* I am grateful to: David Alm, Johan Brännmark, Fritz-Anton Fritzson, Wlodek Rabinowicz and Caj Strandberg, for their useful comments. I would also like to thank Lars Bergström, whose comments (on another paper of mine) forced me to think more about the issues discussed in this work. This is a revised version of my paper of the same title, which was published in Frans Svensson and Rysiek Sliwinski (ed.)

Las razones son hechos, es decir, están constituidas por hechos. Este planteamiento de "razón qua hecho" tiene mucho respaldo en la literatura reciente. El presente artículo aborda algunas cuestiones que surgen cuando aplicamos esta idea a la distinción entre razones normativas neutrales al agente y razones normativas relativas al agente. Principalmente consideraré dos puntos de vista sobre la naturaleza de los hechos. Según un punto de vista popular, que concibe los hechos como entidades abstractas, la dicotomía neutral-relativa no suele considerarse necesariamente problemática.

Así, en el llamado enfoque fino, es posible que algunas razones qua hechos sean neutrales al agente y algunas sean relativas al agente. En un segundo enfoque, menos popular, el llamado enfoque grueso, los enunciados de razón se refieren a entidades concretas, i.e. son entidades gruesas más que aspectos finos de tipo proposicional las que constituyen razones.

En el desarrollo del artículo, una vez que examinemos algunas de las aparentes ventajas del enfoque fino, argumentaré que resultan problemáticas. Pero, más importante aún, a pesar de algunas ventajas más o menos evidentes, el enfoque fino es inferior al enfoque grueso, por lo menos en algunos aspectos importantes. En primer lugar, el enfoque fino trivializa el debate sustancial sobre si existen razones neutrales al agente o razones relativas al agente. En segundo lugar, se argumenta que el enfoque fino implica que el agente enfrenta un sinfín de razones; si no se puede resistir esta implicación, el enfoque delgado resulta estéticamente menos atractivo que el enfoque grueso. Sin embargo, el enfoque grueso no es perfecto y al final discutiré un par de objeciones al mismo.

Razones neutrales al agente, razones relativas al agente, razones completas, razones normativas.

Reasons are facts, i.e., they are constituted by facts. This "reason quo fact" claim is much endorsed in recent literature. This paper addresses some issues that arise when we apply this idea to the distinction between agent-neutral and agent-relative normative reasons. I shall mainly consider two views of the nature of facts. Given a popular view, which conceives of facts as abstract entities, the neutral-relative dichotomy is often regarded as not being particularly problematic.

Thus, on this so called thin approach, it is possible that some reasons qua facts are agent-neutral and some are agent-relative. On a second, less popular approach, the so-called thick approach, reason statements refer to concrete entities; i.e., it is thick entities rather than thin propositional-like features that are constitutive reasons.

In the course of the paper, once we examine some of the apparent advantages of the thin approach, I shall argue that they come out as problematic. But, more importantly, despite some more or less obvious advantages, the thin account is inferior to the thicker account in at least some important aspects. First, the thin approach trivializes the substantial debate on whether there are any agent-neutral or agent-relative reasons. Second, it is argued that the thin account implies that agents face endless reasons; if this implication cannot be resisted the thin view appears aesthetically less appealing than the thick account. The thick account is not flawless, though. At the end, I will discuss a couple of objections to it.

Agent-neutral reason, agent-relative reason, complete reason, normative reason.

Reasons are facts, i.e., they are constituted by facts. This "reason qua fact" claim is much endorsed in recent literature. In what follows I would like to address a couple of issues that arise when we apply this idea to practical normative reasons. My interest in this matter dates back to something I recently argued for, namely that we should be cautious about endorsing the distinction between agent-neutral and agent-relative reasons, if we understand this dichotomy in terms of facts (rather than, say, statements or principles). I shall not here repeat my rather extensive argument.

However, I shall discuss some issues that arise for the distinction depending on what sort of view we take on the nature of facts. So although I shall not try to answer the question –'can we make sense of the dichotomy if we have normative practical reasons qua facts in mind?'– my contribution will at least outline some issues that need to be addressed before we answer this question.

En passant, the questions that I will address are of a magnitude that could easily discourage even the bravest of inquirers. For this reason, I would like to make it clear at the outset that since, inevitably, I will be cutting corners I have no illusions of settling anything at all when it comes to the issues that I will discuss.

