Normative Reasons and the Agent-Neutral/Relative Dichotomy

Abstract

The distinction between the agent-relative and the agent-neutral plays a prominent role in recent attempts to taxonomize normative theories. Its importance extends to most areas in practical philosophy, though. Despite its popularity, the distinction remains difficult to get a good grip on. In part this has to do with the fact that there is no consensus concerning the sort of objects to which we should apply the distinction. Thomas Nagel distinguishes between agent-neutral and agent-relative values, reasons, and principles; Derek Parfit focuses on normative theories (and the aims they provide to agents), David McNaughton and Piers Rawling focus on rules and reasons, Skorupski on predicates, and there are other suggestions too. Some writers suspect that we fundamentally talk about one and the same distinction. This work is about practical reasons for action rather than theoretical reasons for belief. Moreover, focus is on whether reasons do or do not essentially refer to particular agents. A challenge that undermines the dichotomy in this sense is posed. After having rejected different attempts to defend the distinction, it is argued that there is a possible defence that sets out from Jonathan Dancy’s recent distinction between enablers and favourers.

1.

The distinction between the agent-relative and the agent-neutral plays a prominent role in recent attempts to taxonomize normative theories.¹ Its importance, however, extends to most areas in practical philosophy. Despite its popularity, the distinction remains difficult to get a good grip on. In part this has to do with the fact that there is no consensus concerning the sort of objects to which we should apply the distinction. Thomas Nagel distinguishes between agent-neutral and agent-relative values, reasons, and principles; Derek Parfit focuses on normative theories (and the aims they provide to agents), in an early paper David McNaughton and Rawling (1991, 2002) are interested in rules and in a later work they go on to discuss reasons, John Skorupski formulates (personal communication) the distinction in terms of predicates, and there are other suggestions too. Some writers suspect that we fundamentally talk about one and the same distinction (see e.g., Dancy 1993 and Ridge 2005). Here focus will be on practical reasons for action rather than theoretical reasons for belief. Moreover, I will be concerned with one particular way of understanding the neutral/relative dichotomy, namely one that centres on whether reasons (or more accurately, reason statements) do or do not essentially refer to particular agents. I will argue that this sense is open to a challenge. The structure of this paper is as follows: After some introductory notes in “Section 2”, I pose in “Section 3” the challenge that I believe undermines the dichotomy. I then consider some possible replies to it in “Section 4”. In “Section 5” I turn to a more positive argument to the effect that all reasons are agent-relative. Finally, in “Section 6” I suggest a defence of the dichotomy that sets out from Jonathan Dancy’s recent distinction between enablers and favourers.

2.

The philosophically more interesting sense of the distinction was formulated by Nagel in The View from Nowhere:

If a reason can be given a general form which does not include an essential reference to the person who has it, it is an agent-neutral reason…If on the other hand the general form of a reason does include an essential reference to the person who has it, it is an agent-relative reason (Nagel 1986, 152–3).

Nagel draws the distinction in terms of what does and what does not essentially refer to a particular person, viz. the “owner” of the reason. Various other writers have employed similar accounts. They have also tried to be more specific about what is required for a reason statement or form to qualify as agent-relative or neutral. Philip Pettit, for instance, describes an agent-relative reason as “one that cannot be fully specified without pronominal back-reference to the person for whom it is a reason” and an agent-neutral one “as one that can be fully specified without such an indexical device”.² Several other writers have suggested similar accounts, which stress that the distinction concerns reasons that do or do not contain ineliminable back-reference to the agent for whom it is a reason. (Cf., Portmore 2001; McNaughton and Rawling 2002; Ridge 2005).

We cannot, of course, be sure that Pettit’s and other similar approaches to the dichotomy fully coincide with Nagel’s. It is, for instance, a matter of interpretation when a statement is “fully specified”. Notwithstanding, this concern about indexicals and ineliminable back-references reflects these authors’ effort to express (what Nagel was driving at, namely) that the distinction is about reasons (or principles, rules etc.) that do and do not involve essential reference to the agent.³ How we do this in our language is an interesting but nonetheless secondary matter; the real issue is still what makes these different kinds of reason statements true.⁴

The “essentialist” approach to the distinction is the interesting one in that it tries to unravel a conceptual difference (one which among other things is thought to help us distinguish between different sorts of moral theories). It focuses on one particular feature, viz., whether or not a reason statement necessarily refers to the agent for whom it is a reason. What is important to realize, then, is that the distinction is silent about the range of agents for whom something is a reason. It leaves this matter quite open.⁵

