (when it would be more accurate to say that it is the content of these attitudes that is expressed). As long as we remain aware of this ambiguity, it should not give rise to any misunderstandings.

The word “express” has several meanings as well. Here, the relevant meaning is connected with the following idea: to say that the value claim p expresses the attitude A is to claim that there is an attitude A such that the evaluator has to be in this A-state for it to be semantically permissible for him or her to make the value claim sincerely.

2. METAETHICS

Metaethics has been dominated by cognitivism and noncognitivism. Cognitivists maintain that value claims characteristically express the beliefs of the evaluator. Noncognitivists contest this claim,[2] but they face a number of more-or-less serious objections and challenges. The most important is that set by the so-called Frege-Geach problem.[3] Various noncognitivist and hybrid expressivist strategies (after Hare 1952) have been developed to deal with this problem, but I think it is fair to say that none is uncontroversial (e.g., see Strandberg 2015).

As I understand it, and for reasons I will come to later on, On-C is a cognitivist position. As a result, it avoids the Frege-Geach problem. There are essentially two kinds of cognitivism: one asserts that what value claims express are at least sometimes true; the other kind, known as error theories, assert that value claims—at least, those that attribute moral properties and relations—are always false (Olson 2014).[4] Again, it seems fair to say that there is no consensus on the success of error theories, and that the same goes for each of the many realist, correctness theories that have been presented.[5]

I would like to stress that if people who express On-C claims are unjustified in believing that it is rational to disagree on metaethics, they should no longer express conditional beliefs when making value claims. Eventually a correctness or error theory, or an expressivist or hybrid view, or some other metaethical theory might make On-C obsolete. It bears repeating, On-C is not based on the idea that all extant metaethical views are equally good or bad. Far from it—some views face greater challenges than others, and some are better supported by argument.

Is there anything to be said for the idea that On-C coincides with the kind of fictionalist advice that error theories have prompted?

Metaethical fictionalists endorse one of two views: (1) the view that, as a matter of fact, we are pretending that there are moral facts when we express value claims, or (2) the view that, while we are not, as a matter of fact, involved in such pretense (and express genuine beliefs in our value claims), we should be, because there are no moral facts (cf. Burgess 1983). What exactly we are supposed to be expressing by these “fiction claims” is not particularly clear. It might be the kind of attitude we take toward cases of fiction. But, again, it is far from clear what kind of attitude this is. Walton (1990) describes it as a sort of pretense, as a make-believe (cf. Yablo 2002).

On-C does not say that we falsely believe there are moral facts. Nor does it say that our claims are fiction claims. According to On-C, we are not participating in a pretense when we make value claims, so there is a noteworthy difference between fictionalism and On-C. When we claim that x is good we do so, according to On-C, because, for instance, we believe that it is a fact that x is good on condition that goodness exists. We are not attempting to make something that is not the case appear true. We are not trying to make x into something that it is not.

In some ways this paper explores unknown territory. For this reason, I cannot deal with all of the many detailed issues it raises. In the interest of presenting and explaining the big picture, I need to focus on the forest rather than the trees. I shall do so by discussing two tasks that proponents of On-C will need to undertake eventually.

3. TWO GENERAL TASKS

The two tasks are:

(T1) On-C must include an account of how conditional beliefs should be understood in relation to bona fide examples of value claims.

(T2) On-C must supply accurate specifications of the content of the conditional beliefs that evaluations such as (1) and (2) express.

The specifications referred to in (T2) should be assessed with reference to their implications for logical, representational, and dispositional/motivational features of value claims. For instance, since the internalism/externalism issue is far from settled, it would be preferable if On-C claims were open to both internalist and externalist interpretations.[6] Given the peculiar nature of conditional beliefs, it is not clear whether On-C claims are. In Section 7, I shall therefore point to a novel kind of internalism—a conditional kind that might be available to On-C internalists.

