MORAL RELEVANCE AND CETERIS PARIBUS PRINCIPLES

Robert L. Frazier

Abstract:

My goal in this paper is twofold: to provide an account of what makes properties morally relevant, and to indicate the role such properties have in our moral thinking.

I propose that a property is morally relevant just in case it must, ceteris paribus, determine the moral status (the rightness or wrongness) of actions having it. The main part of the paper concerns the conditions under which the ceteris paribus caveat is satisfied, that is, when other things are equal. I argue that the caveat is satisfied when, with respect to a proposed set of morally relevant properties, an act differs from its alternatives at most in the degree to which it has one of those properties.

Since other things are seldom equal, it is natural to wonder why what is true when they are equal should be important when they are not. That is, why is moral relevance, as I characterize it, a useful moral notion? I suggest that it is only by recognizing the moral relevance of properties that we are able to engage in useful moral thinking about the future.

Introduction

One form of ethical pluralism, which I shall call `property ethical pluralism', is the view that irreducibly different properties of an agent's alternatives are morally relevant, that is, relevant to whether an alternative is morally required, permitted or forbidden for the agent.¹Examples of candidates for inclusion on a list of these properties are being the keeping of a promise and being the preventing of some harm. W. D. Ross is perhaps the best known advocate of this kind of pluralism.²

Theories incorporating this kind of pluralism have an obvious advantage: they reflect the general structure of common-sense morality. For surely one of the most conspicuous features of common-sense morality is that it involves competing moral demands of seemingly different kinds.³ Constructing a theory incorporating property ethical pluralism also faces a number of challenges, two of which will be the topic of this paper.

The first challenge is to provide an account of what makes a property morally relevant. I will argue for an account that is a distant relation to G.E. Moore's isolation test for intrinsic goodness.⁴ The proposal is that what makes a property morally relevant is that it is the kind of property that in some idealized situation always determines the moral status of actions having it. The idealized situation that is the appropriate one is captured by the ceteris paribus caveat. So, roughly, the proposal is that a property is morally relevant just in case, other things being equal, it determines the moral status of any action that has it. Explicating the caveat `other things being equal' in a moral context is the core of meeting the first challenge and will be the main focus of this paper.

There is another challenge to the view that I shall defend. It has recently been emphasised by, for example, Jonathan Dancy.⁵ I am going to characterize a property as morally relevant when, in some idealized situation, it always makes a moral difference. The challenge is to explain how what is true in some idealized situation has any bearing on what is true, or morally relevant, in the ordinary, complicated, non-idealized situations in which we find ourselves. For example, let us say that, other things being equal, lying is morally wrong. Why should we take this to give us any useful information about actual cases of lying in situations where other things are not equal? It certainly does not tell us that we should never lie. What does it tell us about what we should actually do?

I will argue that what properties we accept as morally relevant should not be based on the accident of the situations in which we find ourselves. Instead, it should be based on something quite general, something that I take to be accommodated in an appeal to idealized situations. In addition, I will argue that this sort of generality is required for useful future based thinking about moral requirement.

When are other things equal, morally speaking?

.

Intuitively, other things are equal when there is only one candidate for being an influence on some outcome. By requiring other influences to be excluded, we are trying to isolate the target property. The truth of a statement with an `other things being equal' caveat depends on what happens when this isolation is achieved. In the present case, we are trying to isolate one property from other purportedly morally relevant properties. When a morally relevant property is isolated in an action, it should decide the moral status of that action. I see the main problem of characterizing moral relevance as giving an account of this isolation. However, there is another problem that should be discussed first.

.

Proposals like mine can be viewed as making the following conditional claim about each morally relevant property: if it is isolated in an action, the possession of the property by that action will decide the moral status of the action. What should we take to be the nature of the conditional?

It would be simplest to take it to be the material conditional, with the appropriate quantifiers. But doing so would be a mistake. The truth conditions for these conditionals allow them to be true if the antecedent is never satisfied. If we allow such vacuous truths, it is too easy for a property to be morally relevant. Any property which is never actually isolated in any action would be a morally relevant property. This includes every impossible property, every necessary property, and every property that, contingently, is never had by an action and every property that, contingently, is had by every act. Alternatively, if we were to add a condition to the proposal requiring that the antecedent is satisfied,⁶ being a morally relevant property is made too difficult. Since, morally speaking, the world is a complex and messy place, there may be properties that we want to count as morally relevant that never, as a matter of contingent fact, occur in isolation. It is just for these reasons that the `other things being equal' caveat is thought to point towards an idealized situation.