Despite the fact that there is no general agreement⁶ concerning how to formulate the “essentialist dichotomy”, it is at least possible to express in a few words what seems to be the core idea behind it. Let P refer to a ‘reason statement’, i.e., a statement expressing a reason. Nagel, Pettit, and others make this move from reasons to “forms” or statements expressing reasons, and it simplifies the presentation if we can sometimes discuss reason statements rather than reasons.⁷ Thus, the idea is the following—where x (and later y) refers to an agent, and Φ to some act:

If P states a reason for x to Φ, then:

P states an agent-relative reason for x if and only if

P contains an essential reference to x,

Otherwise, P states an agent-neutral reason for x to Φ.

Moreover,

P contains an essential reference to x if and only if

P is not logically equivalent with any other statement Q that does not refer to x.

The following statement expresses, on the essentialist approach, an agent-neutral reason for x (in case x ≠ y): (Pn) ‘Person y is drowning’. The fact that someone is drowning constitutes a reason to save y that is not agent-relative but agent–neutral. The statement (Pr) ‘My daughter y is drowning’, expresses in its turn a typical example of an agent-relative reason; it gives y’s parents a reason to save y, that is different from the reason, say, that I who am not the father of y, have to save her.

The above picture is oversimplified, though. The “relative/neutral” status of these reason statements, Pr and Pn, seem to depend on whom they are addressed to. For instance, if Pn is directed to person y, then Pn should be seen as stating an agent-relative reason for y. And while Pr may state an agent-relative reason for y’s parents, it would in many people’s view also state a reason for other people than y’s parents, since it seems to be an agent-neutral reason that we should save people from drowning. This is quite confusing and it is one of the sources of why the distinction is so tough to get a hold on. That is, given that we have in mind a certain kind of reasons—which I will get back to later on—it is clear that whether a reason statement expresses something agent-relative or agent-neutral depends in part on to whom it is addressed. On the other hand, if the truth of a reason statement is not relativized to how the person to whom it is addressed understands the statement then the above confusion concerns something else, namely what people take to be reasons. This distinction between apparent and real reasons, as I will refer to it, will occupy us in a while.

3.

In order to approach the question whether there are kinds of reasons that might run into problems (i.e., not admit to be divided in accordance with the neutral/relative dichotomy) with regard to the essentialist way of understanding the agent-relative/neutral reasons dichotomy, it will be convenient to proceed by raising a challenge to the dichotomy. It sets out from a feature of practical reasons—what I will refer to as the Personalizability-feature (PEF)⁸—that has apparent relevance for the distinction in question, viz. that all reasons are never only reasons period; they are always reasons for x to Φ.

That reasons are always reasons for someone seems to have clear consequences for the agent-relative and agent-neutral reason distinction. The challenge sets out from the following observation: (PEF) Since all reasons apparently are reasons for someone and a reason is only a reason for someone, if it somehow involves or refers to this someone, it follows that all reasons are by their very form reasons that refer to the person who has the reason to Φ. This, in its turn, is just another way of saying that all reasons are on entirely formal grounds, agent-relative reasons.

The challenge then undermines the dichotomy not by suggesting a kind of substantive argument (say, from a prudential or egoistic perspective) against agent-neutral reasons. The challenge implies something else, namely that the very distinction does not make sense, since all reasons in virtue of being reasons for x have to be agent-relative. The question is of course in what sense it is true that “a reason is only a reason for someone, if it somehow involves or refers to this someone”. In particular, I will be focussing on how it should be understood in terms of the essentialist approach. As I shall be arguing, the challenge does in effect pose something of a threat.

First, a clarification: The neutral/relative reason dichotomy is usually applied only to normative rather than to explanatory reasons, where the former ones are reasons that justify rather than explain why an agent does some act. Here too focus will be on normative reasons. The precise nature of these is a much debated issue that I will not get into here. However, I will suppose that they are facts of some kind—which is in accordance with how many influential writers regard reasons these days.⁹ For the time being, the nature of these facts may also be left unspecified—they may be natural or, say, evaluative (moral). However, it is important that facts should be understood as entities that are independent of what people believe to be the case (facts about people’s beliefs are no exception; if such facts are reasons, they are so independently of whether they are believed to be the case). This is also in line with numerous views on normative reasons, qua facts.

