1  

Toward a Principle of Instrumental Transmission1

An action’s being (or seeming to be) a means to another can be normatively significant in two

ways. As Kant observed in the Groundwork, and as the idea has been developed (in markedly

different ways) by Korsgaard 1997, Bratman 1987, and Broome 2002, it can be a matter of

rationality: a matter of the coherence, or unification, of the deliberating agent’s state of mind.

Insofar as one wills an end and believes that something is a necessary means to it, one is, in

refusing to will those means, at odds with oneself. Or perhaps this is putting it too lightly; one’s

action, or very status as an agent, is threatened by disintegration, in more than one sense of the

word.

However, as Raz 2005a and b observes, the normative significance of instrumentality can

also be seen as a matter of the transmission of reasons. Here what is at issue is not the coherence

of the agent’s deliberation, viewed as a psychological episode, but instead the structure of the

subject matter about which the agent is deliberating. If there is reason for one to pursue an end

(whether or not one pursues it),2 and if an action is in fact (or relative to the better information of

an advisor or onlooker3) a means to that end (whether or not one knows it), then it is the case

that, because of this, there is reason for one to adopt the means—quite apart from any

                                                                                                               1 [Identifying note 1] 2 This formulation is agnostic on whether reasons for ends derive from our desiring those ends, or from the relation of those ends to things of independent value. However, desire-based theorists may deny, against Hubin 1999, that their theory is a combination of a principle of instrumental transmission and the principle that reasons for ends are provided by desires. Instead, they may say, there is just one principle, a principle of, if you will, instrumental transmutation: if one desires the end, then one has reason to take the means. They view instrumental transmission as an illusion in the sense of discussed in connection with General Production in section 11: an idea that delivers extensionally correct results, but plays no explanatory role. 3 See note 11 and the preceding text.

2  

considerations about one’s psychological coherence.4 This is reflected in familiar patterns of

advice. If I know that there is reason for you to lower your blood pressure, and if I know that

your taking this medicine will do that, then I am liable to tell you, if I can, that there is reason for

you to take this medicine—whether or not you intend to lower your blood pressure, or know that

this medicine will do that. Furthermore, insofar as what one ought to do varies with the reason

there is for one to do it, a similar sort of transmission of ‘ought’s from ends to means might be

expected. If I know that you ought to lower your blood pressure, I may well conclude that,

because of this, you ought to take the medicine.

This paper asks what principles govern this instrumental transmission of reasons and

‘ought’s. A now burgeoning literature offers us many proposals. The most popular are:

Ought Necessity: If one ought to E, and M-ing is a necessary means to E-ing, then one

ought to M5

and:

                                                                                                               4 See, for example, Bedke 2009, Bratman 2009, Darwall 1983 16, 46–48 and 2006, Hubin 1999, Raz 2005a and 2005b, Kolodny 2007 and 2008b, Price 2008 7, 138–139, Schroeder 2009, Setiya 2007, Vranas ms. (who distinguishes it from the kinds of support relevant to the validity of imperative inference), and Way 2010. Passages in Scanlon 2004 and Wallace 2006 assume or grant its correctness. [Identifying note 2] 5 “If you should do E, all things considered, and doing M is a necessary means to doing E, you should do M, all things considered” (Setiya 2007, 660). “If X objectively ought to do A, and to do A X must do B, it follows that X objectively ought to do B” (Schroeder 2009, 239). And the principle is often implicitly invoked: “because you promised to do A and cannot do A without doing B, you ought to do B” (Korsgaard 2009b, 38).

One might think that the “end-relative” account of ‘ought’ presented by Finlay 2009 and forthcoming is also committed to Ought Necessity, since that account implies that, whatever else one ought to do in a given context, one ought to take necessary means to the “end” specified by that context. However, what is substituted for E in Ought Necessity need not be this contextually specified end itself. If the contextually specified end is, say, maximizing expected value (see Finlay 2009 326 n. 26), then it might be the case for Unlucky, discussed in section 7, that he ought to get a Ph.D. (=E), but not the case that he ought to take the necessary means of applying to graduate school.

3  

Strong Necessity: If there is reason for one to E, and M-ing is a necessary means to E-ing,

then there is at least as much reason for one to M,6

which entails:

Weak Necessity: If there is reason for one to E, and M-ing is a necessary means to E-ing,

then there is some reason for one to M.

However, Ought Sufficiency, Strong Sufficiency, and Weak Sufficiency, which substitute

“sufficient means” for “necessary means,” have also been suggested.7

This paper’s negative aim is to show that these Necessity and Sufficiency principles not

only fail to provide a satisfying theory (section 6), but also are falsified by counterexample. The

first two counterexamples are anticipated, like so much else, by a deceptively casual remark of

Anscombe 1957. In passing, she observes that the intention expressed by, “I do P, so that Q,”

can be “contradicted” either by saying, “But Q won’t happen, even if you do P” or by saying,

“But it will happen whether you do P or not” (36). In what will be my terms, P can fail to inherit

the reason to Q because P fails to “probabilize” Q—Q won’t happen even if you do P—or

because P is “superfluous” with respect to Q—Q will happen whether or not you do P. The basic

problem with Strong and Ought Necessity is that necessary means need not “probabilize” the end

                                                                                                               6 “Reasons for me to make something my end are, owing to the hypothetical imperative, equally reasons for me to take the necessary means to it” (Darwall 1983, 16). “If one has conclusive reason to believe that one will E only if one Fs, then one has reason to F that is at least as strong as one’s reason to E” (Kolodny 2007, 251). “If X has an objective reason to do A and to do A X must do B, then X has an objective reason to do B of equal weight to X’s objective reason to do A” (Schroeder 2009, 245). “If R is a practical reason in favor of X, X is attainable by the agent, and M is a necessary means to or necessary constitutive element of X, then R is a practical reason in favor of M” (Bratman 2009, 424). “If you have a reason to do A and doing B is a necessary means to doing A, you have a reason to do B which is at least as strong as your reason to do A” (Way 2010, 225). See also Millsap ms a. 7 Way 2010 224 and Bedke 2009 687 n. 10 endorse Weak Sufficiency. The “logic of satisfactoriness” of Kenny 1966 validates an analogue of Ought Sufficiency for “fiats,” which are something like expressions of intention (and so “verdictive” or “all things considered” in the way that ‘ought’-judgments are).