An obvious way of understanding the conditional so that what is morally relevant does not depend on the contingencies of what actually happens in this way is to understand it as a counterfactual conditional.

Indeed one counterfactual account is fairly prominent, appearing to underlie the views of, for example, W. D. Ross and Frank Snare.⁷This account claims that each actual act that has a morally relevant property could have had it in isolation, and, if it had, its possession of that property would have determined its moral status. It is crucial to note that this is a claim about how each actual action could have been. Consider a possible example. Assume that the only two morally relevant properties are being an instance of not causing harm and being an instance of keeping a promise. That an act involves keeping a promise will not always decide its moral status because some such acts will involve causing a lot of harm. This counterfactual account claims that each act of which this is true could have been one in which promise keeping is isolated, and, if it had been, this fact would have decided its moral status.

An alternative counterfactual account says that a property is morally relevant when some action could have had it in isolation, and, if it had, it would have decided the moral status of the action in which it was isolated. This version differs from the previous one in that it does not require any actual action to be such that it could have had the property in isolation. It just requires that each morally relevant property be such that some action could have had it in isolation.

Although there are particular problems with the first counterfactual account,⁸the very idea of giving a counterfactual account seems misguided to me. For a property to be a morally relevant property, counterfactual accounts only require that it decides the issue in some situations in which it is isolated in an action. Consider the following counterfactual: If the property P had occurred in isolation, it would have decided the moral status of the action that had it (in isolation). For this counterfactual to be non-vacuously true, it need not be the case that all situations in which P occurs in isolation are ones in which it decides the moral status of actions having it. Of all of the possible situations in which P occurs in isolation, the only ones relevant to the the counterfactual's evaluation are those that are most similar to (or relatively similar to) the actual situation.⁹ This would mean that what happens when a property occurs in isolation, but in a situation remote from actual situations, is irrelevant to whether a property is morally relevant. This feature of counterfactual accounts seems unsatisfactory.

My view is that an account of moral relevance should require that for a property to be morally relevant it must be possible that it occur in isolation, and, in every possible situation where it does, it must decide the moral status of the action in which it is isolated. Consequently, I believe that the appropriate way to look at the conditional is as one bound by a necessity operator. So, a morally relevant property is such that, necessarily, if it is isolated in an action, the possession of the property by that action will decide the moral status of the action.

.

When does an action have a property in isolation? It is a situation in which there are supposed to be no rival properties that could affect the moral outcome. Since the absence of all other morally relevant properties would guarantee this, an account that requires this is a natural starting place. Consider the following account of a morally relevant property:

MRP1:

A property, P, is a morally relevant property (MRP) iff P is such that

(1) Possibly, there is an agent, s, and action, a, such that a is available to s, a has P, a has no other MRP and none of s's other alternatives have P or other MRPs; and

(2) Necessarily, for any agent, s, and action, a, if a is available to s, a has P, a has no other MRP and none of s's other alternatives have P or other MRPs, then a's possession of P decides the issue morally for s.

MRP1 has two glaring problems. The first is that it is circular. The second is that it may be that properties that we want to count as morally relevant cannot occur in the absence of other morally relevant properties.

It is circular because it isolates one property in a class of properties by simply stipulating that the others in the class are being excluded. This circularity is not acceptable when the isolation of the property is essential to characterizing membership in that very class of properties about which the stipulation is being made.

Circularity would be avoided by enumerating the other morally relevant properties. For example, instead of saying that the action does not have any other MRP, we could say that the action does not have any of the enumerated properties P1,..., Pn. But adopting this strategy has an unacceptable cost: it would require a commitment to just those enumerated properties as the morally relevant ones. I believe that this would make the account of moral relevance unacceptably specific. It would be better if the account were neutral among views about which particular properties are morally relevant.¹⁰

There is a strategy for eliminating the circularity that I think will work.¹¹Remember, the problem is to isolate conceptually the property under consideration from other morally relevant properties without appealing to the moral relevance of the other properties (circularity) or enumerating them (too specific). We overcome this problem by giving an abstract specification of the class of morally relevant properties.