The idea that all reasons are reasons for an agent does appear to create a problem for a view according to which reasons are facts, i.e., states of affairs which obtain (or even better, obtaining states of affairs).¹⁰ However, before explaining why this is so, a caveat is in place; it is important not to mix up the notion of a normative reasons with the related notion of an “apparent reason”.¹¹ The latter refers to that state of affairs which as a matter of fact appears to an agent as a normative reason. Few are ready to say, though, that just because we believe that something is a reason, by this very fact it is also a normative reason. Only on a highly controversial metaethical view (a crude sort of relativism) are apparent reasons constitutive of normative reasons.¹²

The distinction between (normative) reasons and apparent (normative) reasons is important. The apparent kind appear to be quite consistent with the dichotomy—that is, the beliefs which the agent has about what are his or her reasons need not always take a form that essentially involves the agent. There is no good reason to deny this. Nothing necessarily prevents an agent x from regarding, say, the alleged agent-neutral fact that there is a person y drowning to be a reason for him or her to try to save y. But again, this does not establish that there are agent-neutral normative reasons—it indicates that the agent believes there to be in some sense (that is different from the essentialist notion) agent-neutral ones.

Let us return to the challenge and the question whether normative reasons, qua facts, are somehow difficult to understand in terms of the agent-relative/neutral dichotomy? To examine this issue, recall first what it would be for a fact to be an agent-relative reason on the essentialist view; it would be for some obtaining state of affairs to involve the agent for whom the fact is a reason. We know this, since a reason statement, Pr, expressing this state of affairs would not be logically equivalent with any other statement that does not refer to the agent. Suppose what Pr expresses to be obtaining is, among other things that the person who is drowning is my daughter. Here it is quite evident which feature that makes this fact into a reason for me, viz., that it is my daughter who is in need of help. An analogous reply does not work when it comes to agent-neutral reasons, qua facts. On the essentialist approach, given the truth of an agent-neutral reason statement, Pn, there is a state of affairs obtaining such that it is a reason for x, and, moreover, it is a reason for x despite the fact that the state of affairs is not in any way about x.¹³ The question “why is this fact a reason for x” cannot therefore be answered by pointing at some feature of the fact that even in a minimal sense concerns x. This is quite remarkable. The conclusion which seemingly is forthcoming is that there is nothing about the fact that makes it a reason for x, but it is nonetheless a reason for x.¹⁴

4.

There is something troublesome with this picture of normative agent-neutral reasons, qua facts. But it might be objected that this is a much too exaggerated worry. Why not simply say that in certain cases, the fact is that everyone has a reason to do the act, and since, say, I am a member of “everyone”, I have a reason to do the act? But this reply is question begging; it will not do if we are interested in the essentialist definition of reason, qua facts. If, say, the fact that a person y is drowning gives me and everyone else a reason to save y, the reason why I should save y is not that everyone has a reason to save y but that it is a fact that y is drowning. The number of people who have this reason is not a reason-making feature. This fact (that y is drowning) may well be such that it gives everyone a reason to perform the act. This is what is in fact under discussion here. If it is a reason for me, there must be some feature of the fact that in a minimal sense concerns me, for its being a reason for me to save y (so at least goes the challenge). Here, to begin with, are two replies that question this line of reasoning:

• (i) Normative reasons need not be reasons for agents.

• (ii) A state of affairs may be about x and still be an agent-neutral reason.

Now, (i) amounts to a simple refusal to accept the challenge. This would not be an unexpected response. After all, I have not offered any argument in support of the observation PEF, from which the challenge sets out, at least not one other than that there is an obvious ring to PEF. And as is well known, one person’s modus pones is another’s modus tollens. But perhaps there is in effect a suggestion that would help break this stalemate, one, moreover, which would leave the distinction intact. The suggestion has it that the question “why is this fact a reason for x?” is not always legitimate (as is it would be if all reasons are reasons for); in the case of agent-neutral reasons, they are not reasons for but “merely” reasons. Accordingly, it would be to misunderstand these reasons to ask “what is it about this fact that makes it into a reason for x?”. In analogy with a well known distinction in value theory, namely the one between good period and good for, and value period and value for, reasons might just be reasons period but not reasons for.¹⁵

But this “analogy-reply” is unconvincing. Reasons are not like values in this sense; reasons, qua facts, do not just “call for” an action, they call for an action by those who are able to perform the action. That reasons can be altogether agent-unrelated leaves us with a too mystical view of reasons. If a fact is a reason it is a reason for at least one agent. That something might merely be a reason period but not be a reason for any present, past or future agent is, to say the least, a strange idea.

