Explaining Reasons: Where Does the Buck Stop?
Heuer, Ulrike, Journal of Ethics & Social Philosophy
IT WOULD BE GOOD IF I could finally knock this paper into shape. That gives me a reason to sit down and work on it. Or is it the other way around? Is it that I have a reason to get on with it and that is why it would be good to do so? According to the buck-passing account of value, it is the latter: Something is good because it gives us a reason. The reverse view has it that one has a reason for doing things because doing so would be good in certain respects. The buck-passing account offers an explanation of the close relation of values and reasons: of why it is that the question whether something that is of value provides reasons for action is not "open." Being of value simply is, its defenders claim, a property that something has in virtue of its having other reason-providing properties. The buck-passing view combines various virtues: First of all, it offers an explanation of the close connection between values and reasons for actions. Yet the account of reasons for actions that it proposes need not make any reference to values. Furthermore, in at least one of its versions, it is a reductive account of values in the sense that it does not appeal to evaluative properties in explaining value. According to reductive buck-passing accounts, values are to be explained neither in terms of evaluative properties, nor by reference to our (perhaps suitably corrected) attitudes. Rather, so the suggestion goes, we can explain value in terms of non-evaluative, natural properties that provide reasons for those attitudes. But unlike rational attitude theories of value (1), which some regard as the historical predecessors of buck-passing (2), the buck-passing account is not amenable to the project of developing a metaphysical reduction of normative properties. The reduction it offers is an intra-normative one: After all, the explanation of values is in terms of reasons for actions.
The generic idea of buck-passing is that the property of being good or being of value does not provide reasons (3), but it is other properties that do. There are, however, various versions of the account which differ in their understanding of those "other properties." Briefly, they are the following ones:
(BP-1) x is of value (is valuable or good in a certain respect), iff it has other non-normative, natural properties that provide reasons for actions.
(BP-2) x is of value, iff it has either (a) other evaluative or (b) non-normative, natural properties that provide reasons for actions.
There are instances of both (a) and (b).
(BP-3) x is of value (is good), iff it has other evaluative properties that provide reasons for actions, but goodness itself is not a reason. (4)
In this paper I am going to explore all three versions of the view. I will begin with explaining the buck-passing account in greater detail, and then investigate T. M. Scanlon's arguments for it (in section I). In section II I will offer an argument, which, I believe, raises serious doubts about a central presupposition of buck-passing in two of its versions (BP-1 and BP-2), namely that non-evaluative properties are reason-giving. Finally I argue that the third version, BP-3, is also unlikely to succeed (in section III).
Let me issue a warning: All three versions of the buck-passing account have at times been ascribed to Scanlon. There is some textual evidence to do so in each case, but not enough to exclude any of them. Be that as it may, my main interest is not in interpreting Scanlon. All three propositions are interesting in their own right, and worth thinking about. So I will keep all of them in play without claiming that one or the other is Scanlon's "true" or even current position. Scanlon's discussion of buck-passing is extremely terse, but it has the merit of forcing us to consider the various possibilities. The important point is that we now have various interesting, but mutually incompatible, articulations of the buck-passing view on the table.
(1) The original account (= BP-1)
The "buck-passing account" (96) is T. M. Scanlon's label for his own account of values which is captured in the following sentence:
... being valuable is not a property that provides us with reasons. Rather, to call something valuable is to say that it has other properties that provide reasons for behaving in certain ways with regard to it. (96) (5)
What makes it a "buck-passing account" is "holding that it is not goodness or value itself that provides reasons but rather other properties that do so" (97). So it passes the buck from values to the other properties which provide reasons. Scanlon explains that those other properties that provide us with reasons are "natural" or "non-normative" properties (96). The claim, as Scanlon expresses it, appears to be conceptual or semantic: "to call something valuable is to say ..." [my emphases].