Instead of directly characterizing properties as morally relevant when actions having them satisfy certain conditions, a general characterization of an entire class of properties is given. We can then say that a property is morally relevant when it is in that class. We do not give conditions for inclusion in the class. We cannot evaluate properties one after another to see if they are in it. Instead, we give some conditions that the entire class must satisfy. It is by the mechanism of considering the entire class that we avoid the circularity problem, while avoiding the problems associated with simple enumeration of the properties. My account will incorporate this.

.

The second problem, however, still needs to be confronted: it may be that no possible action can have one property in the class of morally relevant properties without having some other property in the class. Perhaps some or all of these properties only come in package deals. If there are morally relevant properties like this, they could not occur in isolation as it is characterized in MRP1. While I am not certain that there are such properties, we should take their possibility into account. Unfortunately this will make things a bit more complicated.

We first have to say that all properties come in degrees, from degree 0 to degree 1. An action's not having a property is the same as its having it to degree 0. If an action has a property that does not come in varying degrees, then the action has it to the maximal degree it is possible to have that property: degree 1. Given these, I hope harmless, stipulations, we can compare the degrees to which any two actions have some property. We can now say that the appropriate isolation occurs when one alternative has one of the properties in the class to a greater degree than any of its alternatives and, of the properties in the class under consideration, the alternatives differ only in the degree to which they have that property. This notion of isolation solves the above problem because when two alternatives differ at most in the degree to which they have one morally relevant property, the other morally relevant properties, to put it crudely, cancel one another out.

It may be that there are some (morally relevant) properties that are such that the relative degree to which actions have the properties cannot vary independently. This would mean that in every possible situation where an action has one of the properties to a particular degree (say n), the action also has to have another of these properties to a particular degree (say m). If it turns out that there are such odd morally relevant properties, then I am forced to treat them as the same property. Or, rather, as an essentially compound property. This seems satisfactory since when one occurs to a particular degree this is then a certain sign that the other occurs to a particular degree.

.

We are now in a position to consider my preferred account of a morally relevant property. (Various relativizations to persons and times are suppressed.)

MRP2:: A property Fi is morally relevant iff
: Fi belongs to one of an ordered pair, C, of classes of properties $\langle$ C1, C2 $\rangle$ (where C1 will be the class of positive morally relevant properties and C2 will be the class of negative morally relevant properties)¹²such that the following claims hold with respect to all these properties:
: (1) if a property Fj is in C1 or C2, then, with respect to the other properties in C1 and C2, it is possible that some action a differs from its alternatives only in the degree to which they have Fj;
: (2) necessarily, if an action, a, has some property, Fj, in C1 to a greater degree than any of its alternatives and, with respect to the properties in C1 and C2, a and its alternatives differ at most in the degree to which they have Fj, then a is morally required and its alternatives are morally prohibited (these are positive morally relevant properties);
: (3) necessarily, if an act, a, has some property, Fj, in C2 to a greater degree than any of its alternatives and, with respect to the properties in C1 and C2, a and its alternatives differ at most in the degree to which they have Fj, then a is morally prohibited and at least one of its alternatives is morally permissible (these are negative morally relevant properties);
: (4) necessarily, if alternatives do not differ with respect to the degree that they have some property in either C1 or C2 then they don't differ in moral status.¹³ ¹⁴

Admittedly, this characterization is a bit complicated. However, the idea behind it is not. What it says is that a property is morally relevant when, under certain conditions, it must determine the moral status of acts having it. More specifically, condition (1) requires that it be possible that each morally relevant property be isolable; condition (2) says that when the properties in C1 (positive properties) are isolated, actions having them them are morally required; (3) says that when the properties in C2 (negative properties) are isolated, actions having them are morally prohibited; and condition (4) says that only the properties in C1 and C2 can decide the moral status of actions.

Discussion of the proposal

.