The success of On-C depends largely on whether we can specify in an accurate and plausible way what a value claim expresses. Here are some tentative specifications of the value claim “x is good”:

 “x is good” expresses that x is good on condition

(a) that goodness exists,

(b) that x’s goodness exists,

(c) that there are criteria for the goodness of x, or

(d) that x’s being good can be rationally assessed (in terms of x’s goodness/conditional goodness).

Any plausible On-C specification will eventually turn out to be a quite complex belief. Arguably (albeit not necessarily always), when we make value claims our beliefs will be complex because we have some idea of what the value-makers of the value bearer are. In what follows, I shall have (a) in mind for the most part. In general, metaethicists agree that cognitivism is the default view of our moral discourse. (Noncognitivists tend to agree, but regard it as a kind of fallacy to regard the default view as conclusively supporting cognitivism.) Given that value language seems, at least, to communicate moral facts, (a) appears plausible. Furthermore, of the above suggestions, (a) expresses the simplest (least intellectualized) and most general approach to such factual communication. For these reasons, I will henceforth focus on (a).[7]

One might argue that On-C is not helpful at all. Consider again On-Ca. Unless we understand the meaning of “good” in its specifying part, On-Ca is not fully informative. True. But recall that On-C is not so much a competitor to extant metaethical theories as it is a theory about the mental state of someone who expresses the opposite of a belief without reservation: a belief, in other words, with reservation—i.e., what could be aptly characterized as a belief with an uncertainty proviso regarding the very nature of the claim being made. This clarification considerably waters down the complaint that On-C is insufficiently informative. Metaethics as we know it would not be put out of business or be made uninteresting just by the fact that On-Conditionalism is true. A supporter of On-C would probably argue that our belief in (i) does not in any important way have the effect that On-C claims it has on what we express with our value claims. It may be that what we express is precisely what one of the extant metaethical views tells us, and that we just do not know what it is that we express. In effect, such a view would be saying that a belief in (i) has no effect on what we are expressing. This I find hard to believe, but as I have already made clear (see note 1), I will not discuss it (or other alternatives to On-C) here.

An On-C-specification should help us determine the assertability conditions under which it would be appropriate to make the value claim. Consider On-Ca again. If we can ascertain what these conditions are by determining what x’s good-making features are, on condition that goodness exists, we have an informative account. There have to be such good-making features (this is a logical requirement). The real issue is rather to determine what these subvenient properties are. I am inclined to say that this is largely a substantive issue; it eventually hinges on how we justify our evaluations. Admittedly, in maintaining this I might be displaying a bias in favour of some metaethical view(s), in which case On-C would not be as available to everyone as I would like to think it is. The drawback would be that someone who thinks, for instance, that x is good, and who also thinks that x’s value-making properties are analytically determined, would not agree that the issue is substantive in nature. However, those who take this combination of views are not going to believe it is rationally fine for me to disagree with their metaethical theory anyway: they are what we might call metaethical dogmatists. On-C does not sit well with metaethical dogmatists.

Since On-C asserts that we express beliefs, one might be tempted to conclude that it is only available to already-convinced cognitivists. But this is a non sequitur. Noncognitivists may have good (albeit inconclusive) reasons for applying noncognitivism to what some people express, but there are not obliged to insist that what people who believe in (i) express is also something noncognitive. Quite the contrary. Noncognitivists have always found it hard to explain away the realism inherent in our ways of talking about value. On-C provides an alternative that is in line with the objective character of value language without committing us to value realism (or error theory). On-C should therefore be attractive to forms of noncognitivism underscored by value irrealism.

But a noncognitivist might insist that, according to On-C, we express something noncognitive by our value claims—where “we” here refers to people who believe, among other things, that noncognitivism is sufficiently flawed to be rationally questioned.

Perhaps, but there are good reasons why a noncognitivist should not insist on this. The most important reason is, of course, that, since it is unclear in what way the conditional form is open to a noncognitive “attitude interpretation,” such a proposal would be seriously problematic from the start.