However, the question is if there is not a rather obvious objection to this reply. Wlodek Rabinowicz has suggested (personal communication) that there might be examples that underwrite the claim that an action ought to be done by someone, without underwriting the claim that it ought to be done by someone in particular. For instance, if there are too many people in a boat, someone ought to leave it. This seems to be an example of a reason for someone that is not a reason for a particular person. This does not mean that everyone ought to leave the boat; it means precisely that someone but not everyone ought to leave the boat.

The example is indeed interesting. But what it shows is quite an open matter. For instance, even if we accepted that we here have a reason for someone that is not a reason for any particular person, this does not automatically qualify it as an agent-neutral reason—at least not on the essentialist approach. Here is why: The statement which expresses the fact of the case, namely that in a boat with persons, say, a, b, and c, someone has to leave, will not be logically equivalent with any other statement that does not mention a, b, and c. So the conclusion seems to be that we have here an odd example of an agent-relative reason, namely one which is a reason for someone but not for a particular agent. However, even this conclusion might be questioned. It appears at least quite counter-intuitive to regard the example as in the first place being about reasons to act. That the boat is overcrowded is a reason for there being someone who leaves the boat. However, this is hardly describable as a reason for an act. A reason for there being someone is not about an action, even though leaving the boat is. It is therefore unclear just what this example shows. If the fact about the boat is a reason, it seems most plausible to be a different kind of reason than reasons for action. Accordingly, even if we concluded eventually that there are no reasons for action that are not agent-relative, we should acknowledge the possibility of there being “no particular person” reasons.

It is always logically legitimate to ask what it is about some fact that makes it into someone’s reason for action. And in the case of a so-called agent-neutral reason, the test is how does P (which expresses this fact), that allegedly contains no reference of any sort to x, express a reason for x to Φ? What is it about the expressed fact that explains why it constitutes a reason for x? What is it that the “for” in “This fact is a reason for x to Φ” signals with regard to the agent, when it comes to agent-neutral reasons? Response (i) should be rejected. Accepting it is tantamount to admitting that reasons for action may quite literally be no one’s reason, which is incomprehensible.

The second response, (ii), suggests that the step from “this fact is (among other things) about x” to “this fact cannot be an agent-neutral reason” is in effect invalid. Facts can be about a certain agent, and nonetheless be an agent-neutral reason for this very agent.

There might be something to this reply. However, as far as the essentialist approach goes, (ii) is a non-starter; it follows from the definition of what is a statement that essentially refers to an agent (see above, the end of “Section 2”) that if a fact is a reason, and it is (among other things) about person x (whether or not it is also about every other agent, y, z…), then the statement expressing this state of affairs, P, will not be logically equivalent with some other statement Q that does not refer to x. Hence, P will contain an essential reference to x. That is, P will be expressing an agent-relative reason for x. In other words, if reply (ii) is to have success it must provide a different meaning than the essentialist one or at least amend it in some suitable way.

It might be pointed out that not all features of a fact are equally relevant. So although a fact may to some degree be about the person for whom it is a reason, this agent-relative-feature is not a reason-making property. And accordingly, the statement expressing the fact, need not mention this agent-relative feature. This might actually be a way of meeting the challenge. In a while I will consider a reply that might be regarded as a version of it that deserves to be taken into a more detailed consideration. However, on the whole, this reply is not obviously compelling; it tries to solve a problem of how to clarify a conceptual issue by introducing a substantive requirement on, in this case, what features of facts may or may not be relevant reason-providers. At best it will convince people who share the same normative outlook, i.e., who agree to the same normative statements. As far as possible, answers to formal questions should not depend on substantive views. Initially we ought therefore to hope for more.

If reasons are all reasons for someone, then the challenge is, in other words, to find what it is about the fact that explains why it is a reason for the agent, if this fact does not at all involve the agent. If there is no such feature, then we are in want of an explanation of why the fact is a reason for the agent. Of course, we might not know or be unsure what the feature is. We might also question that the fact is in effect a reason in the first place. But this is not what is at issue here. The question is why is this reason a reason for x? To maintain that a fact which does not involve person x, might still be reason for x is to admit that there is no explanation for why the reason is a reason for x. That is, if x is not involved in the fact, it becomes inexplicable that the fact is a reason for x.

