1 Reasons and balancing

Deciding what you should do very often involves considering various trade offs. Different options have different costs and benefits. Ross (1930) explained this idea, at least in the moral case, in terms of prima facie duties. It is much more common these days to talk about pro tanto reasons bearing on the options. What you should do, according to this approach to normative theorizing, is determined by the reasons bearing on the options open to you. In particular, it’s determined by competition between these reasons. This competition between reasons is how we understand the trade offs that are ubiquitous in thinking about what to do.

Though the idea that what you ought to do is determined by competition between reasons is widespread, there has been very little work on actually developing the details.Footnote 1 The most common approach is to appeal to a metaphor of balancing or weighing.Footnote 2 We imagine a scale with pans corresponding to the options. The reasons for those options are represented by weighted marbles that are placed on the pan corresponding to the relevant option. The option you ought to choose is the one corresponding to the lowest pan—since it has the most weight in it—once all the reasons are taken into account.

This metaphor highlights at least three very important features of the competition between reasons. First, just as the marbles can be of different weights, some reasons are weightier or more important than others. Second, we see that reasons for one option compete with reasons for other options, since as one pan goes down, the others go up. Third, multiple reasons for a single option can combine to support the option more strongly than either does individually, and so can combine to outweigh a relatively weighty reason for a competing option, just as multiple lighter marbles in one pan can outweigh a single heavier marble in another.

But everyone knows that this is just a metaphor, and that it is incomplete in several ways. For example, it suggests that the strengths or weights of multiple reasons for a given option will combine in a strictly additive way, like the weights of marbles. This is most likely false. It also suggests that the weights of all reasons can be compared; but many people believe in incomparability or incommensurability of certain reasons. These issues are widely recognized, and do not tend to lead philosophers writing about reasons astray. The thought is that when we are able to compare some reasons, they compete in much the same way as marbles on a scale: the reasons provide support for the relevant options, and some will provide more support, and thus win the competition.

In this paper I will focus on a different way in which this picture is incomplete, and which I believe may lead us astray in our theorizing about reasons. Following Scanlon (1998) and many others, we know that a reason for an option is a consideration that counts in favor of that option. The reasons for an option are very naturally represented by marbles that go on the pan corresponding to that option. But what you ought to do depends on more than the reasons for the options open to you. It also depends on the reasons against those options. The problem is that there is no equally natural way to represent reasons against an option, within the balancing metaphor. I believe this observation, plus a commitment to something like the balancing metaphor (while recognizing its limitations, of course), explains why there has been so little work on reasons against, as opposed to reasons for, despite the recent explosion of work on reasons. So the first point I want to make is the simple one that it is important to account for reasons against in our theorizing about reasons, and that relying too heavily on the balancing metaphor makes it easy to overlook them.

The view I focus on in this paper is one that is endorsed by some of the small number of philosophers who have thought about reasons against, and more importantly is a very natural one to come to, if you start with the balancing metaphor and then try to accommodate reasons against. This is the simple view that reasons against are really just more reasons for. In particular, this view holds that reasons against an option A just are reasons for certain alternatives to A. Reasons against A count against A by counting in favor of alternatives to A. The central attraction of this view—put in terms of the metaphor—is that it lets us straightforwardly represent reasons against: we already know how to think about reasons for, as weighted marbles that go on the relevant pans.

I have raised the main issue for this paper, and will often talk, in terms of the balancing metaphor, because doing so makes the discussion intuitive and easy to follow. But it is important to emphasize that my primary interest here is in giving an adequate (partial) theory of the metaphysics of reasons, not in criticizing the balancing metaphor or finding a replacement. That said, what we ultimately want is a theory of the competition between reasons, which the balancing metaphor is meant to represent. The significance of this paper for the larger project is clear: if I am correct that reasons against are distinct from reasons for, then this theory will need to include an account of how reasons against factor into the competition between reasons. The few authors who have started working out theories about the competition between reasons have not shown how to accommodate reasons against, as distinct from reasons for.Footnote 3

So in less metaphorical terms, the central attraction of the simple view is that it simplifies our theory of reasons in two ways. First, we can get by with only one reason relation—the reason for relation. Second, we only need to make sense of one kind of competition between reasons—the competition between reasons for conflicting options. If we reject this simple view, on the other hand, we will need an account of the reason against relation in addition to the reason for relation. We will also need to understand at least two different kinds of competition between reasons: (i) the competition between reasons for competing options (e.g., the wine is complex but the beer is refreshing), and (ii) the competition between the reasons for and reasons against a particular option (e.g., the burger is tasty but it is also unhealthy). To come to a judgment about what we ought to do, all things considered, we will presumably also need some way of aggregating the results of these two kinds of competition. This is because the reasons against our options are as important as the reasons for them. The simple view that reasons against one option are just reasons for alternatives is much simpler.