Another worry is whether On-C is consistent with the idea that if people disagree over “x is good,” they will also be in disagreement if they understand this claim in terms of On-C. It would be undesirable to find that On-C is vulnerable to the kind of objections that semantic subjectivists (who cannot account for value disagreement) face. But this worry can be put aside. Obviously, if a claims that x is good on condition that goodness exists, and b claims x is not good on condition that goodness exists, a and b disagree with each other. In effect, there will also be some kind of disagreement between an ‘on-conditionalist’ and a person who claims that x is not good and also believes that goodness is, say, a sui generis property. They will be in disagreement over whether x exemplifies a property that the conditionalist ascribes to x on condition that there is goodness. Furthermore, the adversaries will in effect be locked in metaethical dispute over the existence of such properties.

4. CONDITIONAL BELIEFS

Conditional beliefs have been discussed in the burgeoning literature on conditionals—and in connected work in philosophy of mind and decision theory. Their role in metaethics has barely been explored, though. To proceed, one might try to first determine the truth-conditions of On-Ca. On-C beliefs have something to do with conditionals, so we might ask, What are the truth conditions for evaluative conditionals?

Of course, it is possible that our On-C belief has truth conditions of the kind characterizing beliefs in material implication (Cf. Grice 1989, and more recently, Jackson 1998). That would mean that to believe that B on condition that A would be to believe the proposition “if A, then B.” But, although I am still uncertain about the precise nature of the On-C attitude, I am inclined to think that what I associate with an On-C attitude is not a material implication. If it were, it would follow, in the case of On-Ca (among other things), that x is good on condition that goodness exists if and only if it is either false that goodness exists or x is good. But this equivalence claim does not capture what I have in mind. Some element in the relation between the antecedent and the consequent is lost when we turn the truth conditions of the conditional belief that x is good on condition that goodness exists into those of a material implication. Since the falsity of the antecedent is sufficient for the truth of the conditional (if we understand it as expressing a material implication), just about anything will come out as good if On-C is about a material implication and the first disjunct is false. Thus, if goodness does not exist, it will follow from On-Ca that “if goodness exists, x is good” is true. Intuitively, there must be more to the proposal than this. But whether or not our On-C belief has the truth conditions of a material implication, to avoid being obliged logically to endorse absurd conditionals, we would be well advised (in light of Gricean conversational implicatures, for example) to try first to determine what would warrant the On-C belief. In other words, we need first to ascertain the assertability of such beliefs. In what follows, I will therefore leave open the question of what, if anything, the truth conditions are. Instead I shall list some suggestions about the nature of On-C beliefs, and discuss two of these in more detail.

It seems natural to begin with the idea that to believe that p on condition that q amounts to having a belief in a conditional—specifically, in “if q, then p.” On this approach,

(I) a conditional belief is (nothing but) a belief in a conditional.

Since there is more than one kind of conditional, (I) needs to be made more precise. I will turn to this need in a moment. Meanwhile, it might be a good idea to list the alternative that I will also consider in due course:

(II) a conditional belief is (not a belief in a conditional but) a sui generis belief.

There are other possibilities. For instance, you might agree that an On-C belief is a sui generis attitude but want to resist the idea that it is a belief. You might, for example think that in order to qualify as a belief the attitude has to be truth apt and you might conclude, because you think conditional beliefs cannot be true and false, that they cannot be genuine beliefs. Instead you could endorse the notion that a conditional belief is a sui generis cognitive attitude but refuse to accept that it is a belief (and hence say it is an example of neither I nor II).

Limited space and the fact that any discussion of this latter example is bound to overlap with a discussion of (II) mean that I shall not discuss it here. I shall concentrate instead on (I) and (II), and eventually I will suggest that neither is satisfactory. However, as I shall argue in Sections 6 and 7, this does not leave us completely in the dark: further considerations will help us to get a better understanding of the nature of On-C beliefs.

5. A BELIEF IN A CONDITIONAL

In the literature on conditionals there is a much-discussed test for examining the conditions under which conditionals should be believed. This so-called Ramsey test is standardly applied to beliefs in indicative conditionals,[8] but the test can be used as a guide to how we should understand other kinds of conditionals as well.[9] Roughly speaking, it says that the conditional proposition “If p, then q” is acceptable (or accepted) in a given state of belief S just in case q should be accepted if S were revised with the new information p.