Perhaps we should accept that certain facts are inexplicably reasons for certain agents. It could then be said that the set of inexplicable reasons coincides with the set of all non-agent-relative reasons? Another possibility is, of course, to maintain that all reasons are agent-relative reasons.

But there are other replies to the challenge. Maybe it rests on an implausible interpretation of the expression “reason for x to do Φ”? Surely something could be a reason for x to be Φ-ing even if the reason failed to “get a grip on” x. This would then suggest that something could be an agent-neutral reason for x to do Φ despite the fact that x would not be Φ-ing even if he were aware of the reason. On such occasions, the reason would still be a reason for x.

This reply leaves the challenge in one piece, though. That something is a reason for x does not, as might be recalled, require that x is somehow aware of x. In fact, that some fact F is a normative reason for x to be Φ-ing is not conditional on its being the case that x would do Φ if he believed that P. That is, the latter might be true of certain facts, but it would be implausible to maintain that this is always the case. Sometimes two persons may have the same beliefs about what are the facts, and still not share the same normative views. More than that: Even if I know a certain fact and know that this fact is a reason for me to do Φ, I might still not do Φ. Moreover, that there are many facts unknown to all of us is something that is generally accepted. In light of this it is reasonable to expect there to be many normative reasons that are unknown to all of us. That reasons are always reason for x should in other words not be understood as implying that x must somehow be aware of or even respond to the reason. The expression “reason for” establishes that there must be something that relates the fact to the person; but the relation in question need not be epistemic. If there is no relation between the agent and the fact then it becomes perplexing how the fact can be a reason for x. If this contention is correct, there is one sense of “reason for x” that appears to be difficult to understand in terms of a normative agent-neutral reason.

But why not say that certain facts involve certain individuals essentially, while other facts only involve people contingently? This looks like a straightforward solution. But it remains, of course, to be shown what sense can be given to the locution “this obtaining state of affairs contingently involves x”. Metaphysicians at least tend to agree that propositional (abstract) entities like states of affairs have all their elements necessarily. As far as I can see, “obtaining states of affairs” are not any different in this respect.

There is the possibility that the challenge sets out from a narrow view of facts. Perhaps there are, say, facts of the following kind: ‘that one ought to save people from drowning’. This does seem to be an agent-neutral reason. It might appear so. However, consider the expression “one ought to…”. Suppose we take it to mean (1) “everybody ought to…” or (2) “regardless of who you are you ought to…”, and by ‘everybody’ in (1) or “you” in (2) mean a determinate set, S, of people, namely every person there is at the time this state of affairs obtains. In that case the reason statement expressing this particular state of affairs can accordingly be expressed as follows: “Each of the persons x, y, z…(i.e., every member of S) ought to…”. Understood in this way, it is evident why this reading will not serve as a way around the challenge. The statement is not logically equivalent to one in which, say, there is no reference to x. Since it is not, it is clear that this fact is not an agent-neutral reason—on the essentialist approach.¹⁶

5.

In all its simplicity, the challenge (from PEF) seems to be a hard nut to crack. It sets off from two ideas: First, that claims such as “This is a reason” and “The reason is…” are, at least in the case of normative reasons, elliptical—they are all shorthand for “This is a reason for x to Φ” or “The reason for x to Φ is...”. Second, if normative reasons are facts, then those facts that are reasons must somehow reflect or explain this “for x” feature of reasons—otherwise it is inexplicable why those facts constitute reasons for the agent. The challenge for adherents of agent-neutral reasons is then to find a fact that does this without involving the person for whom it is a reason.

The challenge has not established that this is impossible. In fact, a further reply quite correctly points out that that since all that has been shown so far is that certain replies do not work, this should call for caution. This is, of course, a sensible warning. However, there is in fact also a strong positive argument (which already has been invoked in this work) for why all normative reasons are in fact agent-relative. It centres on a generally accepted feature of practical reasons, namely what sometimes is boiled down to the essential (R) “Reason to act implies ability to act”.

Suppose, then, it is true that to have a reason to Φ you have to be able (physically and perhaps psychologically) to Φ. Furthermore, suppose that this “ability constraint” on reasons is valid for all practical reasons. This restricts in its turn the kind of facts that constitute reasons: Fact F is a reason for x to Φ, only if F involves at least in part x’s ability to Φ. A full description of the fact (that is a reason) would in other words mention among other things that the agent is able to act.