In describing this simple view, I have been deliberately vague in saying that reasons against A are reasons for “alternatives to A”. To get a theory we can evaluate, we need to spell out this idea in more detail. In the next section, I will consider four different implementations and argue against each of them. These different implementations say which alternatives to A the reasons against A are reasons for. In Sect. 3, I will consider a way of rehabilitating two of these implementations by appealing to reasons transmission—when reasons for one option transmit to other, related options. I’ll argue that this is unsuccessful. Finally, in Sect. 4, I will sketch a competing view according to which reasons against are distinct from reasons for. On this view, there are (at least) two distinct normative reason relations: the reason for relation and the reason against relation. I’ll show that it can explain the initial appeal of some implementations of the simple view without inheriting their problems.

2 Reasons against as reasons for

The simple view says that reasons against are really just more reasons for; in particular, reasons against A are reasons for alternatives to A. In this section, I argue against four ways of spelling out this idea by specifying which alternatives the reasons against A are reasons for. These are: (i) reasons against A as reasons for some alternative to A, (ii) reasons against A as reasons for ¬A, (iii) reasons against A as reasons for the disjunction of alternatives to A, and (iv) reasons against A as reasons for each alternative to A.

An important point to keep in mind is that the versions of the simple view I consider here say that reasons against just are certain kinds of reasons for. This is a stronger claim than one which says merely that reasons against an option are also reasons for certain alternatives. This latter claim is consistent with the rejection of the simple view, since it is consistent with a view (like the one I will defend) on which the reason against relation is distinct from and not reducible to the reason for relation.

2.1 Implementation 1: reasons for an alternative

The first implementation is that a reason against A is just a reason for some alternative to A. So the set of reasons against A consists of the union of the sets of the reasons for B, reasons for C, reasons for D, and so on. If we begin with the balancing metaphor, this is a natural first thought. Recall that reasons for an option support it by adding weight to its pan, and thus lowering the pan. An option wins this competition if it corresponds to the lowest pan, once all the reasons have been taken into consideration. A crucial part of the metaphorical balance (and of real balances) is that as one pan goes down, the others go up. If winning the competition requires corresponding to the lowest pan, then things that make a pan go up make it harder for the corresponding option to win the competition, and thus count against it. So it is natural, at least given the balancing metaphor, to think of reasons against A—considerations that count against A—as reasons for an alternative to A.

Moreover, and relatedly, this implementation is closely related to the commonsense idea of an opportunity cost. Among the reasons against an option are the good things you’ll miss out on by failing to do the incompatible alternatives. For example, the high salary that comes with being a lawyer is a reason to be a lawyer. It is also intuitively a reason against being a philosopher, instead, since if you’re a philosopher (and not a lawyer) you’ll miss out on the high salary that comes with being a lawyer.Footnote 4

Nevertheless, this implementation is unsatisfactory as an account of reasons against. As Greenspan (2005) emphasizes in a slightly different context in her defense of the sort of view I defend in this paper, when we consider paradigmatic reasons against options—the long hours that come with being a lawyer, the long wait times at a particular restaurant, the high prices of the coffee at that shop—they seem to target the options they are reasons against. Reasons against A highlight some bad (in some sense) feature of A. But reasons for an alternative to A do not necessarily target A in this way. This is easiest to see if we think in terms of the balancing metaphor (which, recall, the defender of the simple view aims to maintain). Suppose the choice is between A, B, C, and D. Now consider a reason for one of the options, say B. That is represented in the metaphor by a counter in the B pan. As we saw above, this does count against A in a sense, since by pushing the B pan downwards, it raises the A pan. But note that it also raises the C and D pans. So the reason for B does not seem to target A anymore than it targets C or D.

To move away from the metaphor, consider one of the paradigmatic reasons against mentioned above: Northpoint has very long wait times, and that’s a reason against going there. There may be several other options, including BeThai and the One Under. If this reason against going to Northpoint is just a reason for one of the alternatives, say, for BeThai, then it seems to count against the One Under just as much as it counts against Northpoint. But the fact that Northpoint has very long wait times doesn’t seem to count against the One Under. So we lose the sense in which at least some reasons against an option seem to target that option. Thus, I do not think this implementation is ultimately a very attractive one.

2.2 Implementation 2: reasons for the negation

The second implementation holds that a reason against A just is a reason for not doing A, or for ¬A. This may be even more natural than the first implementation. In fact, of the few philosophers who have suggested accounts of reasons against, this is the most popular view.Footnote 5

The primary advantage of this implementation over the first is that we can easily explain the sense in which reasons against A target A. What the reason counts in favor of, on this view, is not doing A. So in the restaurant case from above, the fact that Northpoint has very long wait times is, on this view, a reason for not going to Northpoint. This seems to much more clearly target the option of going to Northpoint than a reason for going to, say, the One Under. In particular, unlike a reason for going to the One Under, the reason for not going to Northpoint appears to treat going to Northpoint and going to BeThai asymmetrically.