Without going into too much detail, we can distinguish in broad terms between two (or perhaps even three) important views on the nature of facts. Given in particular a view that conceives of facts as abstract entities, the dichotomy is not particularly problematic.

We might run into problems when it comes to identifying which facts are reasons and which are not, but the very dichotomy itself seems sound. So on this, what I will refer to as, thin approach it is at least a possibility that some reasons qua facts will be (in some sense of the dichotomy) agent-neutral and some will be agent-relative.

The second approach, which I will comment on, the thick account, is, I believe, much less popular. However, as I shall argue, it would be better if we could understand the dichotomy even if we had this notion in mind. It would be preferable because it is better if our notion of a reason is consistent with a wider rather than narrower set of plausible metaphysical views. But, more importantly, it would also be preferable because I think the thin approach, as I will explain in section 3 (the thin account), trivializes an interesting issue among practical philosophers. I will also argue that, as an additional drawback, the thin account is in one respect less appealing than its thick cousin. The thick account is not flawless, though. At the end, I will discuss a couple of objections to it.

Before we continue, a caveat is in place. As already mentioned, the dichotomy has been understood in a number of ways. I will set out from what is sometimes referred to as the essentialist sense of the distinction.

This is what I take to be Nagel's (but not necessarily e.g. Derek Parfit's) sense¹. Nagel draws the distinction in terms of what does and what does not essentially refer to a particular person, the owner of the reason. Various other writers have employed similar accounts.

In order to understand the reason-qua-fact idea, we need to take a short detour and formulate some idea about what kind of statements express these reasons. It is not obvious, of course, just how we should formulate a reason statement, i.e. a statement expressing a fact that constitutes a reason. Moreover, since there are different kinds of fact, and we should be open to the possibility that they can all be reasons, most likely there will be different kinds of reason-statement. To simplify matters, I suggest that we simply understand them as expressing what may figure in a that-clause. I therefore propose we set out from the following examples:

AN: A person a is drowning in circumstances S.
AR: My daughter is drowning in circumstances S.

Both statements refer to what we typically would refer to in order to answer questions of the form "Why should I/you ∅ (e.g., jump into the water)?"

Presumably, both AR and AN, describe only one aspect of a more concrete state of affairs². Let us for the moment focus on the one expressed by AR. Other features of this state of affairs would be, for example, that the person (referred to in AR as drowning) has curly hair, lives in Barcelona, and desires to be rescued in S. Needless to say, the list of features can easily be made very long. In fact, there seems to be no end to how many different descriptions we can give of a concrete entity.

Understanding the statements in this way raises the question just what is the reason-constitutive element in AR (and AN)? What precisely is it that constitutes the reason? Prima facie, only a few options seem reasonable: it might be the thin feature(s)³ explicitly expressed by the statement –that my daughter is drowning in S– or it might be the thicker state of affairs of which that my daughter is drowning in S is only one aspect out of many?

What I will refer to as the thick account of reasons suggests that reason statements refer to concrete entities; i.e., it is thick entities that are the reason-constitutive parts rather than the thin features. The state of affairs is, then, concrete in the sense that it contains a number of features that may be abstracted from these entities⁴.

Some of these features are more salient than others, and what e.g. AN and AR express is these salient features rather than the whole picture. On the thick approach the propositional element in a reason statement is like the tip of an iceberg; it displays only a tiny fraction of the whole entity.

The main challenger is the thin account. This approach, which I believe is more widely accepted, comes in two versions. However, to begin with it is important to notice that it should not be understood as denying, e.g. that the statement that my daughter is drowning in S describes a feature of a thicker state of affairs. But, for our purposes, the important thing to stress is that according to the thin account, the reason-constitutive element is only effectuated by the thin feature, and not by the thick state of affairs. So whether or not there is some other more concrete state of affairs involved, what makes AR or AN into a true reason statement is an aspect or a conjunction of aspects of a state of affairs rather than the concrete state of affairs.

One thing needs to be made clear. In this work I distinguish between "state of affairs" and "the obtaining of a state of affairs". This allows me to regard states of affairs as one kind of entity and obtaining of states of affairs as a second kind of entity. Since there are, say, no mermaids, there is no obtaining of the state of affairs that mermaids live in the sea.

Moreover, since we are supposing that AN and AR express something true, they must therefore be understood as referring to an obtaining of the state of affairs rather than to a state of affairs. In what follows I will sometimes speak about these obtaining as concrete state of affairs.