So if the credo R is indeed a formal feature of reasons (or more accurately of reason statements), it seems to a fortiori strengthen the claim that there are no agent-neutral normative reasons (This somewhat cautious “it seems” claim is warranted, as we shall see in a moment).

It might be argued, perhaps, that the “possibility constraint” is an uninteresting, trivial or somehow disregardable feature of reasons. And that we may or should divide reasons in different groups without paying any notice to this shared feature. This might be true, but it does not change the fact that these groups would exclusively consist of agent-relative reasons; you cannot uphold the reason dichotomy discussed here if you believe in the doctrine that practical reasons imply ability or possibility to act. Or so it seems.

6.

Jonathan Dancy’s idea that we should distinguish between reason considerations that are “favourers/disfavourers” and those that are “enablers/disablers”¹⁷ suggests how the positive argument might be refuted.¹⁸ Favourers, we are told, are features (facts) that count in favour of acting in one way or another. This cannot be said about enablers, though; they do not favour acting. Instead, what they do is to clear the ground for the favourer—they make it possible for a certain feature to count in favour of acting. In short, the enabler allows the favourer to do its job. Here are Dancy’s own examples: “(1) I promised to do it. (2) My promise was not given under duress. (3) I am able to do it. (4) There is no greater reason not to do it”.¹⁹ Of these (1) is a favourer and (2–4) are different kinds of enablers.

The distinction between favourers and enablers comes in handy; even if reasons imply ability it does not follow that all reasons are agent-relative. A way of describing an agent-neutral reason would then be to say, that it is some fact that on condition that it is enabled by the fact that the agent is able to do the act, favours (The precise nature of this fact that favours on condition may be left open with one exception, namely that it must not involve the agent x for whom it is reason. Otherwise the challenge would not be met).

This is a plausible response to the positive argument. However, just how Dancy’s distinction should be understood in more detail is still very much discussed. Joseph Raz, for instance, finds “the category of being an enabler /…/, so diverse as to be of little use, and likely to mislead.”, (Raz 2006, p. 106), and more generally he detects a number of tensions in the distinction between favourers and enablers.²⁰

Admittedly, Dancy’s distinction is fundamentally a phenomenological one, and so it should not come as a surprise that it gives rise to interpretative tensions. Notwithstanding, as far as I can see the idea that not all “reasons features” play the same role paves the way for a defense of the neutral/relative dichotomy which is by far the most promising I am familiar with. Notwithstanding, I want to voice two worries. The first one has to do with the fact that Dancy takes reasons to be contributory reasons. I am prone to believe there is at least another kind of reason, namely something that we could refer to as a complete reason—something that resemblances Ross’s idea that ‘duty is toti-resultant’—and I suspect that such reasons that are individuated by “all” the facts (i.e., the obtaining states of affairs at time t) are still open to the challenge. Dancy does comment on Ross’s idea. Thus, he says “Ross lumps all relevant features together, as part of the right-making base (the ground, as one might call it), and this metaphysical picture of the situation is far too indiscriminate” (Dancy 2004, p.45; see also pp. 91–92).²¹ Dancy’s choice of term here (“indiscriminate”) is noteworthy. As “indiscriminacy” is not a feature of the world but more accurately of how it is described, Dancy is here most likely concerned with descriptions or explanations of reasons. His objection to Ross’s idea is therefore not one that immediately bears on the notion of reason that I have been examining (namely the idea of reasons qua facts). We should certainly agree that it is not very helpful to start enumerating all the relevant facts if we want to explain, cite or (more generally) describe a reason to someone. But this is quite consistent with the idea that there is one and only one thing I have a reason to do at t1, and that this is determined by how the world precisely is at t1.

Not that we (should) ever aspire to take into consideration all there is to know about the world in the moment we think there is a reason to do something. Surely that would be preposterous. The world is epistemically muddy, our knowledge of, say, what our actions will result in is fallible. But if at every given time there is a set of (relevant and most likely endless) facts, and it is true that what we have reason to do at least sometimes is determined by not a single fact but by a number of facts, then it is hard to see how we can rule out the notion of a complete reason on purely conceptual grounds. That we cannot aspire to know what facts constitute a given complete reason is something we have to live with. But it does not prevent the notion of a complete reason to play some role, though.