Yet, it seems clear that there is a metaphysical claim involved as well: For it to be true that something is valuable, it has to have certain non-normative properties which "provide reasons for behaving in certain ways with regard to it." Metaphysically speaking, the use of the concept of value has true applications only if there are appropriate natural properties and reasons, but it does not presuppose the existence of values. The buck-passing account offers a metaphysical and even a conceptual reduction of values; yet, it explains values in terms of reasons. So, it is not--and is not intended to be--a reductionist account of all normative concepts. (6) Scanlon illustrates his view by giving the following examples:
... the fact that a resort is pleasant is a reason to visit it or recommend it to a friend, and the fact that a discovery casts light on the causes of cancer is a reason to applaud and to support further research of that kind. These natural properties provide a complete explanation of the reasons we have for reacting in these ways to things that are good and valuable. It is not clear what further work could be done by special reason-providing properties of goodness and value, and even less clear how these properties could provide reasons. (97)
Here there are the "natural" and "non-normative" properties of being pleasant or of casting light on the causes of cancer, which give us reasons of various kinds regarding those things that instantiate the properties. (7)
There is a prima facie difficulty with this idea since according to a common philosophical understanding being reason-giving just is being normative. How can non-normative properties be reason-giving then? In formulating the position Scanlon relies on an intuitive way of distinguishing normative from non-normative properties. This in itself may not be a problem. But it follows that if buck-passing is correct we will have to rethink the common theoretical understanding of normativity.
(2) The next stage (= BP-2)
Jay Wallace in his discussion of Scanlon's book (8) raises an obvious doubt:
To say that a resort is "pleasant," for instance, is a way of adverting to the distinctively positive qualities of experience that are enjoyed by a visitor to the resort. It is not merely an evaluatively neutral description of the natural properties of the resort or of the experiences induced by the resort in its visitors, and this is what makes it appropriate to think of pleasure itself as a concrete category of evaluation. (448)
Wallace suggests that instead of maintaining that values cannot provide reasons, Scanlon's point might have rather been that the "abstract" property of being good does not provide reasons, whereas "concrete" evaluative properties might well do so. Hence, he suggests passing the buck in explaining reasons only from "abstract" goodness to more concrete values, but not from values to non-normative, natural properties. Scanlon's reply (9) to this suggestion reads as follows:
I agree with Wallace's reinterpretation of my "buck-passing" account of value. My thesis was that goodness is not itself a property that provides reasons, not that the underlying properties that do this are always natural properties, and I should not have written in a way that suggested this. He is quite right that more specific evaluative properties often play this role. (513)
The quote suggests a modified version of the buck-passing account. The correct view, Scanlon now says, is a hybrid one: Both natural and evaluative properties can provide reasons. If something has properties (of either kind), which give us reasons, it is good--but goodness itself does not give us reasons. Let me call this the modified or the hybrid view (or BP-2) which differs from the original one in that it endorses a broad understanding of the reason-providing properties. (10) I will now investigate Scanlon's arguments in favor of buck-passing.
(3) Scanlon's arguments for buck-passing
Scanlon makes the case for buck-passing in two stages. His arguments purport to show that buck-passing must be true because the only alternatives to it are quite implausible. If being good is a reason, then it either is the only reason, or it is an additional reason--so goes the general assumption that seems to underlie the argument as a whole. In the first part (hereafter: "first argument"), he argues that it is not an additional reason; in the second (hereafter: "second argument"), that it is not the only one. Together, the theses established by the first and second argument are deemed to be necessary and sufficient for buck-passing.
First argument. Scanlon claims that we need not invoke the property of being good in order to explain our reasons, but only natural properties such as being pleasant or casting light on the causes of cancer. That the cancer research is good or valuable, Scanlon maintains, would surely not give us any further reason to support it in addition to the fact that it casts light on the causes of cancer. Adducing the property of being valuable in the explanation of our reasons is like putting in a fifth wheel that does not pull any weight. Hence, the suggestion that being good, rather than providing us with reasons, is explained by the fact that we have reasons, which are in turn provided by other properties. Goodness therefore is not an additional reason. (11) Let me call this the explanation argument. As you probably noticed, the argument is presented in terms of the original--not the modified--account, which is unsurprising since its formulation pre-dates the revision. The next question is therefore how does the modification affect the argument. Remember, the modified view claimed that some of the reason-providing properties, like being pleasant, are evaluative properties (and not non-normative, natural ones). What then is the relation of goodness and other evaluative properties, like being pleasant? Surely, goodness itself is also an evaluative property. But couldn't we nonetheless claim that there is no role to be played by goodness in explaining our reasons? All we need to refer to in order to explain our reasons are the more specific evaluative properties. The thesis would then be that something is good, if and only if it has other evaluative (and / or non-evaluative) properties that give us reasons. I will investigate this suggestion below.