According to the second, and I believe, more widespread version of the thin account, we should rather understand reasons as being the content of possible judgements that are true.

This suggestion sets out from a different perspective than the one expressed in the first version. At least, it would be possible (many would even say reasonable) to regard the first version of the thin account as being about the truth-maker of the proposition which the second version considers to be the reason-constitutive entity.

In what follows, to simplify matters, with the "thin account" I will have the latter, propositional version in mind. These accounts (as well as a combination of both) merit different treatments, and so my treatment of the thin account will necessarily be incomplete.

However, discussing all accounts would have made this paper considerably longer. Moreover, since, I believe, the propositional approach is the more popular, I have chosen to discuss it rather than the "feature approach". (Later on I will briefly point out why the latter approach is problematic).

There are advantages as well as disadvantages with the propositional account of the thin account. The most obvious advantage, I suppose, is that it certainly allows us to draw the distinction between agent-neutral and agent-relative reason statements (in the essentialist sense). Since the reason dichotomy is intuitively plausible, we should therefore welcome it. Moreover, it allows us to draw the distinction in a fairly simple way: We all agree that any statement contains either at least one indexical (e.g., 'I', 'my', 'he') or no indexical at all.

Hence, it would be absurd to deny that at least some possible proposition essentially refers to the person who issued one of these statements and thus endorsed the proposition. And since it is safe to say that at least some such propositions are true, we have cleared the ground for the existence of agent-relative reasons for a view that takes reasons to be thin entities. Some of these true, essentially referring, propositions will (it would at least seem so), be reasons.

In a structurally similar way, we can reason about statements and propositions that are true but are not essentially referring to the person whose reason they express. Some of these are suitably agent-neutral reason propositions. So if the true statement (proposition) refers essentially to the agent for whom the proposition is a reason, then it is an agent-relative reason; otherwise it is an agent-neutral reason.

This, then, is obviously a plus. But, in my view, it is a somewhat stained advantage. The discussion whether there are any agent-neutral reasons or, for that matter, any agent-relative ones, does not only turn on whether we are able to formulate true statements (and hence express propositions) that contain or do not contain ineliminable references. Certainly, the controversy concerns an evaluative or normative matter; it concerns, among other things, whether an ineliminable reference to an agent gives rise to a particular kind of reason, viz. an agent-relative reason. It does not concern whether a certain true statement or proposition involves such an ineliminable reference. That is not an evaluative issue.

Therefore, to suggest that we can draw a distinction between agent-neutral and agent-relative reasons if we acknowledge that some statements refer essentially to the agent and some do not, might clear the ground for the distinction but it seems nonetheless to miss out on something important. It trivializes an interesting debate, which is unfortunate. Later on, in section 5 (problems for the thin account), I will substantiate this claim in greater detail. However, for the time being, a couple of other possible advantages ought to be mentioned.

One thing that might be talking in favor of the thin account is the following: It is generally acknowledged that reasons play or at least should be able to play an important part in our lives. However, if reasons are concrete obtaining of state of affairs, clearly they cannot play the sort of direct "in the head" role as a thin propositional entity appears to play. If reasons are the content of possible judgements that are true it becomes easier to understand how reasons could move us to actions, and hence play a more direct role in our lives. Or so the argument goes.

However, this "In-the-head advantage" is actually not much of an advantage⁵. Since we are talking about reasons as the content of possible thoughts that are true, then, if we have at least a classical view of truth in mind, reasons obviously will be entities with extrinsic features. They are therefore necessarily entities that relate to that which makes the thought true. Being, then, essentially a relational entity to something outside the agent (or at least to something outside the thought that p), suggest that the "in-the-head-advantage" is not really an advantage after all. What is in the head is more accurately (or, at any rate, just as correctly) described as the grasping of the fact or feature, but not the fact or the feature itself.

An advocate of the thick account can therefore rightly point out that she too can agree that the grasping, e.g. the believing that Anna is drowning, goes on in the head. She will just deny that this constitutes the reason, and that this should be recognized as well by the supporters of the thin account. After all, in their view, reasons encompass relate, one of which is external to the thought that p.

Notice, by the way, that the above objection is, I believe, consistent with the substantial idea that the only facts that constitute reasons are facts about the agent's beliefs and desires. Unless we want to say that the facts are in the head of persons, the above point stands unaffected by the idea that only facts about the agent's beliefs and desires constitute reasons for her. Or more guardedly put: the point is not that facts are not in the head, but rather that it is at least as plausible to say that they are not in the head as it is to say that they are in the head⁶.