My approach to the understanding of moral relevance and the ceteris paribus caveat is to give the conditions that must be satisfied for other things to be equal. Since these conditions appeal to merely possible situations, it might be thought that it would be an advantage if we could make sense of moral relevance and the ceteris paribus caveat without requiring that conditions be given for when other things are equal. Paul M. Pietroski, for example, offers a view that does not require that they be given. He says that the ceteris paribus clause is to be understood as offering `a ``promissory note'' to the effect that there is a correct explanation for each exception to the lawlike statement it modifies'.¹⁵His account has the consequence that a property is morally relevant if and only if we are committed to there being an explanation of each time its possession fails to determine the moral status of an action possessing it.

Consider the lawlike statement `other things being equal, one ought to keep one's promises'. On my view, promise keeping can only be morally relevant if it is possible that it occur in isolation from other morally relevant properties. Pietroski could account for its moral relevance even if it could not occur in isolation. This is because each time it fails to determine the moral status of an action having it, there is an explanation of that failure.

Nonetheless, I think that my approach is better. Pietroski's view seems to have the consequence that every ceteris paribus principle is true, or every property is morally relevant. This is because when any possession of any property does not determine the moral status of an action possessing it, there is, presumably, some explanation of why it does not. Consider the following lawlike statement: other things being equal, if a person has blue eyes, that person ought to be tortured. Being blue eyed, I hope that this lawlike statement is false. Yet each instance of a person's having blue eyes and not warranting torture can be explained. The fact that torture causes pain would do as an explanation.

.

Since in MRP2 the notions of being morally required and being morally prohibited are used in characterizing moral relevance, I take the former to be conceptually more `basic'. This means that it is not possible to give a criterion for being morally required in terms of something like having a morally relevant property that `overrides' others. Although some might think that this is problematic, I doubt that such a criterion can be given, and if there are necessary and sufficient conditions for being morally required or for being morally prohibited, they will be disjunctive, with each of the disjuncts being so complicated that they are beyond comprehension. Consequently, I regard it as no disadvantage that my account of moral relevance appeals to moral requirement and prohibition.

.

Given my account, there is not going to be a unique set of properties that satisfies MRP2, that is, whose members are morally relevant. There are at least three reasons why this is so. The first reason is that some complex properties constructed out of the `basic' properties in the set will also be in the set. The second is that there may be a variety of equivalent sets that satisfy MRP2. For example, I hold that the making of promises is morally relevant, but could account for this either by saying that that being an instance of keeping a promise has positive moral relevance or by saying that being an instance of breaking a promise has negative moral relevance. Finally, the pair $\langle\{$ being-morally-required $\}$ , $\{$ being-morally-prohibited $\}\rangle$ will, trivially, satisfy MRP2.¹⁶

It would be relatively easy to give conditions that exclude complex properties constructed out of `basic' ones and for choosing between equivalent sets.¹⁷Conditions could also be given that exclude the trivial satisfaction of MRP2 by the properties of being morally required and being morally prohibited. However, I do not see a need for any of these conditions.

If we allowed classes with generated complex properties, we would still allow classes without them. Also, properties such as being morally required and being morally prohibited are morally relevant and it would be a mistake to think otherwise. It is just that the recognition of the moral relevance of the complex generated properties and properties such as being morally prohibited is not very informative. Being a pluralist, I believe that there is another, more enlightening, pair of classes of properties that satisfy the conditions for moral relevance.

Nonpluralists might also want to allow the possibility of there being different sets of properties that satisfy MRP2. For example, consequentialists who hold that being an instance of maximizing utility can satisfy MRP2 could also hold that individual properties that are intrinsically good or intrinsically bad are morally relevant.¹⁸They would be looking at two different pairs of classes of properties. Although the properties in one pair of classes might be more fundamental than the properties in the other, there is no reason to deny that the properties in each are morally relevant. My conclusion about uniqueness is that if it is required, it can be had by tinkering with MRP2. However, it just does not seem to be required.

.

I believe that the fundamental moral principles are principles of prima facie obligation and prima facie prohibition. I will characterize principles of prima facie obligation in terms of the properties that are morally relevant.¹⁹Each of the positive morally relevant properties, the properties in C1 of C, founds a principle of prima facie obligation with the following form:

: If a person, s, can do an action, a, a has a property, Fi, and Fi has positive moral relevance (satisfies the conditions for being in C1 of C of MRP2), then s is prima facie obligated to do a.