Nevertheless, I don’t think this implementation is, so far, satisfactory. This is because many of our choices are more fine-grained than simply a choice between doing an option A and not doing it, or doing ¬A. Often we have to choose between A, B, C, D, and so on. The problem is that we do not yet know how a reason for ¬A bears on these more fine-grained choices between A and several incompatible alternatives. But reasons against A are clearly relevant to such choices: they count against doing A. According to the simple view, for these considerations to be reasons against A just is for them to be reasons for alternatives to A. That is, for the considerations to count against A just is for them to support alternatives to A. The problem for Implementation 2 is that in fine-grained choices, like that between A, B, C, and D, the action for which these considerations are supposed to be reasons for—¬A—is not actually one of the alternatives. So Implementation 2 lacks an account of how reasons against A bear on these more fine-grained choices.Footnote 6

Though this second implementation nicely captures the sense in which reasons against A target A, it does not yet explain how reasons against A are relevant for more fine-grained choices between A and several incompatible alternatives. In Sect. 3 I will consider a way to respond to this problem, but first I will consider two other potential implementations of the simple view.

2.3 Implementation 3: reasons for the disjunction of alternatives

Another natural implementation of the simple view is that reasons against A are just reasons for the disjunction of the alternatives to A. Like Implementation 2—that reasons against A are reasons for ¬A—and unlike Implementation 1—that reasons against A are reasons for some particular alternative to A—this implementation explains the way in which reasons against A seems to target A. It does this because A is treated asymmetrically from all the other options: it is the only option not included in the disjunction for which the consideration is a reason.

But like Implementation 2, this view faces the important question of how reasons against A bear on relatively fine-grained choices, like the choice between A, B, C, and D. Implementation 2 treats reasons against A as reasons for ¬A. But since ¬A isn’t an option (the choice is more fine-grained than that), we need some way of understanding how this reason against A bears on this choice. The implementation under consideration now faces this same question. Reasons against A are supposed to be identified with reasons for the disjunction, BCD. But just as ¬A isn’t an option, neither is BCD. So a reason for this disjunction does not appear to bear on the choice between A, B, C, and D. In Sect. 3, I will consider such an account, but argue that it fails.

2.4 Implementation 4: reasons for each of the alternatives

The final implementation of the simple view I want to consider is that reasons against an option are reasons for each of the alternatives. Implementation 1 holds that reasons against an option are reasons for any alternative and Implementations 2 and 3 hold that reasons against an option are reasons for a particular alternative—the negation of that option, or the disjunction of alternatives, respectively. The first does not explain how reasons against an option target that option. The second and third do not explain how reasons against an option bear on relatively fine-grained choices between that option and several alternatives to it.

This final implementation avoids both of those problems. Reasons against A target A because it is the only option that these reasons are not reasons for. In terms of the balancing metaphor, such reasons go on every pan except the A pan. This also lets us see how these reasons bear on choices between A and a number of alternatives. They support each of the individual alternatives, and not just their disjunction.

Nevertheless, this implementation faces another serious problem. To appreciate it, I need to briefly say something general about how I am thinking about reasons for action. On the picture I prefer—and which many others accept, as well—reasons are provided or explained by what I call objectives. What the objectives are will depend on which more substantive theory of reasons is correct. On Humean, individual desire-based views, objectives are the desires (or objects of the desires) of the relevant agent. On ‘democratic’ Humean views like that in Manne (2016), the objectives are desires of either the agent herself or of others. On value-based views, objectives are values like justice, honesty, friendship, happiness, or even a sui generis property of goodness. When a given action stands in a specific relationship to one of these objectives, that objective provides or explains reasons bearing on the action. On desire-based views, as well as some value-based views, this relationship is usually taken to be promotion. Actions that promote the objects of your desires are the actions for which you have reasons. On some value-based views, this relationship may be something like respect or honor.Footnote 7 When some action would appropriately respect or honor a given value, that value provides or explains reasons for the action. Most often, views which fit this broad pattern appeal to explanation in giving their constitutive accounts of when some particular consideration is a reason bearing on a given action. One straightforward version of this picture says that when some consideration r explains why performing action A would promote/respect objective O, then r is a reason for A, explained by O.Footnote 8

On this picture, reasons for and against actions are explained by objectives. If we accept it, then it’s natural to think that just as reasons for A are explained by objectives with respect to which A does well, in terms of promotion or respect, reasons against A are explained by objectives with respect to which A does poorly. The relevance for evaluating Implementation 4 of the simple view, that reasons against A are reasons for each of the alternatives to A, is that it seems that we can have reasons against A, explained by an objective, even if some other option does even worse with respect to this objective. In such cases, it is not at all plausible that the reason against A is a reason for this other, inferior (as far as the objective in question goes) option. But Implementation 4 holds that a reason against A is a reason for all of the alternatives.