One thing that also perhaps talks in favor of the thin account is the fact that some reasons are hard to understand as anything but thin entities. Consider the following way of talking about reasons, which I borrow from Schroeder (11): suppose someone asked us why there are no circles with some points further away from their centre than others. We might reply the reason is that being a circle means having all of the points equidistant from the centre.

The "because fact" that we allude to here is more easily understood, I suppose, as a proposition (period); it is at least not clear to me what the concrete obtaining of a state of affairs would be in such a case. Its truth-maker seems to be some thin rather than concrete entity. We need not get into the complex matter of what makes such propositions true. But whatever it is, it does seem like a thin rather than a concrete object. Suppose this is correct. In that case, should we perhaps count this as evidence for the hypothesis that reasons are thin entities? Yes, I suppose; it is some evidence.

However, we should not be impressed by this sort of quick reasoning, which, unless qualified, simply boils down to a hasty generalization. Perhaps some reasons are like the above examples, but I very much doubt that all practical reasons are.

In section 5 (problems for the thin account), I will bring up what I take to be two major disadvantages with the thin account. First, however, we need to get a better grip on what reasons actually involve according to the thin account.

It is customary to look upon propositions involving an indexical (typically e.g. 'I' or 'my') as if they are someone's proposition, i.e., they are what a certain person expresses at a given time and at a given place. For the sake of uniformity, I suggest that we introduce this "owner" element whenever we talk about propositions. There is really no consensus on how propositions should be understood. However, in this context, I think this manoeuvre is rather inoffensive and non-committal. If we therefore keep this feature in mind, we may now list the key elements involved in a reason, according to the thin account, in the following way:

p is the true proposition; the function of the two kinds of parenthesis is to stress that their content is of a different kind than what is otherwise listed; x refers to the possible person entertaining the proposition, y refers to the person for whom the proposition is a reason, and ∅ is the act which y has a reason to perform in virtue of the fact p. In some cases, however, x and y may as a matter of fact refer to one and the same person. For instance, this would be an example of an agent-relative reason, where TRR refers to me.

Supposing (ii) is correct, we can now state why it expresses an agent-relative reason (on the essentialist account); the proposition that constitutes a reason for TRR contains an ineliminable reference to the person for whom it is a reason. There is a cross-reference between the person referred to as "my" and the person (the second TRR) for whom the proposition is a reason⁷.

On the other hand, the following would, then, be an example of an agent-neutral reason:

There is no cross-reference involved between TRR in (iii) and the proposition, and so, if this proposition (that a dog is drowning) constitutes a reason it constitutes an agent-neutral reason.

The above examples display the key elements involved, when something is a reason for a particular person, i.e. if we embrace the thin account.

So far, I have assumed that it is the true proposition p in (i) (i.e. x [p] y (∅)) that is the reason-constitutive element. However, this might be questioned. One might argue that the reason is not actually p but rather the whole substantial proposition p*: that x [p] y (∅), i.e. the entire content of (i).

However, I think there are good reasons to be wary about such a suggestion. First, recall that all I said about (i) was that it lists elements involved when some-thing is a reason. I did not say that (i) itself expresses a proposition. But perhaps there is indeed a proposition lurking in the vicinity. Maybe (i), in fact, expresses, say, that "x endorses the true proposition p for y" (or something to that effect), and that it is this fact, i.e. this true proposition, that constitutes a reason for y. This sounds somewhat awkward, however. It is not clear what it means to endorse a proposition for someone.

Maybe (i) should instead be read as expressing an even more complex proposition, viz. what is expressed by the following (N): "x endorses the true proposition p, and p is R-related to y" where R, then, specifies some kind of relation between (the endorsement of) p and the person for whom p is a reason. This (naturalistic) suggestion is not without its merits. Let us refer to it as the Agent-endorsement account of reasons, or AEAR. For instance, explanatory reasons are often said to refer to this sort of AEAR propositions, i.e., propositions like the one expressed in (N). When we want to understand why an agent acted as she did, we tend at least to invoke at some point a belief of the agent, especially a belief whose intentional content will have some motivational force for the agent⁹.