Second argument. Scanlon claims that it is a "source of support for a buck-passing account ... that many different things can be said to be good or to be valuable, and the grounds for these judgments vary widely. There does not seem to be a single, reason-providing property that is common to all these cases" (97f). If we assume, as Moore did, that goodness is a simple property, then it cannot be a reason. Even in those cases where goodness figures as a candidate reason, it is not the only reason, but there is a plurality of properties that explain the fact that something is good and they provide us with reasons. (And as the explanation argument has already shown, goodness is not a further reason in addition to them.) Let's call the second argument the argument from pluralism. And again the question is how the argument is affected by the revision. If there is a plurality of evaluative properties that provide us with reasons, the question becomes what is the role of goodness within a pluralistic understanding of values. I will address this question below (in section III). To be sure, Scanlon's observation shows that goodness as G. E. Moore understands it--as a simple property--does not explain our reasons.
II. Against buck-passing (BP-1 and BP-2): Can non-evaluative properties provide reasons?
I now want to develop an argument that will show why buck-passing in both of its versions (the original and the modified view, or BP-1 and BP-2) is unlikely to succeed.
According to the modified view, it does not matter whether natural or evaluative properties provide reasons, as long as we understand that goodness does not. It grants that being pleasant is an evaluative property, but nothing much turns on that. But the difficulty with the original buck-passing account is not only that it ignored (and, in cases like pleasant, perhaps misinterpreted) many evaluative properties. There is a difficulty in claiming that non-evaluative properties can provide reasons at all, and that difficulty affects the modified account as well. Or so I will argue. Thus, in this section I will not deal with the generic idea of buck-passing directly--the idea that goodness does not provide reasons--but with an assumption that is shared by the first two versions of the account: that non-normative, natural properties do provide reasons (either exclusively, according to BP-1, or among others, according to BP-2).
I will start with some observations about pragmatics and context (paragraph 1), then take a closer look at the role of specific evaluative properties, such as being pleasant, in explaining reasons (paragraph 2), and thirdly explore the possibility of non-evaluative properties providing reasons (paragraph 3). The concluding fourth paragraph of the argument works out some differences between my view and buck-passing.
Virtually anything can be cited as a reason in appropriate circumstances.
Amy asks Bert, "Why did you leave last night's party in such a hurry?" "Because it was almost midnight," Bert replies. "Ah, yes, of course ...," says Amy, apparently satisfied with the answer. Is the fact that midnight approaches Bert's reason for leaving the party then? After all, it provides a satisfactory answer to a question that was inquiring for the reason. But it does not seem that being whatever time can in itself provide a reason for doing anything. If we were to specify Bert's reasoning in a more complete way, it might go something like this:
1. Bert has promised to be home at 12:30 at the latest.
2. There is no available transportation that will get him home in less than half an hour.
3. Bert has reason to make sure that he leaves at midnight.
4. It is almost midnight now.
Bert concludes that he will leave the party now. If Amy knows some of the other facts--(1) would suffice--she may be able to fill in or guess at the missing bits, and (let's stipulate) this is why she is satisfied with the answer. So she actually received all that was needed to understand why Bert left the party.
But what exactly is his reason to leave? Is it the conjunction of (1), (2), (3) and (4)? We could perhaps call all of (1), (2), (3) and (4) parts of his complete reason for forming the intention to leave. However, it would not be very helpful to do so. After all, each of (1) to (4) could be mentioned as a reason in an appropriate context, but even taken together they may not form a complete reason. (1) and (2) are sufficient for (3) only if Bert's promise is valid, for instance. I did not introduce the example to legislate whether a certain consideration may or may not be called a reason, but to show that many features which can be cited as reasons are not, taken by themselves, reasons for doing anything. This is surely true of (2) and (4). They can, however, figure as premises (like (1), (2) and (4)) or as derivative steps (like (3)) in a chain of reasoning. The point of the example is to illustrate that whenever a consideration such as (4)--which by itself is not a reason to do anything--is cited as a reason, there are further considerations in the background that make it that, in the given circumstances, it may well be called a reason. It does not matter much whether we claim that Bert's reason for leaving the party is really just (1), and (2) and (4) are part of the circumstances, or switch it around, and regard (4) as Bert's reason for leaving, but only in circumstances which contain also (1) and (2). The everyday use of the notion of a reason is very versatile and open-ended, and calling any of the properties mentioned in the propositions above a reason is not offending against the normal use of the term. But if (4) is a reason for leaving the party in these circumstances because of Bert's promise to be home at a certain time, it must be a reason to do so, whenever these circumstances obtain. This, I believe, simply follows from the ordinary understanding of reasons. (12) In the given circumstances, (4) can be regarded as Bert's reason for leaving. Taken by itself, however, (4) is not a reason for doing anything.