Here is another version of the restaurant case from above, to illustrate. Suppose I am trying to decide where to go for lunch, and am deciding between three restaurants: Northpoint, the One Under, and BeThai. Northpoint is very crowded at this time of day. That’s a reason against going to Northpoint. According to Implementation 4, this reason against is really just a reason for each of the alternatives. That is, it is a reason for going to the One Under and a reason for going to BeThai. But suppose that, while BeThai is not crowded at all, the One Under is even more crowded than Northpoint. In this case, it is not plausible that the fact that Northpoint is so crowded is a reason for going to the One Under. Nevertheless, the One Under being even more crowded does not seem to prevent the crowdedness of Northpoint from being a reason against going there, especially given that there is a non-crowded option, namely BeThai. So we should reject Implementation 4.

We may question whether this is really a problem for Implementation 4.Footnote 9 That Northpoint is crowded is a reason against going to Northpoint, and that the One Under is even more crowded is a reason against going to the One Under. Very plausibly, the reason against going to the One Under is stronger than the reason against going to Northpoint. According to Implementation 4, then, though the fact that Northpoint is crowded is a reason for going to the One Under, the fact that the One Under is even more crowded is an even stronger reason for going to Northpoint. Thus, we will not get the result, on Implementation 4, that we ought to go to the One Under, assuming that crowdedness is all that matters.Footnote 10 The defender of Implementation 4 must reject the claim that the fact that Northpoint is crowded is a reason against going to Northpoint but not a reason for going to the One Under. But she can explain why considerations of crowdedness nevertheless favor Northpoint over the One Under.

If all we wanted out of a theory of reasons was a way to generate the correct rankings of options, then Implementation 4 might be adequate. But though this is an important task for a theory of reasons, I think we want more than this. In particular, our theory of reasons should explain the claims we are willing to accept about reasons. One such claim, I take it, is just the claim I appealed to in the argument, that if the One Under is more crowded than Northpoint, then the fact that Northpoint is crowded cannot be a reason for going to the One Under.Footnote 11 Relatedly, we ordinarily think of reasons for an option as things that tend to make that option more rational or choiceworthy than it would otherwise be.Footnote 12 But the fact that Northpoint is crowded is not the kind of thing that can make going to the One Under, which is even more crowded, more rational than it would otherwise be. These may not be decisive points, since most philosophical theories will have to reject some ordinary claims we want to make. But I find these claims sufficiently compelling that I count it as a serious mark against Implementation 4 that it must reject the them. As we’ll see, the view I suggest can capture them easily.

2.5 Summing up

To sum up this section, I have argued against four natural interpretations of the simple view that reasons against are really just certain kinds of reasons for. Though this view, in one or another of these implementations, would simplify our theory and allow us to maintain the attractive balancing metaphor for the competition between reasons, it appears that actually spelling out the simple idea is a serious challenge.

By seeing the problems that face each of the implementations, we can draw out three criteria for a successful theory of reasons against. First, we saw from the failure of Implementation 1 that at least paradigmatic cases of reasons against an option need to somehow target that option: they should bear on that option differently than they bear on the alternatives.

Second, we saw from the challenge facing Implementations 2 and 3 that reasons against an option need to bear on relatively fine-grained choices between that option and a number of alternatives. It is not enough to simply hold that reasons against A are reasons for ¬A, or for the disjunction of alternatives to A, if those relatively coarse-grained actions are not options in the choice. Since reasons against an option are relevant for these fine-grained choices, any theory of reasons against must say how they are relevant.

Finally, we saw from the failure of Implementation 4 that reasons against A cannot be reasons for each alternative, since some of those alternatives may do even worse with respect to the objective that explains why the consideration is a reason against A. So if reasons against one option are reasons for alternatives—even if they are not simply identified with reasons for alternatives, as the simple view holds—they cannot be reasons for every alternative.

At the end of the paper, I will sketch a theory of reasons for and against that satisfies these three criteria by abandoning the simple view, and treating reasons against as distinct from reasons for. First, though, in the next section I will consider one way to answer the challenge facing Implementations 2 and 3, of saying how reasons against an option bear on relatively fine-grained choices.