But although this might be one way to understand explanatory reasons, I think (N), or something to this effect, should be avoided as an analyses of normative reasons. Admittedly, my reasons are mainly substantial. AEAR strikes me as simply unconvincing; why should we believe that of all the kinds of fact that might constitute a reason, it is only facts about what persons endorse (believe, think, etc.) that constitute reasons. That my dog is drowning would not on this account be a reason, only that someone endorses that my dog is drowning. There are further problems with this sort of view, but I think we may leave it and move on.

The real worry with the thin account is that it locates the difference between agent-neutral and agent-relative reasons only in our descriptions of the world. I have already hinted at why I suspect this. The time has now come to spell out my worry in more detail. If substantiated it would certainly be, in my view, a serious drawback for the thin account.

Consider the (claim expressing the) proposition "Buster is drowning". Suppose we accept the proposition as an agent-neutral reason maker for me, i.e. TRR. According to our schema, the following list, then, captures the core elements involved in an agent-neutral reason for TRR, assuming that Buster is a dog and that we have a reason to save a dog that is drowning:

But, if (iv) is true, so is, ex hypothesi, the agent-relative (ii) TRR [my dog is drowning] TRR. That is, TRR's proposition "my dog is drowning" constitutes an agent-relative reason for TRR. Both "Buster is drowning" and "my dog is drowning" refer, that is, to features of one and the same obtaining state of affairs (viz. one in which a particular dog that is called Buster and is owned by TRR is drowning). What is more, if (ii) and (iv) are true, so is, ex hypothesi, the following (where ARR refers to TRR's daughter):

The thin account says, on the essentialist reading, that whether a true proposition is agent-neutral or not for TRR depends on whether the (corresponding statement expressing the) proposition p contains an adequate proper name ((i.e. one that corresponds to the x in (i)), or an adequate indexical¹⁰.

However, to say something more substantial the thin account has to say, as a minimum, that there is a normative difference between (ii), (iii), (iv) and (v). It should normatively matter, in the sense that it should at least be feasible that only one of these might be a reason for TRR, or at least might be a stronger/weaker reason for TRR than the alternatives. Otherwise, the distinction would not be interesting or important. However, I fail to see that such a normative difference is detectable, and so I find the distinction trivialized.

Now, this point is far from obvious. Let me therefore explain how it resists the more obvious rejoinders. First, one might object that, depending on who will act on these propositions, it will certainly in some cases make a difference whether the person believed the proposition in for instance (iv) rather than the one in (v); so although we assume that both are made true by (or, if you like, are features of) the same concrete state of affairs, it is easily conceivable that grasping the proposition in (iv) might have a different effect on y than if she believed the proposition in (v).

For instance, suppose TRR falsely beliefs that ARR is not his child but that he correctly believes that Buster is his dog. In that case, it is quite expectable that, depending on whether the agent believes (iv) or (v), it will have a different impact on what she will do (or even take herself as having a reason to do or be, as a matter of fact, motivated to do). As is well known, the intentional context of a proposition is crucial when it comes to the effect that a proposition has on the agent's behavior.

I definitely do not want to deny this. And so, if we could reduce normative reasons to what goes on in the agent's mind, in terms of propositional attitudes, it is clear that we would have an interesting reason notion, and, in its train, a plausible reason dichotomy. But since, I think it is safe to say, the normative still resists this kind of reduction, this sort of rejoinder currently misses its target.

It might be objected that this objection to the thin account rests on an unwarranted assumption. At least, it seems as if I assume here that what makes the proposition true is some concrete state of affairs. But, it might be argued, it is in effect not some concrete state but rather only the abstract feature(s) that have this role. It could accordingly be argued that since these features, corresponding to, say, what makes (i)-(v) true, are different, we should also say that (i)-(v) express different reasons. And since these features may in different possible situations call for different sorts of actions, we should conclude that this sort of view handles the objection from above.

I am not convinced by the above reasoning, though. Suppose we agree that it is different features, period (rather than features of a concrete entity) that make, say, (iii), (iv) and (v) true, and that we accordingly should say that these three true propositions are or at least refer to different reasons. This in itself does not meet the challenge; it still needs to be argued that individuating reasons in this way is normatively interesting.

This shouldn't simply be assumed. In fact, when these features make the propositions true in the same world, at the same time, it still appears puzzling to me how one could say that it normatively matters in the sense outlined above. I still do not see that it does.