But, it might be objected, even if there are complete reasons in this sense, couldn’t we still differentiate between those complete reasons whose favourer(s) do not involve the reason holder and those that do? And so couldn’t we just say as before that the first kind was an agent-neutral one, and the second an agent-relative reason? But this response is not really thought through. A complete reason is composed by all the (relevant) facts, and these do not constitute a favourer (of their own)—at least not in what I believe to be Dancy’s sense, viz., a favourer on condition. But, more importantly, even supposing that the complex fact constitute some sort of favourer, such a (what might be referred to as a complete or unconditional) favourer, might well be constituted by facts that on their own constitute contributory reasons that favour other things or even disfavour, say, the act that the complete favourer (reason) calls for. The correspondence between the two kinds of favourer cannot at least be assumed to be necessary without any argument. It is therefore not obvious how the notion of a favourer on condition applies to the idea of a complete reason. Hence, the “favourer/enabler” suggestion is therefore not really helpful when it comes to one kind of reasons.

So although we may still among the features constituting a complete reason tell apart those that are favourers and those that are enablers, we cannot obviously use this to differentiate between agent-neutral and agent-relative complete reasons. It remains unclear what the relationship is between these favourers/enablers (that constitute different contributory reasons), and the complete reason. So even if a given complete reason will in a sense encompass, say, a number of different favourers (in Dancy’ sense), some of which will involve the agent and some which will not, it is not clear what the relation is between these favourers and the complete reason. Hence, it is unclear why the nature of a complete reason should eventually be determined by a certain favourer rather than some other one.

The above point needs to be amplified, though. We should first distinguish between two kinds of complete reasons both of which can be characterized as being individuated (and thus constituted) by the facts at a certain time. Thus, it might be that the complete reason at t1 is (i) constituted by all the obtaining states of affairs at t1. This would be one way of conceiving what it would be to exemplify a complete reason. However, it is hard to think that there ever would be such a complete reason. Among other things, it would mean that at a given time all facts were relevant, which is hardly imaginable. A more realistic idea is that a complete reason is (ii) constituted only by all the relevant facts. However, since we are dealing with the notion of a complete reason here, one of these facts would have to be a somewhat peculiar fact, namely the fact that all other facts are irrelevant, i.e., they have nothing to do with the reason. The justification for this proviso is then the following observation: a reason at t1 would hardly exemplify the notion of a complete reason qua facts if there were further facts that might alter or have an influence on what we have a reason to do at t1.

A complete reason of the “all the relevant facts” type or of the former (more implausible) (i)-kind, should not be understood as the reason that results from weighing different (contributory) reasons against each other. Rather, the idea is that the (relevant) obtaining states of affairs at a given time all together equally constitute the reason. The following analogy might be helpful: A jigsaw puzzle J1 (displaying, say, some prescriptive instruction) consists of equally shaped pieces that together constitute an image. Another jigsaw puzzle J2 (also displaying, say, some prescriptive instruction) is not really finished—there remain some pieces that we do not have access to. Still, J2 contains enough pieces for us to give us a good idea of what it would look like (or what is prescribed). However, J2 is made up by pieces that are all differently shaped. Since the pieces are differently formed we might want suggest that some pieces play roles that other pieces do not (e.g., some are better to begin with than others in that they enable you to see, say, the corners). This would not make sense to say about J1, whose pieces are all uniform.

Now although we never as a matter of fact seem to be in a position to say with certainty what a complete reason is, such a notion of a reason is not pointless. For instance, it may well be what best explains our motivation to look for more facts even if we think we have a contributory reason for doing a certain thing. We are not, at least not always, just looking for yet another contributory reason. We are looking for (what best approximates) the reason, i.e., the reason constituted by the complete picture.

Now, my first worry is, then, that such a complete reason squares badly with the idea that reasons are favourers on condition. Each fact does its equal reason constitutive share (cf., jigsaw puzzle J1). The explanation why it does not seem to be a favourer on condition, is then that a complete reason does not in any obvious sense have (or more accurately, does not admit of having) an enabler. On the notion of a complete reason outlined here, any fact will either be (1) a fact that together with all other facts equally constitute the reason (which would be the case if “complete reason” were understood in the unrealistic sense that all facts were relevant) or (2) a fact that is either relevant or irrelevant, which would be the case if we read “complete reason” in the sense of (ii). Since it is not obvious how one and the same feature can be constitutive of a reason and be an enabler for this very same reason, there is ground to be concerned that this sort of reason notion is still open to the challenge from PEF.²²