Each of the negative morally relevant properties, the properties in C2 of C, founds an analogous principle of prima facie prohibition:

: If a person, s, can do an action, a, a has a property, Fi, and Fi has negative moral relevance (satisfies the conditions for being in C2 of C of MRP2), then s is prima facie prohibited from doing a.

The number of principles satisfying these schemata depends on the number of properties in C1 and C2. Each is a fundamental moral principle: they are indirectly defined in terms of moral obligation and prohibition, which are themselves not defined.

Why does being morally relevant matter?

.

My proposal is an account of moral relevance that is supposed to be a modification of W. D. Ross's view. Jonathan Dancy has argued that it is a defect of Rossian notions of prima facie obligation and concomitant characterizations of moral relevance, such as the one I have proposed, that they do not show how a property relevant in a particular case is generally relevant, and why a generally relevant property must be relevant whenever it occurs.²⁰

If we consider my account, we can put the problem this way: Why should the mere fact that a property determines the moral status of an act in some merely possible situation mean that it is morally relevant to the moral status of any actual act? And why should the fact that a property is morally relevant in some actual act mean that it is relevant in every possible situation in which an act has it? Accounts like mine are supposed to fail to capture something important we want to say about a property's being morally relevant. What is this important something? Dancy, I believe, thinks that theories of moral relevance must be committed to the view that whenever a property is morally relevant it will `make a difference'.²¹ And accounts like mine do not have this as a consequence.

If moral relevance is about making a difference, then one might expect that the idea of actual moral relevance is the idea that a property does make a moral difference and the idea of general moral relevance is the idea that such properties always make a moral difference. Such a view would take actual moral relevance as basic and characterize general moral relevance in terms of it.

.

A type of model that accepts this is what I shall call the `force model'.²² One version of the force model says that diverse morally relevant properties determine moral upshots in much the same way that diverse physical forces affect the trajectories of projectiles: the moral upshot in some situation is a function of the moral `forces' resulting from the particular occurrences of morally relevant properties. Another version accepts that morally relevant properties are, in some sense, like physical forces, but denies that there is any (straightforward) function from these properties (or forces) to what is morally required or prohibited.

One way of attacking models of this type is first to argue that their acceptance rests on establishing some sort of analogy between the roles of forces in moral theory and in the natural sciences, and then to argue that the analogy does not hold.²³ Although I reject the force model, this approach seems inconclusive because it isn't clear just how close an analogy there is, or needs to be, between moral and physical forces. My reason for rejecting the model has to do with what I take to be the nature of the claim that some property is (generally) morally relevant.

Both versions of the force model propose that actual relevance is basic, and say that claiming that a property is generally morally relevant involves some sort of induction on actual cases. This is not how I see things. I take general moral relevance to be the basic notion. The main reason for thinking that this is so is that otherwise the truth of normative theories would be merely contingent. I do not want the fact that being the keeping of a promise is a morally relevant property to depend on the contingent fact that people sometimes make promises. This would mean that if no promises were actually made, then being the keeping of a promise would never have actual moral relevance, and consequently, would not be morally relevant at all.²⁴Since normative theories can be distinguished by what properties they say are morally relevant, if we allow that there could have been different generally morally relevant properties, depending on what properties were actually relevant, we would end up with the possibility that some different normative theory could have been true.²⁵But what normative view, if any, is correct is not generally taken to be a contingent matter.²⁶ One way to avoid the possibility of there being different morally relevant properties at different worlds is to characterize them in the way that I have, which makes true statements of general relevance conceptual claims.²⁷ So my account takes a notion of general moral relevance as the basic one and defines actual moral relevance in terms of it.²⁸

Utilitarians such as G. E. Moore, for example, are not merely claiming that it is contingently true that intrinsic goodness and badness are the most basic morally relevant properties (other than moral rightness and moral wrongness); they are claiming that they are the only properties that can make an act obligatory. Similarly, pluralism, of the sort I would want to defend, claims that there is a plurality of properties that are generally relevant. I need not, as Ross did not, claim that I know what they all are. But whatever they are, it is not a contingent matter that they are morally relevant. It may even be true that some of these properties never determine the moral status of any actual act.²⁹At least nothing I have said requires it. It is enough that they must do so whenever they occur in isolation.