These thin features (expressed by iii, iv and v), I take it, have relational properties to other features and in particular to each other in the case they are true in the same world, such that they will make true some further propositions that make it clear, with regard to the examples we are discussing, that, for instance, ARR is TRR's child. So I have a hard time seeing that invoking these features as separate reasons makes a normative difference.

One might object in a different way. I have said that at present it is still unclear whether normativity can be naturalistically reduced. Besides this qualification, I have left it open just what makes a fact into something normative. But suppose we endorsed an account along the lines of John Broome's (2004) interesting suggestion that a normative reason for ∅-ing is what explains why the agent ought to ∅. In that case, it seems as if it would make a difference whether we offered the explanation in terms of, say, "the dog of ARR's father is..." rather that in terms of "TRR's dog is..." This reply works, I think. If normativity is understood in this way, then it is conceivable that it normatively matters whether we (iii), (iv) or (iv) is a reason.

However, although appealing, Broome's suggestion is not an obvious one. What is an explanation in one context need not be so in another; explanation is a success notion in a way that the notion of a reason is not. Intuitively, it is at least conceivable that I could have a reason for ∅–ing, even if every attempt to explain why I ought to ∅ failed. Perhaps we can get round this sort of intuitive objection. At present I do not see how, though.

The second problem with the thin account is that it is aesthetically less appealing than the thick account. Recall the examples from above (ii, iii, iv, v).

The unattractiveness derives from the following: since it is easy to imagine that one and the same concrete entity displays endless variants of the aforementioned features, the thin account will seemingly have to say about our example that the agent faces an endless list of reasons for saving the dog.

This strikes me as an odd thing to say. But perhaps accepting this might, after all, be quite inoffensive. Perhaps we might just say that there is only one reason but lots of versions of it. How inoffensive such claims are depends, I suppose, on whether we take each of these reasons (or versions of reasons) to have normative strength, and, in particular, whether we think these strengths somehow add up.

The latter seems hard to justify, however. After all, it would have some rather counter-intuitive implications. If each of these reasons carries a certain normative strength, we appear to face the strange conclusion that at any given time that an agent has a reason to perform an act, he or she has infinitely strong reasons to do so–strength here being interpreted as the sum of the strength of each reason. This also strikes me as odd.

An advocate of the thin account is therefore well-advised to deny that all these reasons somehow add up in strength. But, of course, in that case, it seems reasonable to ask why we should in the first place crowd the canvass with countless reasons (it gives us a lot of unnecessary trouble, such as finding out which true propositions carry normative strength that may add up with the normative strength of some other proposition); it seems better (i.e., gives a cleaner picture) to say that there is actually only one reason (or, in the last example, two pro-tanto reasons), viz. one that is constituted by a concrete fact that can be described in numerous correct ways.

The idea that reasons are concrete is, in one respect, a simpler model; it involves fewer elements, as the following list shows:

Here O refers to the obtaining of a state of affairs. There is no mention of any proposition or -what is more important- the person whose proposition it is. However, the thick account introduces the complicated notion of an obtaining, i.e. a concrete entity. Needless to say, this idea is much less transparent than what could be wished for.

But it is important not to overlook that whatever fault we take this to be for the thick account; it will most likely also be a flaw for the thin account. In any case, it will not be a simple and straightforward matter to do without the distinction between a state of affairs and its realization (i.e. the obtaining of a state of affairs).

So it is far from clear that the thin account can profit from whatever flaws we detect in the thick account's core notion. But the thick account can, on the other hand, take advantage of the faults we detect in the thin account due to its idea that reasons are true propositions. Propositions play no constitutive part in the thick account.

As expected, the thick account faces in its turn some fairly common objections. Perhaps the most common one is the following: Although we often speak of facts, the idea that we can separate one fact from another is puzzling; there is actually only one concrete entity that can be described as the obtaining of a state of affairs, and it is the one which encompasses how the world is at a present point in time (or something close to this).

But this, then, seems to lead to the absurd idea that my reason for, say, saving a dog from dying, is constituted by this monstrous entity that encompasses all the features of the world at the given time that I have a reason to rescue the dog. The thick account should therefore be rejected.

There are different variants of this sort of reductio ad absurdum (and I must confess, not all the details are equally clear to me); basically they all involve two ideas that must not be conflated. The first one puts it to us that what constitutes a reason cannot be too complex or exceed a certain limit; e.g., it cannot incorporate too much or too many items. But, the argument goes; the problem with the thick account is that this is precisely what it does. The sheer size of what constitutes a reason, on the thick account, is so overwhelming that it should be evident that my reason for, say, saving the dog simply cannot be constituted by such a colossal entity.