Just what this revised belief state amounts to is not obvious. Minimally, the revision would amount to giving up beliefs that are contradicted by p, leaving everything (or nearly everything else) intact in S. So, for example, to determine whether “If it rains in Berlin, I will believe there are wild tigers in Germany” is credible, I should ask myself, What if I were to believe that it was raining in Berlin? Should I accept that there are wild tigers in Germany? It seems not, so this is a conditional I should not accept. When we express an acceptable indicative conditional, we seem to be giving voice to our disposition to change our beliefs in the light of new information.

Ramsey’s test gives us some indication of what characterizes On-C beliefs.[10] We could test whether a certain On-C specification expresses a Ramsey conditional (i.e., one that is acceptable as outlined above). Suppose I endorse On-Ca and that I then learn that goodness exists. What should we expect in such a case? I think it is safe to say that when I have revised my information about the existence of goodness, I should now think that x is good (on pain of being illogical).[11]

But whether it is a belief in an indicative or a subjunctive conditional, it seems that it should be something that would stand the Ramsey test. So, if you accept On-Ca, the following should be acceptable:

(Pos) You ought to believe that x is good, if you learn that goodness exists.

But what about the following idea?

(Neg) You ought not to believe that x is good on condition that goodness exists, if you learn that goodness does not exist.

While I am ready to endorse Pos (supposing, that is, that On-Ca is correct), I am less confident about Neg. It seems I can go on believing that x is good on condition that goodness exists even if I believe that goodness does not exist. Of course, it is perfectly understandable that others may become upset with me if I withhold the fact that, actually, I also believe that goodness does not exist. But this has more to do with Gricean conversational implicature than it does with what we can or cannot believe.

Consider an analogous case. I might believe that x is a mermaid on condition that mermaids exist. The x with which I am confronted satisfies all the conditions that I believe a mermaid logically must satisfy to be a mermaid: it has a fish tail, lives in the sea, has the upper body of a woman, and so forth. But, although I am unaware of this, x, or at least x qua mermaid, is not a real being (e.g., because x is a hologram). Suppose I then come to believe that mermaids do not exist. If this belief does not upset other beliefs I have about what something has to be in order to be a mermaid, it seems that I could still believe that x is a mermaid on condition that mermaids exist. After all, I am saying not that x is a mermaid, but that x is a mermaid on condition that such creatures exist.

This seems to be a case in which we endorse a conditional about something where it is a merely contingent fact that this something does not exist. As far as I know, mermaids could exist. My initial inclination to accept that x is good on condition that goodness exists, and to maintain this belief even if I come to know that goodness does not exist, is perhaps likewise dependent on the belief that goodness could exist. The difference is, I suppose, that I am convinced that mermaids do not actually exist (it is a topic about which I am presently not willing to agree to disagree with someone who thinks they exist), but I am not convinced that goodness does not exist.

Prima facie, it is somewhat strange to think that goodness exists qua possibility but not in the actual world. This suggests that I have a reason to give up my intuition that I could still believe that x is good on condition that goodness exists even when I am sure that goodness does not exist. Perhaps it is simply too odd to imagine that goodness exists in a possible world but not in the actual world—which would suggest that we should give up talking about the existence of goodness in the first place. This would be fine, I think. The On-C pattern is not tied to the existence proposal (a), which is discussed here in part because it helps us to unravel some of the issues that On-C runs into. But, be that as it may, there might be a way of modifying our first response to this issue and rendering it slightly less odd.

Goodness claims are supervenience claims, so we might interpret the claim “Goodness does not exist” as entailing “(Presently) there are no (things such that they have or are) goodness-makers.” This claim strikes me as too implausible to be accepted, but it is somewhat less implausible than “Goodness does not (necessarily) exist.” Hence it might be possible to explain how one could go on with the conditional belief “x is good on condition that goodness exists” even if one happens to believe that goodness does not exist. For one could persist in the conditional belief if one were ready to believe that x might after all turn out to be good if goodness existed; and one could believe this because one thinks that if goodness existed, the properties of x would in effect be successful goodness-makers.[12] One could have thought that certain properties (a fish tail, etc.) made x into a mermaid on condition that mermaids exist. Similarly, one would think that x being generous, helpful, and so on, are goodness-makers on condition that goodness exists. The fact that goodness does not exist is no reason to retract the conditional belief.