.

Since, according to my account, recognizing that a morally relevant property occurs in a particular situation does not tell us that it makes a difference in that situation, this notion of moral relevance is fairly attenuated. So what good does it do us? It seems to me that without this notion, it would be very difficult, if not impossible, to do any useful moral thinking about the future.

Our thoughts about the future have to be fairly general, because we are not in an epistemic situation that would allow us to know the full details about particular situations that we will face. We cannot imagine future situations in full enough detail to know exactly what will make a moral difference. The contingencies in which we find ourselves are just too complicated. Consequently, moral thinking about the future--for example, planning, prediction and education--all have to rely on our recognition of general features of situations. Recognition of what properties can make a moral difference gives us some of the information that we need for this future based thinking. For example, the reason I try to avoid making promises that I cannot keep is that I know, other things being equal, that I ought to keep each of my promises. That is, I know that the making of a promise is morally relevant to what I should do. Yet I also know that it is possible that I will find myself in situations where I ought to keep few, if any, of the promises that I actually make.

My claim is that the properties on which we focus when we do this future directed thinking are all properties that we think are generally relevant to moral outcomes. They are properties that we think can make a moral difference. It also seems that the very same properties that we invoke in moral thinking about the future are the properties that we invoke in explaining the moral status of present actions and past actions. Consequently, it is hard to see how we could do without some appeal to a notion of general moral relevance, and it is hard to see how we would go about explanation and justification involving present moral situations without appeal to the very same properties that we use in future directed moral thinking.

Conclusion

Providing a complete defence of property ethical pluralism is not an easy task. Many of the important challenges to such a theory have not even been broached in this paper.

However, before the view can be defended, it must be stated. Central to a statement of a pluralistic view is an explication of the notion of moral relevance. I have explicated what I believe to be the appropriate notion: a property is morally relevant just in case (i) it is possible that it occur in isolation and (ii) it must determine the moral status of any action in which it is isolated.³⁰

Christ Church
Oxford OX1 1DP

References

Bibliography

Becker, 1992: Lawrence C. Becker, `Places for Pluralism', Ethics, 102 (1992), p. 707 Becker, `Places for Pluralism'
Dancy, 1983: Jonathan Dancy, `Ethical Particularism and Morally Relevant Properties', Mind, 92 (1983) p. 530 Dancy, `Ethical Particularism
Dancy, 1993: Jonathan Dancy, Moral Reasons (Oxford: Basil Blackwell, 1993) Dancy, Moral Reasons
Gay, 1985: Robert Gay, `Ethical Pluralism: A Reply to Dancy', Mind, 94 (1985), p. 250 Gay, `Ethical Pluralism
Hare, 1981: R. M. Hare, Moral Thinking, Its Levels, Method and Point (Oxford: Clarendon Press, 1981) Hare, Moral Thinking
Kagan, 1989: Shelly Kagan, The Limits of Morality (Oxford: Oxford University Press, 1989) Kagan, Limits of Morality
Lewis, 1973: David Lewis, Counterfactuals (Cambridge, Massachusetts: Harvard University Press, 1973) Lewis, Counterfactuals
Moore, 1903: G. E. Moore, Principia Ethica (Cambridge: Cambridge University Press, 1903) Moore, Principia Ethica
Pietroski, 1993: Paul M. Pietroski, `Prima Facie Obligations, Ceteris Paribus Laws in Moral Theory', Ethics, 103 (1993), p. 489 Pietroski, `Prima Facie Obligations'
Ross, 1930: W. D. Ross, The Right and the Good (Oxford: Oxford University Press, 1930) Ross, Right and the Good
Ross, 1939: W. D. Ross, Foundations of Ethics (Oxford: Clarendon Press, 1939) Ross, Foundations of Ethics
Snare, 1974: Frank Snare, `The Definition of Prima Facie Duties', Philosophical Quarterly, 24 (1974), p. 24 Snare, `Definition of Prima Facie Duties'