The above "monstrosity objection" involves a related but nonetheless different idea. Even if we somehow accepted that there is no limit to the size of a reason-constitutive entity, the thick account is open to a perhaps more serious objection. The sort of entity that it suggests is constitutive of a reason would simply involve features that it would be implausible to include in the supervenience base of the reason. Whatever it is that makes it into a reason that the dog should be saved, it certainly does not involve facts about, say, people and objects on the other side of the planet. This, as we might call it, "too inclusive" objection seems to me to be more serious than the former one. However, to begin with, let me briefly comment on the former objection.

I think there are mainly two explanations close at hand for why anyone would suggest the monstrosity objection. The first stresses something that we have already commented on, namely that reasons should be in some sense accessible to us. Simplifying, the point can be expressed as follows: For something R to be a reason for a person x, R must at least be graspable by x; otherwise x could not have acted for the reason R. Such a claim would most certainly have to be qualified. However, there is no need here to go into details. Essentially, what it says is that there is a necessary (conceptual) link between "R is a reason for x" and "x could act for a reason R", i.e. "x could be motivated by R".

The problem with this sort of view is that it simply does not seem to be true that there is such a necessity. Perhaps I am wrong about this, though. Once we qualify the circumstances, and disregard abnormal cases, irrational persons, etc., it will perhaps become manifest that there is such a tight relation between 'being a reason' and 'being such that a person in circumstances C could act from it". But again, this remains to be shown.

My suspicion is that reasons need not be graspable. However, since I think (as will become clear in a moment) the thick account actually is consistent with "the graspable requirement", for reasons I will get to in a moment, I will not argue for this here¹¹.

The second explanation is quite different. It sets out from the distinction between a pro-tanto, or contributory reason, which it takes to be constituted by some aspect(s) of the world, and a complete reason, which it understands as what would be called for if we took "everything" into consideration. It then argues that the thick account somehow assumes that we can have a "complete reason". But, as the argument has it, the notion of a complete reason is not coherent, and so we should give up the idea that reasons must be understood as thick entities.

For sure, if we cannot find a reasonable way of understanding "complete reason", we are handed a strong incentive for giving up the thick account. However, I am not sure that the notion of a complete reason is really inconsistent. For reasons I will get to in a moment, I think it would be unwise and certainly premature to set this idea aside as being incoherent.

The too inclusive objection carries greater force than the two above objections. The gist of it is that we should reject the thick account because otherwise we would end up with a set of subjacent features that simply cannot be what makes something into a reason. For instance, ex hypothesi, in this world where Buster is drowning, there are numerous features that are irrele-vant. We do not want to have to say that my reason for saving the dog involves, among other things, facts about, say, the number of leaves on one particular tree 10,000 miles away from the dog and me. This would definitely be absurd to accept.

Notwithstanding this, I think the aforementioned objection can be discarded. One way to do so would be to admit that the supervenience base includes a somewhat peculiar kind of fact, viz. a fact about facts. If what calls for me acting in a certain circumstance are not only certain positive features but also the absence of countervailing features, we have a response to the "too inclusive objection".

Thus, what makes something into a complete reason are certain salient features, together with the peculiar feature that any remaining feature is irrelevant. We do not have to give up the idea that among the reason-making characteristics of the world, we should only include those that are relevant for the explanation of why the agent ought to ∅. It is just that one of these relevant features is the feature that there are no other relevant features. So if at a given time the facts are such that certain features positively call for some action ∅ by me, and it is the case that there are no other features that would discard this call, then I have a complete normative reason for acting.

I am inclined to accept this sort of "complete reason". The drawback is that it seems that in order to explain normative reasons, we will have to do so in terms of what appear like normative notions, namely "relevance/irrelevance". Perhaps this should make us look elsewhere¹².

I have outlined two ways of understanding the idea that reasons are (constituted by) facts. I have also argued that it would be preferable if we could understand the agent-neutral and agent-relative reasons dichotomy in such a way that it is consistent with both these views. The idea behind this contention is that the thin account is inferior to the thicker account in at least some aspects. Of course, this conclusion does not suggest that the reason dichotomy is in the clear. The thick account might lead to complications for the distinction in a way that the thin account avoids. In fact, I believe this to be the case. However, this is not something that I will argue here.