Facts that are frequently cited as reasons are often of this kind: The propositions that express them do not state the complete reason, but rather mention some feature that is salient because of what is of interest in the communication situation. They can, however, successfully answer a question about reasons, provided it was asked by someone who is aware of the broader context. The answer presupposes a background of shared knowledge. While the quest for a reason can be successfully answered by mentioning virtually any kind of property of an action or just any feature of the situation, this observation does not vindicate the claim that non-normative properties can be reason-giving. It leaves the possibility that the cited properties figure as reasons only in circumstances which contain normative or evaluative properties that make it the case that they are reason-giving. I mention this perhaps trivial sounding claim because it will be relevant later.
2. The Conceptual Link.
Pragmatics aside, how can evaluative properties explain reasons? Before I address the question whether non-evaluative properties provide reasons--which is the crucial presumption of BP-1 and BP-2 that I am going to challenge--it will be useful to first explore the option that the modified account adds to the original one: that specific evaluative properties provide reasons (but "being of value" or "being good" does not).
As Scanlon notes, it is part of our conceptual understanding of values that they provide reasons. More specifically, he writes that a person who does not understand that certain actions and attitudes are (in-)appropriate with regard to something that is of value, does not understand the value of the thing itself.
Having recordings of Beethoven's late quartets played in the elevators, hallways, and restrooms of an office building ... would show a failure to understand the value of music of this kind. What I am suggesting is not that this would show a lack of respect for this music, but rather that it shows a lack of understanding of what one should expect from it, and in what way it is worth attending to. (13)
A person who understands the value of a work of art will understand that its value provides reasons for attending to it in certain ways, but not in others. Those reasons are provided by the evaluative features of the music itself: They are non-derivative reasons. A reason is non-derivative if its being a reason need not be determined by there being other reasons and its relation to them. (This leaves space for the possibility that if there is a reason to [phi], there may always necessarily be a reason to [psi] without the reason to [psi] deriving from the reason to [phi]. If you have a reason to lend your car to your neighbor (say), you may also have a reason to make sure that it is in good working condition. But the reason for lending the car does not derive from the reason for making sure that it works, or the other way around. After all, you have reason to look after your car because of your own safety, and not just because of your neighbor's.) The claim which I will call "the conceptual link" is this:
The Conceptual Link. The (even partial) understanding of any evaluative concept requires understanding some of the non-derivative reasons that the evaluative property that the concept refers to provides. (14)
Note that the Conceptual Link thesis is readily available to buck-passing versions BP-2 and BP-3 (15) (and, as the quotation above shows, Scanlon himself appears to subscribe to it), except that the buck-passer denies that it applies to the most general property of being good or being of value itself. I will therefore not rely on it with regard to being good or being of value until the final section (section III) when I will explain why it applies to that property as well. While being consistent with buck-passing, the thesis is shared by philosophers who do not accept any version of the buck-passing account. (16) Being common ground among both defenders and opponents of buck-passing I will not argue for the thesis but take it as my starting point. However, some clarifications may be useful. As mentioned, I will discuss the implications of the thesis only with regard to the modified account and its claim that specific evaluative properties provide reasons (whereas goodness does not).
But, you may object, the reasons which are provided by the properties of an artwork (say) are conditional reasons, i.e. they are reasons to attend to it only for a person who in fact enjoys a certain kind of music. Does not that make them derivative reasons, because there would not be a reason to attend to the music if there were not a reason to enjoy oneself?