3 Reasons transmission

The challenge facing Implementations 2 and 3 is to say how reasons against A bear on relatively fine-grained choices between A and some number of alternatives, like a choice between A, B, C, and D. Many of our choices are like this; just think of choosing between universities, restaurants, vacation destinations, or even just plans for how to organize your day. In choosing between Northpoint, BeThai, and the One Under, the fact that Northpoint is crowded is a reason against going to Northpoint. According to Implementation 2, this is just a reason for not going to Northpoint. But this doesn’t yet tell me how to factor this reason into the choice I face. According to Implementation 3, the fact that Northpoint is crowded is a reason for either going to BeThai or going to the One Under. But again, I don’t know yet how to factor this into the choice I face. On the simple view, the fact that Northpoint is crowded is supposed to factor in by supporting alternatives to going to Northpoint, but the alternatives identified by Implementations 2 and 3 are not among the options in these more fine-grained choice situations.

A plausible thought here is that the reason against going to Northpoint bears on these other options derivatively. That is, they are reasons for those other options, if they are, because they are reasons against going to Northpoint. Recently, there has been lots of work on the ways in which reasons for one option transmit to become derivative reasons for other, related options. That is, there has been lots of work formulating transmission principles for reasons. So the suggestion I’ll explore in this section is that reasons against A bear on relatively fine-grained choices between, say, A, B, C, and D derivatively, via a transmission principle, because they are non-derivative reasons for either ¬A (on Implementation 2) or BCD (on Implementation 3).

If we’re sticking with the simple view, that reasons against are really just a certain kind of reason for, then it seems that reasons against A must bear on these fine-grained choices by supporting alternatives to A. Notice that the individual alternatives to A are ways of not doing A, or ways of performing the disjunctive action. A simple and at least initially attractive example of a general transmission principle for reasons for that tells us how reasons transmit from actions to ways of carrying out those actions is the following:

Facilitative Principle (FP): if r is a non-derivative reason for A, and doing B facilitates doing A, then r is a reason for B.

By calling the reason for A ‘non-derivative’, I just mean that it is a reason for A, and not because it is a reason for anything else. Derivative reasons, on the other hand, are reasons for an option only because they are reasons for something else. By saying that doing B ‘facilitates’ doing A, I mean something quite broad. It is meant to include taking the means to doing A, so that starting your car facilitates going to the store. But it also is also meant to include relations of constitution: driving to the store isn’t naturally described as a means to going to the store; rather, it constitutes going to the store. This kind of constitution is what is most directly relevant for thinking about how reasons for ¬A or reasons for BCD could bear on choices between A, B, C, and D, since performing one of these individual alternatives constitutes performing either ¬A or the disjunction of alternatives.

This principle explains many plausible cases of transmission. For example, the fact that it’s your birthday is a reason for me to bake you a cake, given the birthday-related conventions of our society. We can suppose that this is a non-derivative reason. Suppose I am deciding what to bake for you, and my choices are baking you a chocolate cake or baking you chocolate cookies. The fact that it’s your birthday is a reason for me to bake you a chocolate cake. This seems to be because this fact is a reason to bake you a cake, and baking you a chocolate cake constitutes baking you a cake. More generally, the FP allows us to explain how reasons for relatively coarse-grained options support more fine-grained ways of carrying them out, over options that are incompatible with the coarse-grained option.Footnote 13

If we supplement either Implementation 2 or Implementation 3 with the FP, then we can see how reasons against A bear on more fine-grained choices. According to these implementations, the reasons against A are really just reasons for relatively coarse-grained alternatives to A—either ¬A or a disjunction of alternatives. Since performing the individual alternatives facilitates performing these coarse-grained alternatives, by constituting the performance of them, the FP gives us reasons for the individual alternatives. This is how reasons against A bear on these more fine-grained choices.

But, as may already be clear, neither the defender of Implementation 2 nor the defender of Implementation 3 should accept this principle. Combining the principle with either of these implementations gives us the result that non-derivative reasons against A are reasons for each alternative to A, in the more fine-grained choices. This is because each of these alternatives is both a way of not doing A and a way of doing the disjunction of alternatives. So by adding the FP in order to explain how reasons against A could bear on more fine-grained choices, both Implementation 2 and Implementation 3 collapse into something very close to Implementation 4, and face the same problem. The problem, again, is that some alternatives to A may be worse than A with respect to the objective that explains the reason against A. In these cases, it is not plausible that the reason against A is a reason for these inferior (with respect to the relevant objective) alternatives.Footnote 14

I have only considered one fairly simple transmission principle, and it will most likely need refining to be acceptable, even leaving aside considerations about reasons against and the simple view. Thus, the defender of one of these implementations may try to give more sophisticated transmission principles in order to get around this problem. But since reasons against, on these views, are meant to really just be certain kinds of reasons for, any proposed principle will have to be plausible for reasons for more generally. In particular, the amended principles will still need to explain the kinds of intuitive cases of reasons transmission that motivate such principles in the first place. The argument of this section suggests that formulating a principle that does this while also giving acceptable results when applied to reasons against (as the simple view understands them) may be difficult. I leave that as a task for the defender of the simple view. I do not have a general argument that this can’t be done. Instead, I will now move on to develop an alternative view, on which reasons against are distinct from reasons for.