¹ Originally Nagel (1970) referred to the distinction as one between subjective and objective reasons. The terminology 'agent-neutral and agent-relative' was introduced by Derek Parfit in (1979). See Thomas Nagel (1986); especially (152-3).
² Does it have to be a unique state of affairs? Not necessarily. I am at least inclined to share the view of those who believe that the "closest world" in fact refers to a set of possible worlds that all share the property that they are equally close in resemblance to the actual world. However, for my purpose in what follows nothing important hinges on whether the thin feature is considered as an aspect of one unique closest possible world or of many equally close possible worlds.
³ The thin account may, of course, take a reason to be constituted by two or more features.
⁴ I would also add that in contrast to abstract features, these concrete entities may consist of objects in a wide sense (persons, things, etc.).
⁵ Observe that if we understand the thin account to be about features rather than the proposition, the advantage disappears.
⁶ Moreover, a view that made the feasibility of our reason notion or reason dichotomy dependent on taking sides on this matter would therefore not be particularly appealing. This, I believe, cautious point will nevertheless not sit well with everyone. But whether or not we are comfortable about discussing where to locate facts, it would be unfortunate if the dichotomy were only applicable by advocates of the idea that only facts about the agent's desires and beliefs are reasons.
⁷ (i) is a simplified picture that easily expands into the following: xt [pt'] y ∅t'' here t refers to the time when x expresses the proposition, and t' refers to the time of the "propositional object", and t'' to when the reason applies to y, i.e. when y's ∅-ing is normatively required. However, in this work I have had to simplify matters. E.g., the earlier suggestion that x in (i) might refer to the 'grammatical subject' has to be further specified, and it is also possible that p must somehow involve the act ∅. But since I believe a discussion of these matters would not substantially affect my arguments against the thin account, I have decided to leave out these qualifications.
⁸ But suppose x has TRR's dog in mind in (iii), and it is a reason for TRR. In that case, should we say that this still is an agent-neutral reason for TRR? I suppose so.
⁹ A caveat is in place: In general AEAR would not suit as a motivational reason; this kind of reason should not be confused with explanatory reasons. What motivates an agent is seldom his belief that p; it is rather that p.
¹⁰ That indexicals might cause trouble for the agent-relative/neutral dichotomy was made clear by Derek Parfit: "Even if you and I are trying to achieve some common aim, we may be in different causal situations. I may have reason to act in a way that promotes our common aim, but you may have no such reason since you may be unable to act in this way. Since even agent-neutral reasons are, in this sense, agent-relative, this sense is irrelevant to our discussion" (143). However, just what Parfit commits to here is a matter I leave open. His understanding of the dichotomy does not quite coincide with the essentialist reading. For more on this, see Rønnow-Rasmussen (2011).
¹¹ There is a weaker condition, which seems more plausible, viz., that the parts of a reason should be accessible to the agent. Although the whole reason might be a real mouthful, impossible to grasp, each of its constitutive parts might be graspable. This suggestion, which I owe to Wlodek Rabinowicz (personal communication), does seem plausible. Moreover, it is quite accessible to advocates of the thick account.
¹² We might avoid the circularity by suggesting that 'relevant/irrelevant features' refers to an evaluative but not necessarily normative feature. For instance, it might be claimed that the distinction is not really referring to what are reason-making features. Rather, a relevant/ irrelevant feature might be described as one that is e.g. good/bad at motivating us, or at least at explaining why you should do a certain act, or, what calls for an action in the most effective way. Just how viable this suggestion is needs to be further examined. However, it is clearly not an option for advocates of so-called Buck-passing accounts or Fitting attitudes analyses of value.

Broome, John. "Reasons". Reasons and value: Themes from the moral philosophy of Joseph Raz. Ed. R. J. Wallace. Oxford: Oxford University Press, 2004. 28-55. Print.         [ Links ]         [ Links ]

Parfit, Derek. "Prudence, morality and the prisoner's dilemma". Proceedings of the British Academy, 1979: 539-564. Print.         [ Links ]         [ Links ]

---. Personal value. Oxford: Oxford University Press, 2011. Print.         [ Links ]         [ Links ]

Svensson, Frans and Rysiek Sliwinski, Ed. Neither/Nor-philosophical papers dedicated to Erik Carlson on the Occasion of His Fiftieth Birthday. Uppsala: Philosophical Studies, 2011. 243-257. Print.         [ Links ]