As an objection to the Conceptual Link, the claim incorporates two steps: (1) the reasons for attending to something which is of value are conditional reasons, and (2) they are derivative reasons in virtue of being conditional. I will investigate these steps in turn.
Does Beethoven's music provide only conditional reasons? If I do not care about classical music, if I get nothing out of listening to it, I hardly have a reason to put myself through the ordeal of sitting through a concert (say). Yet, a person who loves this music may well have such a reason. Is "enjoying the music" a condition for having a reason to attend to this kind of music in the first place? There is some plausibility to this, as it accommodates some of the ways in which we reason about values. If someone hates a certain kind of activity, there often seems little point in making her do it. In what way is the ability to enjoy oneself in doing something a condition of having a reason? It is, I will suggest, not a condition in the normal sense in which there may be conditions on having reasons. If there will be frost tonight, I have a reason to bring in the pot plants from my back garden. The occurrence of frost is a condition on having a reason to bring in the plants. (17) But enjoyment during a concert is not quite like this. It is not in the same sense that if I will enjoy myself, I have reason to go to the concert. After all, I may enjoy myself during the concert because I like watching other people or because I brought some popcorn. But this is not the relevant kind of enjoyment. The value of the music provides me with a reason to go to the concert only if I enjoy myself in listening to the music. But, if so, it seems that rather than the enjoyment being a condition of the reason to go, it is part of what I have reason to do: I have reason to listen to the music in a certain way, namely with enjoyment. The adverbial construction brings to the fore that enjoyment is, in this case, only a qualification of the relevant kind of listening and not an independent condition. The reason provided by the value of the music is a reason for a certain kind of listening. Unlike the overnight frost which occurs whether or not I keep my plants in the back garden, the relevant enjoyment at the concert is not any enjoyment that may occur whether or not I listen to the music, but only the enjoyment in listening to it. Therefore, the reason in question is an unconditional reason to listen to the music with enjoyment, and not a conditional reason after all. But if it is an unconditional reason, it clearly is not a derivative one.
But possibly there are other conditions. Perhaps I have a reason to prevent the destruction of an artwork (say) only if I know (or at least could have known) that there is a threat that it may be destroyed. But while being conditional, the reason to prevent the destruction does not derive from my knowledge nor does it appear to derive from any other reason. Thus, conditional reasons are not ipso facto derivative ones. Being derivative is a special case of being conditional, (18) and there is no reason to assume that all (or even any) of the reasons that are provided by evaluative properties are derivative ones.
The upshot of this discussion is that understanding something which is of value requires understanding at least some of the non-derivative reasons for action that it provides. Thus, there is a conceptual link between understanding values and understanding reasons, which allows us to explain how evaluative properties provide reasons: the conceptual understanding of these properties is of them as being reason-giving. A person who does not understand why there are reasons not to be cruel (say) does not understand the concept of cruelty. If the property that "cruel" refers to can be instantiated at all, then there are reasons not to be cruel.
The Conceptual Link also helps with the pragmatics example above since it explains which features of the circumstances make it the case that there is a reason to act in certain ways: If the circumstances include evaluative (or normative) features which are necessarily reason-providing, the structure becomes clear, and we begin to understand why "it is almost midnight now" can be Bert's reason for leaving the party. It derives from his reason to keep his promise, which in turn will be explained by the value of promise-keeping.
There is a common way of specifying some of the reasons that are provided by evaluative properties: They provide reasons, first, for preserving whatever has the property, and preventing its destruction, and, occasionally, and secondly for choosing it or seeking it out. Hence, there are some general descriptions of the reasons that evaluative properties provide. Knowing in this general and vague way which reasons certain evaluative properties provide is a matter of understanding (even partially) the evaluative concepts that refer to those properties.