4 Reasons against as distinct from reasons for

In this section I will sketch a theory of reasons for and reasons against that treats them as distinct. After doing so I will show that this meets the three criteria for a theory of reasons against I listed at the end of Sect. 2, and that it can explain the appeal of the various implementations of the simple view. Finally, I will briefly compare the view suggested here to Greenspan’s (2005) view, which also distinguishes between reasons for and reasons against.

The starting point is the idea, mentioned in Sect. 2, that reasons are provided or explained by the promotion or respect of objectives—usually desires or values. Often views that accept this picture are put in terms of binary notions of promotion or respect. If an option promotes or respects an objective, there are reasons for it; if not, there are not. But I think we do better to put things in comparative terms—in terms of how well a given action promotes or respects the objective, compared to the alternatives.

Here is the argument for this claim, in brief.Footnote 15 Theories that accept a binary promotion or respect condition on reasons struggle to explain cases in which some consideration is a reason for A when the alternative is B, though not a reason for A when the alternative is a different option, C. For example, that I’m trying to get in shape is a reason to cycle to work when the alternative is driving, but not a reason to cycle to work when the alternative is jogging, assuming that jogging is better exercise than cycling. In fact, if the alternative is jogging, then the fact that I’m trying to get in shape is plausibly a reason against cycling to work, since jogging is better exercise. If we accept a binary promotion condition, then for the fact that I’m trying to get in shape to be a reason to cycle to work, when the alternative is driving, cycling to work must promote or respect the relevant objective (for example, my getting in shape). But if it does, then it seems that this consideration would also be a reason to cycle to work when the alternative is jogging—after all, cycling promotes or respects (in binary terms) my getting in shape. On the other hand, if we accept a comparative condition, such that an action has to promote or respect an objective better than the alternatives, we can explain cases like this. Since cycling promotes or respects the relevant objective better than driving, I have a reason to cycle when the alternative is driving. But since it does not promote or respect the relevant objective better than jogging, I do not have a reason to cycle when the alternative is jogging.

So here are the analyses of reasons for and reasons against that I propose:

For: r is a reason for A when r explains (or is part of the explanation) why A promotes/respects some objective better than all of the alternatives.

Against: r is a reason against A when r explains (or is part of the explanation)Footnote 16 why A promotes/respects some objective less well than some alternative.

Recall that the objectives referred to here will depend on the correct substantive theory of reasons. They may be desires, objective values, or something else. There are also important questions about what it takes for one alternative to better promote or respect an objective than another that I cannot address here. For example, a natural idea is that one action better promotes an objective than another if the first makes the instantiation of the objective more probable than the second. But I do not want to assume such a probability-raising view here.Footnote 17

There are variants of this kind of view which make different choices about how many of the alternatives an option has to do better or worse than, with respect to a given objective, for some consideration to be a reason for or against it. My view puts a strong condition on reasons for and a weak condition on reasons against. For some objective to provide a reason for an option, that option has to do the best with respect to the objective. For some objective to provide a reason against an option, that option only has to do worse than some alternative. We could instead hold that there are reasons for an option when it does at least as well as any other option, or when it does better than at least one alternative. Similarly, we could hold that there are reasons against an option only when it does worse than every other alternative. I think that the analyses I give do a nice job of achieving a balance between capturing intuitive claims about reasons and making theoretically attractive predictions, though I cannot defend that claim here.Footnote 18 I will illustrate some of the attractive features of this view below, when I show how it satisfies the three criteria on a successful theory of reasons against that we saw in Sect. 2. I will also address what I take to be the most serious objection to this view at the end of Sect. 4.2. This is the objection that, given For, we do not get reasons for options that do better than all of the other options, but are on a par with one another.

The most important feature of this view for the purposes of this paper is that it allows for reasons against an option that are not reasons for any other alternative. Consider the following variant on a case from above. I am trying to decide where to go for lunch, and the options are Northpoint, the One Under, and BeThai. Once again, Northpoint is very crowded at lunchtime, and BeThai is not at all crowded. But now—unlike in the case above—the One Under isn’t crowded, either. In this case, the fact that Northpoint is very crowded is a reason against going there, explained by (for example) my desire to avoid a long wait, since Northpoint does worse than at least one of the other alternatives (and in fact, worse than both) with respect to this objective. Since neither the One Under nor BeThai are crowded, they do equally well with respect to this objective, and so by For, this objective will not provide reasons for either one. We only get a reason against going to Northpoint.