So far it is hard to draw any conclusions at all from what has been said above. Perhaps the most we can conclude is that the idea cannot be excluded that a conditional belief is a belief in a conditional that in some cases might meet a Ramsey conditional. That is not much to go on, but it suggests a direction future work on On-C might take. However, as I shall argue later on, there is more to On-C beliefs than the idea that they are Ramsey-conditional beliefs. Meanwhile, it will be helpful to consider a proposal to the effect that a conditional belief is a sui generis attitude.

6. A SUI GENERIS ATTITUDE

Leitgeb (2007) argues that we should resist the idea that a conditional belief is (nothing but) a belief that entertains a conditional proposition. However, his conclusion is different from others who also have questioned the equation. Leitgeb does not want to give up the idea that conditional beliefs are in fact beliefs. His suggestion is that they are merely beliefs of a special kind that do not have a single propositional content.

Leitgeb takes insights from the logical systems for conditionals, probability, and belief-revision to support the idea that we should regard conditional beliefs as mental states amounting to a special kind of belief.[13] These mental states, he thinks, are conditional, even if, as he believes, they lack conditional content (in the sense that they are not beliefs in a conditional). Leitgeb depicts these conditional beliefs as cognitive attitudes that are “neither true nor false and which have two propositional contents rather than just one” (Leitgeb 2007, p. 116).

The idea that conditional beliefs are not beliefs in a proposition is a recurrent theme in Leitgeb’s paper. As he puts it, “it will be an integral part of our theory to claim that a conditional belief is not a belief in a unique proposition of whatever sort” (Leitgeb 2007, p. 118).

Leitgeb’s account is intriguing. Not all of its details are clear to me, though. For instance, I am not quite sure whether he actually identifies a conditional belief with the above dispositions or whether instead he takes conditional beliefs to be attitudes that result from, or necessarily involve in some way, a disposition to believe that B on condition that A is believed (Leitgeb uses the variables A and B). In light of the fact that what we often express when we make a value claim is an occurrent thought (with a certain natural-language content), I shall suggest in Section 7 that we should distance ourselves, when advocating On-C, from the idea that conditional beliefs are identical with the dispositional state Leitgeb has in mind. But before this, I want to set out an intuitive argument for the claim that there is more to conditional beliefs than Leitgeb’s idea that a “B on condition that A” belief consists in a “disposition of acquiring the belief in B given the circumstances that A is believed” (Leitgeb 2007, p. 124).

The Ramsey-conditional dispositional state on which Leitgeb focuses fits well the description of a belief that we would have (but do not occurrently have) if certain conditions obtained. But beliefs of the kind I am interested in are different: they are occurrent beliefs that I have, but which are conditional in nature. Intuitively, there is a distinction between these and the dispositional state that Leitgeb describes—for the following reason. Suppose I think (a) that the ice cream melts on condition that the temperature does not fall below 0°C. If I occurrently believe what I think, we should expect that I will take precautions depending on whether I want the ice cream to melt or not. It seems clear that we do indeed take such precautions that involve such thoughts. But suppose next that I am in Leitgeb’s dispositional state (b). If the temperature does not fall below 0°C, I shall believe that the ice cream melts. In that case, things are quite different. As for being in the dispositional (b) state, I would only take some relevant precautions if I occurrently believed either that the temperature was below 0°C or that it was above 0°C. So, if we think it is credible to say that we might take different kinds of precaution because we occurrently think (and I shall assume) believe that something is the case, on condition that something is not the case, then we are moving beyond a pure Ramsey-conditional dispositional view.