3. Normative Significance.
Thus far, the buck-passer of the BP-2 variety can agree (I hope). But the buck-passing account rests on a further presupposition. Both BP-1 and BP-2 require that non-evaluative properties too are reason-giving. If they are, can we come up with a similar explanation how and why they are? Is it part of the conceptual understanding of some, if not of all, non-evaluative properties that they are reason-giving too? Surely, not all properties provide non-derivative reasons for actions. The initial discussion about the pragmatics of citing certain considerations as reasons has already shown that: even though they can be cited as reasons in certain circumstances, they are not by themselves reasons. But possibly there are some non-evaluative (non-normative) properties that do provide non-derivative reasons. Take as an example Scanlon's proposed reason for admiring or fostering a research project "because it sheds light on the causes of cancer." As opposed to being pleasant, "shedding light on the causes of cancer" is a non-evaluative property. Does it provide a reason? We could again tell a story about pragmatic considerations and circumstances like the following: If someone did not quite understand why "casting light on the causes of cancer" makes a research project admirable, we could go on explaining: "Don't you see, cancer is a terrible illness--and basically everything that promises progress towards curing it is in that respect a good thing"--thereby making explicit the evaluative aspects of the example, which are normally implicitly understood. But we could not proceed explaining, "Don't you understand? The research was conducted by a team of three people"--which, let's stipulate, is another of its non-evaluative features. That will not help to dispel the puzzlement. Generalizing from this, we could then say: Whenever non-evaluative properties are cited as reasons, they are reasons in the instant circumstances because those include some evaluative features. This analysis of the example would take it to be another case where the consideration cited is not by itself a reason, but figures as a reason only in conjunction with evaluative features which form part of the circumstances. But this move seems less obvious and less convincing here than it was in the party example above. Furthermore, the generalized claim would be warranted only if we had reason to believe that non-evaluative properties cannot provide reasons except when this is due to the presence of some evaluative feature. But this, precisely, is the question--and nothing said so far allows answering it.
In what follows I want to explore the possibility that some, if not all, non-evaluative properties provide reasons. Let's call this kind of non-evaluative properties, borrowing a term of Parfit's, "normatively significant properties" (19)--just to distinguish them from reason-providing properties in the more general sense, which includes evaluative or normative ones as well. Normatively significant properties are those that provide reasons without being normative. Supposedly, "jumping into the canal as the only way to save my life" (20) (for example, when the building is on fire) is such a property. It provides me with a reason to perform any action that has this property.
At first blush it may seem odd to say that reason-providing properties can be non-normative. As mentioned earlier, on a common understanding of normativity, being reason-providing is being normative. But the puzzlement dissolves once we allow for a distinction between "being a reason" and "being a reason for": reasons understood as reason-providing properties and reasons as the relation of there being a reason for a person to do something (in the case of reasons for action). Ordinary language does not always distinguish between the two. But the distinction helps to understand why normatively significant facts may well be natural, non-normative ones. The fact that something has a normatively significant property is one of the relata. But it is the relation of "being a reason for" that is normative (and, possibly, irreducibly so). This suggestion allows us to make sense of the idea of being a realist about reasons without assuming that reason-providing properties need to be normative. The distinction is sound and useful. But of course it does not settle the question whether non-normative (non-evaluative) properties provide reasons. It only helps to understand in which sense they do, if they do.
Do they provide reasons then? How about the example? If I had a strong or even conclusive reason to commit suicide, the fact that jumping into the canal would be the only way to save my life would then be a reason against any action which has this property. If I have a reason to commit suicide, but also a reason to stay alive, the fact that jumping is the only way to save my life would count both in favor and against jumping at the same time. This paradoxical sounding implication may make it doubtful that "jumping ..." by itself is a reason. What we are missing here is the possibility of explaining how a normatively significant property provides reasons in the way that the Conceptual Link made possible explaining how evaluative properties do.
Nevertheless, calling "jumping into the canal as the only way to save my life" normatively significant seems adequate in a different sense: (21) After all, the situations where this property will not be highly relevant in deciding what to do are extremely rare. It sticks out as the one feature of the situation that matters most to most people. By contrast, the time of day is in general a very insignificant feature of the circumstances, and becomes relevant only in very circumscribed situations (as in my pragmatics example).