4.1 Criteria for reasons against

The three criteria on a successful theory of reasons against from Sect. 1 were the following. First, reasons against an option need to target the option they are reasons against, by bearing on it differently than they bear on other alternatives. Second, reasons against an option need to bear on relatively fine-grained choices between that option and a number of alternatives. Third, reasons against an option are not reasons for alternatives that do worse with respect to the objective that explains the reason against. I will now show that the view I’ve sketched here satisfies these criteria.

First, if r is a reason against A, it explains why A does relatively poorly with respect to some objective. In particular, it explains why some other alternative does better than A. Consider another variant of the restaurant case. Northpoint is very crowded, the One Under is also very crowded, though not quite as crowded as Northpoint, and BeThai is not at all crowded. The fact that Northpoint is very crowded is a reason against going there, since this fact explains why going to Northpoint would do worse than at least one other alternative (and in fact, worse than both) with respect to the objective of avoiding long waits. But note that this fact—that Northpoint is very crowded—does not explain why going to the One Under would do worse than some alternative with respect to this objective. This fact doesn’t really have anything to do with the One Under. So the reason against going to Northpoint—that Northpoint is very crowded—does seem to target that option.

But things are actually not quite so straightforward. Notice that even if some alternative B does better with respect to a given objective than A, it does not follow that that objective does not provide reasons against B. It may be that another alternative, C, does even better than B. It will then follow from Against that the objective provides reasons against B. In the case above, the objective of avoiding a long wait will not only provide reasons against going to Northpoint, but also reasons against going to the One Under. So though the reason itself—that Northpoint is very crowded—targets going to Northpoint, there is a sense in which the objective that provides that reason doesn’t really target Northpoint, since it also provides reasons against going to the One Under. But I believe this is actually the correct result. The objective of avoiding a long wait does seem to provide reasons against going to the One Under, given that there would be a long wait there, and would not be a long wait at BeThai. But we still do get a kind of asymmetry, even at the level of objectives, since for an objective to provide reasons against an alternative, by Against, there must be some alternative that it does not provide reasons against.

Second, it is clear that reasons against can bear on relatively fine-grained choices. The analysis of reasons against makes reference to the alternatives, so no matter how coarse- or fine-grained they are, we can apply the analysis to see what the reasons for and against each of the options are. This holds both for relatively coarse-grained choices, like a choice between A and ¬A or a choice between A and BCD, and more fine-grained choices like a choice between A, B, C, and D.

Finally, reasons against one option, explained by some objective, are not reasons for alternatives that do worse with respect to that objective. Suppose r is a reason against A, explained by objective O. This means that there is some other alternative that does better than A with respect to O, and that r is part of the explanation for this. Now suppose that another alternative, B, does even worse than A with respect to O. Then obviously B does not do the best with respect to O, and so, by For, there will not be reasons for it, provided by O. In particular, the reason against A, r, will not be a reason for B.Footnote 19

So the view I’ve suggested here, on which reasons for and reasons against are distinct, meets the three criteria (or, at least, three of the criteria) for a successful account of reasons against. Now I want to show how this view explains some of the appeal of the various implementations of the simple view.

4.2 Explaining the appeal of the simple view

Part of the appeal of Implementation 1—that reasons against A are just reasons for an alternative to A—comes from the idea of an opportunity cost: that among the costs of an option are the good features of incompatible alternatives that you’ll miss out on. The view I’m defending here captures this idea, but in a way that allows for a distinct category of reasons against. It follows from For and Against that anytime there’s a reason for an option, explained by some objective, there will also be reasons against the alternatives, explained by that objective. From For, if r is a reason against A, explained by O, then A does the best with respect to O, and r helps explain this. This means that the other alternatives do worse than A, and so by Against, O will provide reasons against these other alternatives. And it is plausible that r will help explain this, as well.Footnote 20 But unlike on Implementation 1, we may have other reasons against an option, in addition to those that are also reasons for the alternatives. In particular, if the alternatives to an option A all do better than A, but are on a par with one another, with respect to an objective, then that objective will provide reasons against A but not reasons for any of the other alternatives.

Implementation 2 holds that reasons against A are simply reasons for ¬A. There is certainly something attractive about this picture, but as we’ve seen, it is problematic to simply identify reasons against A with reasons for ¬A. We need to know how these reasons for ¬A can bear on more fine-grained choices, and spelling this out is not straightforward. The view I have suggested here can explain the appeal of this implementation. On this view, reasons against A will also be reasons for ¬A in the coarse-grained choice between A and ¬A. It is easy to see that this follows from For and Against. But as noted above, we also have an account of how these reasons bear on more fine-grained choices, since For and Against can be applied regardless of the granularity of the set of alternatives.