In the next section, I shall tentatively suggest that the further element the On-C account requires (over and above a purely dispositional element) is not an account of beliefs that we would have if a certain condition obtained, but rather an account of conditional beliefs as I have outlined them here. Such beliefs might have to be understood in terms of some dispositional states. In fact, they might turn out to have something in common with what Leitgeb has described. But I shall not try to clarify what such a state would involve. I think I am in good company (for very much the same reasons that Leitgeb has suggested) when I say that we presently lack the theoretical tools to describe the content of these conditional beliefs, and their dispositional bases, fully.

7. WHAT WE MIGHT EXPRESS BY “X IS GOOD”

To explain why On-C requires something more than a belief that we would have if some condition were met, we need to make a few distinctions that were brought into play in the above outline of conditional beliefs. It is customary to distinguish between occurrent and dormant attitudes (e.g., Audi 1972), and thus to separate occurrent and dormant believing. Arguably, occurrently believing that p is a manifestation of the dispositional state we are in when we dormantly believe that p. I shall also assume that a dormant belief is an unmanifested dispositional state (at least, as regards the corresponding occurrent belief), and that the corresponding occurrent belief is not necessarily the only manifestation of this dispositional state.

Consider again On-Ca and what follows if we accept the above outline. Accepting the outline, we see that the belief we express according to On-Ca is arguably either an occurrent conditional belief that is the manifestation of a dispositional state or indirectly related to this state in some relevant (causal) way.[14] From a representational point of view, acknowledging this is consistent with acknowledging that the intentional content of the attitude state in which we are when we think x is good is in effect a conditional content of the kind “B on condition that A.” Applied to the case of “a is good,” this just is a way of saying that what we express in asserting “x is good” is a belief in a conditional that is directly or indirectly caused by some dispositional state the nature of which we do not presently fully grasp.

A caveat is in place at this point: what I am considering here is not the denial of Leitgeb’s claim that a “conditional belief is not a belief in a unique proposition of whatever sort” (Leitgeb 2007, p. 118). The belief that I suggested might have the content “B on condition that A” should not be identified with the conditional belief to which Leitgeb refers. It might be intimately related to it, but it is not identical with it.

Thus, we might agree that a belief qua dormant dispositional state does not have a single propositional content. Let us refer to this dispositional state as DS. However, we might also think that when we do express something, such as that x is good, this expression must be the product of a manifest occurrent dispositional state.[15] Let us refer to this occurrent state as ODS. Arguably, there is more than one way for a DS to turn into an ODS. I shall therefore assume that it makes sense to suppose that the ODS can either be directly caused (or constituted) by the dormant dispositional state DS or be an occurrent dispositional state indirectly caused or constituted by DS.

If we acknowledge this, we can admit that the belief that B on condition that A does not fully capture the dispositional state (whatever that means) of an On-Conditional belief. Consistently with this, it might still (indirectly or directly) be the product of this state.[16] Given this, we should perhaps accept that the intentional content of the attitudinal state in which we are when we think x is good is in effect a conditional content of the kind “B on condition that A.” The price of this is high, though. What we express does not fully disclose our dispositional state.

Leitgeb might be right in thinking that there is a dispositional state of the sort he describes, one that allows us to discern a determinate kind of mental state—or, what he refers to as a conditional belief. He might even be right that no single belief in a conditional can be said to correspond to this state, if we take this to mean that the thought fully discloses what we are disposed to do when we are in this kind of state. But I am not quite sure why anyone would deny the latter. At least, it does not seem far-fetched to think that a thought with a single conditional content might be the product of such a dispositional state.

The second possibility I would like to briefly address is perhaps more plausible. What I have in mind is an alternative that takes the content of the belief to be categorical rather than conditional. So, what we express in making value claims is, in fact, attitudes with a categorical (i.e., nonconditional) content. This has the important advantage that it helps explain why people continue to make unconditional value claims even when they express conditional beliefs.