This contrast between Parfit's example and the pragmatics example is easily explained. Being alive is the most fundamental condition for having reasons in the first place. Continuing one's life is in most cases a condition of being able to comply with one's reasons, such as reasons to pursue one's projects and relationships (provided they are worth pursuing). As Bernard Williams pointed out, (22) some reasons to pursue one's goals or projects are conditional reasons in the sense that they are reasons for specific actions only if one happens to be alive at the relevant time. But a person may also have goals or projects that provide her with reasons to stay alive in order to pursue them. If so, and if a person has any reasons (other than reasons to commit suicide), at least some of them may ipso facto be reasons to stay alive, as a condition of complying with them. Only a person who not only has a reason to commit suicide, or even a conclusive reason to do so, but no reason whatsoever to continue with her life, will have no reason to jump. In other words: Everyone who has any reason to continue living (however forceful her reasons for not doing so may be) has a reason to jump. (23) This is why "jumping ..." is normatively significant: Almost everyone will at least have a reason to do so. There are other facts which are normatively significant in a similar sense, such as the availability of food and shelter as conditions of staying alive.
But if this is right, then the difference between the pragmatics example and normative significance is not a principled one. Just as "leaving the party now as my only way to get home in time" provides a reason only on the condition that I have a reason to get home at a certain time, "jumping into the canal as the only way to save my life" provides a reason on the condition that I have reason to continue my life. What makes "jumping ..." normatively significant is that it is a default assumption that the condition is satisfied: We will always assume that a person has reason to jump, unless there is firm evidence to the contrary. Another difference is that "jumping ..." will appear relevant to anyone who thinks about the reasons pertinent to deciding what to do in the situation in question. "Jumping ..."--unlike "it is almost midnight"--is significant without presupposing shared knowledge about any specific features of the situation. It does, however, presuppose common knowledge about human life and the role of reasons within it. The upshot of this is that normatively significant properties by themselves do not provide reasons. Like pragmatic considerations, they are at best derivative reasons or aspects or parts of a more complete reason.
Assume that I am wrong about this. Assume that normative significance cannot be understood on the analogy with stating some propositions as reasons when there is shared background knowledge of further reason-providing considerations. We are then left with the claim that certain properties provide reasons, where this cannot be explained by the Conceptual Link, since they do not provide reasons in all circumstances. But can we explain when there is a reason to jump and when there is not without adverting to the evaluative properties that continuing one's life instantiates? I take this to be the challenge that a defender of the claim that non-evaluative properties provide reasons has to answer.
She may of course refuse to do so, insisting that certain facts just are reasons, and there is nothing more that can be said. But while explanations may come to an end, it seems quite unsatisfactory to stop at a point where we have properties that sometimes provide reasons and sometimes do not. It should be possible to explain why and when it is the one or the other. But if the buck-passer tries to do so, she may run into a regress.
She may claim that "jumping ..." provides a reason only in appropriate circumstances. Hence, the circumstances determine what it is a reason for, and whether it is a reason for or against certain kinds of actions. But how do the circumstances do this?
a) One possibility is the one that I proposed: The circumstances are relevant because they include certain evaluative properties. (24) E.g., the circumstances might include that the person has reason to continue with her life because it is well worth living.
b) Another possibility is that the circumstances settle the case, not because they include evaluative properties, but rather further non-evaluative ones.
Obviously, in order to defend that non-evaluative properties provide reasons, one has to rely on (b) alone. (25) Let's assume that the non-evaluative features F and G are part of the circumstances which determine whether a person (call her Claudia) has a reason to jump. Let F be that her long-term relationship broke up, G that she lost her job on the very same day, and P (as before) that jumping into the canal is the only way to save her life. Does Claudia have a reason to jump? There seems no way to answer the question (yet). Might P in the presence of F and G not be a reason? Possibly. If so, there would be no reason to jump whenever P, F and G. But now imagine that Claudia's relationship was an abusive one, and that she was continuing her job despite the fact that the work was boring, repetitive and badly paid, only because the relationship had been wearing her out to the point that she simply could not muster the effort of applying for a more rewarding job. In this case, while she may have contemplated not to jump in the absence of F and G, it seems that their presence ensures that she has a reason to jump. However, it is not the conjunction of P, F and G that settles the question whether Claudia has a reason to jump, but the further (evaluative) facts that her relationship was abusive and her work unrewarding. But if P in circumstances which contain F and G were a reason to jump, it would have to be the very same reason, whenever P is present in conjunction with F and G. In other words, it is necessarily the case that P (in those circumstances) is a reason to [phi]. It follows that P, if F and G, is not a reason. However, if present together with a more complex combination of features of the circumstances, P may be a