Now, Dancy discusses at length some reason notions that differ from his own “contributory reasons”.²³ In particular his discussion of Joseph Raz’s notion of a complete reason is of interest:

[o]n Raz’s account of a complete reason, whose parts (which he thinks of as premises, and which I would think of as reasons, some of them) are not reasons, it will be true that whatever is a reason is always a reason, and always on the same side. (Dancy p. 97)

Dancy takes Raz to endorse (what Dancy calls) “cluster atomism”, and it is really this idea that is bothering Dancy. A cluster atomist not only regards (at least) some reasons as clusters of features, he or she also maintains that “a cluster that plays a certain role in one case, must play that role wherever it appears” (p. 97). This means, then, that if something is a reason for x in a certain situation, it will be a reason for x in any situation. This Dancy then takes to be inconsistent with his own view about reasons, holism, which is the idea that a feature which is a reason in one case may be no reason or even be a reason against in another case.

Now, the atomism–holism issue is an important one. However, thinking that there are complete reasons in the sense I have outlined here does not commit you to either atomism or holism. In fact, it might quite plausibly be argued that the issue does not arise in the first place, the explanation being that what is a complete reason at one point of time will not be the complete reason at any other time (i.e., given the plausible assumption that the set of facts changes over time). In fact, in my view we should follow Dancy and reject atomism. That is, I share Dancy’s view that one should reject atomism when it comes to contributory reasons. But not all reasons seem to be contributory. As far as I can see, Dancy’s objections to “complete reason” do not affect an idea such as the one outlined above.

The distinction between favourers and enablers is important. But even if we accept it, which I think we should, we need not accept that all reasons are favourers on a condition. We must make room for the possibility of a non-contributory kind of reason, namely a complete reason. Since such a reason cannot be depicted as a favourer on condition we cannot avail ourselves of Dancy’s distinction to meet the challenge.

The atomism/holism notions apply more naturally to apparent reasons, i.e., to what we believe are our reasons rather than to what reasons are. This brings me, then, to my second worry, which concerns an idea that several writers have had, namely that Dancy’s distinction is fundamentally a pragmatic rather than a metaphysical one. Favourers and enablers are notions we employ when we cite or explain what we think our reasons are. Something that is a favourer in one context might then be an enabler in another depending on what we are interested in emphasizing.²⁴ Raz, for instance, is ready to read Dancy along these lines: Consider for instance the following:

/Dancy/ protests against the claim made by Crisp (2000, p. 44), that citing a favourer may be good enough an explanation of one’s reason for acting as one did (pp. 47–48 and 1pp. 95–97). That seems to me true since explanations can have different objectives as well as different objects. Not infrequently citing the so-called favourer is, given one’s interlocutor’s interest, the best explanation of one’s reason. But this only illustrates the importance of distinguishing between the explanation of a reason and the reason itself. The notion of a complete explanation probably does not make sense. There can always be additional puzzles calling for additional explanations. It does not follow that the notion of a complete reasons does not make sense. One should not conclude from the fact that human questions have no end that how things are in the world is equally open ended and in the same way (Raz 2006, 110; my italics).²⁵

If there is something to these suggestions then the distinction between favourers and enablers concerns what reason considerations can do rather than what facts (features) can do.²⁶ The distinction has more to do with how we explain (our) reasons, than with what (our) normative reasons are. What an explanation will look like in terms of favourers and enablers might vary depending on, say, to whom the explanation is directed (e.g., a child or an adult; cf., Raz 2006, p.101). Since the challenge concerns what reasons are rather than how we explain them, it might be argued that Dancy’s distinction will not help us defend the neutral/relative reason dichotomy. So if these suggestions are correct, the favourer/enabler distinction lacks any metaphysical base; hence, “enability” and “favourabilty” most likely are but features we ascribe to facts depending on what appears to us as salient given our desires and interests, at the time. These ascribed features are then better characterized as belonging to apparent (normative) reasons rather than to normative reasons qua facts independent of our beliefs.

Still, I am not certain that Dancy’s distinction is, or more importantly, must be contextual in this sense. Perhaps it need not be. However, it is difficult to see what would eventually settle this issue. And so it is best to be cautious about whether Dancy’s distinction can be used to defend the dichotomy. Notwithstanding, Dancy has supplied us with the tools for illuminating it.