Footnotes

... agent.¹: Ethical pluralism, of course, comes in many forms. See [Becker, 1992] for a discussion of its variations.
... pluralism.²: See [Ross, 1930] and [Ross, 1939].
... kinds.³: See [Kagan, 1989] and [Hare, 1981].
... goodness.⁴: [Moore, 1903].
... Dancy.⁵: See [Dancy, 1993], pp 97-98 and [Dancy, 1983], p. 540.
... satisfied,⁶: This additional condition would simply be the antecedent bound by existential quantifiers.
... Snare.⁷: See [Ross, 1930] and [Snare, 1974].
... account,⁸: Despite the fact that the identity conditions for actions in counterfactual circumstances are not clear, it is dubious that every actual action that has a morally relevant property could have had it in isolation.
... situation.⁹: How to rank possible situations by degree of similarity, and what degree is relevant to the evaluation counterfactuals in unclear. For a discussion of these topics, see [Lewis, 1973], esp. pp 91-95.
... relevant.¹⁰: Something like this is also held by Frank Snare, [Snare, 1974], p. 236.
... work.¹¹: I take this strategy from Frank Snare, [Snare, 1974]. Although I reject his approach to characterizing moral relevance in terms of counterfactuals, there is much that is good in his paper.
... properties)¹²: Peter Vallentyne's comments led me to distinguish between positive and negative properties in this way.
... status.¹³: The nature of the necessity operator in MRP2 is not perfectly clear. I am inclined to think that these are conceptual truths, and that the necessity is conceptual necessity.
... ¹⁴: Jonathan Dancy has suggested that this account does not allow for the possibility of a very robust kind of supererogation. At the most, this is so only in situations where a property is isolated in an action. My account says nothing about situations where there is no isolation. Something similar is true of moral dilemmas. My account may preclude them in situations where properties occur in isolation, but says nothing about more complex situations.
... modifies'.¹⁵: [Pietroski, 1993], p. 495. Although he seems to accept the result that other things may never be equal, it is not absolutely clear that he does so (see p. 505).
... MRP2.¹⁶: Thanks to Fred Feldman for pointing this out to me.
... sets.¹⁷: Snare does this. See [Snare, 1974], p. 241.
... relevant.¹⁸: MRP2 might be altered to give something akin to a Moorian isolation test for intrinsic goodness.
... relevant.¹⁹: I am following Ross and Snare here.
... occurs.²⁰: [Dancy, 1983], pp 534, 541. See [Gay, 1985] for one attempt to defend Rossian proposals against these charges. (Dancy also objects that we cannot learn about something's being generally morally relevant from its being morally relevant in a particular case. That issue is quite thorny, but will not be dealt with here. All epistemological questions are being left for another time.)
... difference'.²¹: [Dancy, 1983], p. 540.
... model'.²²: See, e.g., [Ross, 1939], p. 86 and [Pietroski, 1993], pp 506-511.
... hold.²³: [Dancy, 1993], pp 100-104.
... all.²⁴: Note that this is not the weaker claim that the account of moral relevance is true at all worlds, but which properties are actually morally relevant differs from world to world. I would accept this view. The proposal I am rejecting characterizes moral relevance in terms of actually making a difference. So if a property does not actually make a difference, then, according to this proposal, it is not morally relevant, tout court. It seems to me that an action's being such that it would result in the extinction of all life may have moral relevance independently of the pain, deprivation, etc., that it would bring about. This would be a case of a property that never has actual moral relevance, but is, nonetheless, morally relevant.
... true.²⁵: Or, using possible worlds talk, at some other possible world, one where different properties make a moral difference, a different moral theory would be true.
... matter.²⁶: Even those who are error theorists or anti-realists with respect to morality do not often take their claims to be contingent.
... claims.²⁷: Cf. [Pietroski, 1993], pp 512-513.
... it.²⁸: Of course this is compatible with the view that we are first aware of actual moral relevance. I am talking about conceptual priority, not epistemic priority.
... act.²⁹: It is not even required that each of them is ever had by an actual act. My characterization allows for the possibility that there are some that are never actualized.
... isolated.³⁰: A much earlier version of this paper was read at the Wolfson Philosophical Society; I thank the participants for their comments, especially Michael Martin. I also thank Roger Crisp, Jonathan Dancy, Fred Feldman, Penelope Mackie and Peter Vallentyne for their comments on various versions. Finally, John Cottingham and Brad Hooker, of Ratio, also provided me with useful comments.

Robert L. Frazier 2009-10-27