Similar remarks apply to Implementation 3, which held that reasons against A are reasons for the disjunction of alternatives. In the coarse-grained choice between A and BCD, any reason against A will be a reason for the disjunction. But again, we have an account of how these reasons bear on more fine-grained choices.

This observation helps to (at least partially) diffuse one objection to the view I’ve suggested here, based on the strong condition on reasons for.Footnote 21 If two options do equally well with respect to some objective, and better than all of the other alternatives, For tells us that that objective will not provide reasons for either of them. It may seem more plausible to hold that it provides reasons for each of them. I think we can mitigate this counterintuitive consequence. Suppose that A and B do equally well with respect to an objective, and both do better than C. Then in the choice between A, B, and C, the objective will not provide reasons for either A or B. But first, note that it will provide reasons against C, but not reasons against either A or B. Second, note that in the related, more coarse-grained choice between A ∨ B and C, the objective will provide reasons for the disjunction. Finally, the objective will provide reasons for A in a choice only between A and C, and similarly, it will provide reasons for B in a choice only between B and C. This may not completely satisfy the objector, but given the other attractions of this view, illustrated above, I hope that it does significantly reduce the force of the objection.

4.3 Comparison with Greenspan’s view

Before concluding, I will contrast my view with the one defended by Greenspan (2005). She defends a view like mine, according to which reasons against (which she calls “negative reasons”) are distinct from reasons for. In fact, there is a sense in which reasons against are more important for Greenspan. This is because only reasons against can ground a rational (or moral) requirement—reasons for are always optional, in the sense that they can be rationally ignored, even in the absence of reasons for alternatives. On the other hand, if there is a reason against an option but no (or even insufficiently strong) reasons for it, then it is irrational to perform the action. This is part of Greenspan’s criticism-based picture of rationality, according to which rationality is primarily about avoiding or answering criticism, rather than about doing what your reasons most strongly support, as a more traditional view holds. This line of thought forms the basis of Greenspan’s argument for a reasons for/reasons against distinction. In brief, she argues that by holding that only reasons against can ground requirements, we can capture two attractive but seemingly inconsistent thoughts: (i) requirements are based on reasons, and (ii) reasons do not rationally compel. The first is true of reasons against, while the second is true of reasons for.Footnote 22

Unlike Greenspan’s view, my view and arguments are compatible with the more traditional picture. So though I have framed my discussion as a criticism of the balancing metaphor, Greenspan’s view is a more radical departure. Part of what the balancing metaphor is meant to capture is that reasons compete in a way that makes talking in terms of their weight at least approximately correct, which is compatible with my view. But on Greenspan’s view, this is just not what reasons do. Instead, reasons against ground criticisms while reasons for answer those criticisms. Relatedly, though both Greenspan and I distinguish reasons against from reasons for, they differ more radically on Greenspan’s view. Grounding a criticism and answering a criticism are quite different things. On the other hand, on my view, both reasons for and reasons against are explanations of facts about how well an option promotes or respects an objective, in comparison with the alternatives. It is thus easier to see, on my view, why the simple view that reasons against are really just reasons for is so tempting. This is only circumstantial evidence in favor of my view, but it does highlight the difference between it and Greenspan’s view.

5 Conclusion

Theorizing about reasons tends to focus almost exclusively on reasons for, but reasons against options are also important. Relying too heavily on the balancing metaphor makes it easy to overlook reasons against. Here, I have argued against various ways of spelling out the simple and natural idea that reasons against just are a certain kind of reasons for. I also sketched a view on which reasons against are distinct from reasons for, and showed that it provides a more satisfactory account of reasons against.

Though a theory like mine does complicate our theory of reasons by adding a separate reason against relation, the particular version I have offered analyzes the reason for and reason against relations using the same kinds of explanatory resources—primarily, (i) promotion and respect of objectives and (ii) explanation. So the added complexity should not be too worrying in this respect.

The most important consequence of this view is that we may need to complicate our account of the competition between reasons. Reasons for competing options compete with one another, and for this competition, the balancing metaphor is apt. But reasons for one option also compete with reasons against that option. If the arguments of this paper are correct, we cannot straightforwardly subsume this kind of competition to the competition between reasons for competing alternatives. This means that the balancing metaphor is not even approximately correct. Unfortunately, I do not have a replacement picture. But whatever replacement we offer should allow for a distinct role for reasons against. Perhaps it isn’t surprising that the simplistic balancing metaphor is inadequate for fully representing the complex interactions among reasons. Given that there are few detailed accounts even of the competition between reasons for conflicting options, I take it that there is a much larger project of working out how it is that reasons compete to determine what we ought to do. What I am suggesting is that this project should include a distinct account of the competition between the reasons for an option and the reasons against it.