This option might appear counterintuitive, but appearances can be deceiving. If we keep in mind the distinction between the content of a belief and what this belief consists in (i.e., a certain dispositional state), and if we then add the second idea that the content of the belief need not fully reflect the full dispositional state, this alternative might be more interesting than the option I discussed above. It seems at least conceivable that a categorical belief might be the (direct or indirect) product of a Leitgebian (or of some other here unspecified) conditional dispositional state.

Actually, there is an indication that a belief with categorical content might be an expression of a conditional dispositional state. Let us suppose that I utter “x is good.” My claim might then be an expression of what could be appropriately referred to as a conditional dispositional state just in case it is not causally dependent on the belief that goodness exists. If it turns out that people who express the belief that x is good do not give up their belief that x is good when they learn that goodness does not exist (which is what we should expect if they had expressed a nonconditional belief), but rather modify their belief by conditionalizing it, then it seems to follow that they were expressing a conditional attitude. Hence what we expressed by “x is good” was in effect elliptical: we believe x is good but good on condition that goodness exists. (For what it is worth, in my experience this kind of response is not that uncommon.)

In fact, we might even go a step further and suggest that what makes the belief that x is good in some cases into a conditional belief is simply that it is not dependent on certain beliefs, period. There is no need to bring in any reference to a dispositional state like Leitgeb’s. What makes it into a conditional belief is (somewhat paradoxically) that it is causally resilient, at least when it comes to some kinds of belief. This suggests that we have two hallmarks of conditional beliefs: they play an explanatory role in precautionary contexts and they are causally resilient.

Both the dispositional interpretation and the nondispositional interpretation suggest, then, how the thought that x is good might in a sense be an expression of a conditional attitude. But the nondispositional interpretation seems to be the more daring (and perhaps daunting) suggestion, since it implies that there are at least two scenarios where we express a thought to the effect “that x is good.” We might give up the belief that x is good when we realize that there is no goodness. If this happens, we may conclude that our original belief was (causally) dependent on some other attitudes. Most of our attitudes are dependent on other attitudes. But (a belief directly or indirectly caused or constituted by) an On-C belief seems to be dependent on beliefs that differ from those on which the categorical belief that “x is good” depends. So, what makes expressing “x is good” into the expression of a sui generis conditional attitude is the fact that it is a causally independent attitude. Learning that goodness does not exist will not always lead to the same result when a person has categorically endorsed that x is good.[17]

Much work remains to be done if we are to clarify the two options. Although they are clearly tentative proposals, they point in interesting directions, and this bodes well for the future of the On-C project.

It is worth highlighting the following intriguing metaethical implication of the second option. Metaethical internalists maintain that if you express the thought that x is good, you favour x in some way (on pain of being insincere, expressing an ‘inverted-commas’ judgment, or giving in to weakness of will). This is hard to account for if one is not a noncognitivist. But even cognitivists who think they can be internalists about “good” face a genuine challenge if they think “x is good” expresses a belief in a conditional. Since “good” is embedded in the conditional, it is not obvious how anyone can account for the motivating feature of “good.” That is, it is not at all obvious why you would be motivated to realize or promote x in virtue merely of believing that x is good if goodness exists. What the above discussion suggests is that the On-C approach does not need to add any further burden to the internalist; if cognitivism is consistent with internalism, then there is hope that even an On-C approach can be an internalist view.

It should be clear that what we have, then, is a novel version of internalism. Plausibly, even if “x is good” is motivating despite being an expression of a conditional belief (i.e., a resilient belief that survives in the presence of certain other beliefs), it is not obvious that the motivation would withstand certain beliefs. We might be motivated at t1 by the thought “x is good,” but this motivation might not withstand the new belief at t2 that goodness does not exist. This new belief might not mean that we give up on On-Ca, but it seems reasonable to suppose that we are not (or not as) motivated by our conditional belief at t2. Thus, our motivation becomes conditionalized. But that it is conditionalized does not mean that externalism is (now) the correct position; the belief still motivates at t1. Externalism tells us that motivation only occurs when something external is added to what we judge to be good. By contrast, On-Conditional internalism tells us that our judgment is motivating, but that it ceases to motivate us in certain cases when something external is added.