4  

(section 7). The basic problem with Sufficiency principles is simply that sufficient means may

be “superfluous” with respect to the end (section 8). A third kind of counterexample, to Ought

Sufficiency (unsurprisingly) and Ought Necessity (surprisingly), arises when means are costly.

In particular, there are cases where one ought to pursue the end, but one ought not to take the

necessary means to it, because certain costs “attach” to the means without similarly “attaching”

to the end (section 9).

These negative points testify to the wisdom of Raz’s:

Facilitative Principle: When we have an undefeated reason to take an action, we have

reason to perform any one (but only one) of the possible (for us) alternative plans that

facilitate its performance (Raz 2005a, 6),

which avoids any reference to necessary or sufficient means. This paper’s positive aim is to

articulate (in sections 1–5) a principle—unimaginatively titled “General Transmission”—that

would capture the spirit of Raz’s proposal, but with perhaps somewhat more generality and

exactness. My hope is that General Transmission avoids the counterexamples raised against the

Necessity and Sufficiency principles, and offers a satisfyingly “unified,” “informative,” and

“complete” theory, which, among other things, delivers whatever other valid principles there are

as special cases.

But—it’s healthy to ask—who cares what the correct principles of instrumental

transmission are? So long as we grant that there is some such phenomenon, why bother to make

its description precise? Perhaps a sufficient answer would simply cite the aim, which fuels so

much of philosophy, of making explicit the conceptual structures that implicitly underwrite our

thought and practice. Instrumental transmission is ubiquitous. It is hard to think of an episode of

practical reasoning that isn’t in some way implicitly sensitive to it. So there is a natural interest

5  

in the kind of self-understanding gained by finally coming to terms with what we everywhere

unreflectively employ.

A more specific reason for caring about what the principles of instrumental transmission

say is that such principles have been pressed into service by various treatments of the

“normativity” of instrumental rationality: treatments of the demand, or seeming demand, to make

our psychology means-end coherent. First, some have suggested that there are reasons to be

instrumentally rational: for example, that one has “wide-scope” reason (either not to intend the

end of, say, starting World War III, or to intend the means of launching a nuclear missile).

Others have objected that, according to some principles of instrumental transmission, it would

follow, absurdly, that one had “narrow-scope” reason to intend to launch a nuclear missile. Yet,

as I argue, these are among the principles of instrumental transmission that I argue are incorrect.8

Second, with Raz and others,9 I am drawn to the conjecture that the normativity of

instrumental rationality is, in some sense, only apparent: merely the psychological shadow, so to

speak, cast by the instrumental transmission of reasons. Since (as one knows) one’s reasons for

ends transmit to means, when one adopts an end (presumably believing that one has reason for

it), it will of course seem to one that one has reason for the means. But it is a mistake to

conclude from this that there is in fact a normative demand to be instrumentally rational as such:

to put one’s intentions for ends and means into a kind of tidy order. However, defenses of this

conjecture have relied on principles of instrumental transmission that, I argue, are false.10

Ultimately, I believe, General Transmission allows us to put on a surer footing both of these

claims about the normativity of instrumental rationality: that there is no reason to be

                                                                                                               8 See Setiya 2007, Schroeder 2009, Raz 2005a and b, and Bedke 2009. 9 See Scanlon 2007, Schroeder 2009, and Bedke 2009. 10 See Kolodny 2007 and 2008; Schroeder 2009; Bedke 2009; and Raz 2005a and b.

6  

instrumentally rational (section 10), and that the seeming normativity of instrumental rationality

is an illusion arising from the instrumental transmission of reasons (a subject for another paper).

Even granting that there is such a phenomenon as instrumental transmission, that it is

worth describing, and that General Transmission describes it, the question remains: What

explains instrumental transmission? Why do reasons for ends flow to means in the way that

General Transmission says? My hunch is that no deeper explanation is forthcoming: that

General Transmission is a—perhaps the—basic fact about the structure of practical reason.

While I have no proof that no explanation can succeed, I conclude by discussing some telling

failures (section 11).

1. Desideratum: means must probabilize the end

In these first few sections, I identify three desiderata for a principle of instrumental transmission:

namely, that reason for an end should transmit to a means just to the extent that that means is (1)

probabilizing, (2) effective, and (3) nonsuperfluous with respect to that end.

Intuitively, reason for an end, E, transmits to a means, M, to the extent that M-ing “makes

E-ing probable,” or “probabilizes E-ing,” at very least in the following two senses. If there’s no

probability at all, conditional on one’s M-ing, that one E’s, then no reason is transmitted. And,

other things equal, if the probability increases, then more reason is transmitted. It is less clear,

however, how this notion of “probabilizing” the end should be extended beyond these two fixed

points.

One might first suggest that more reason is transmitted to the extent that M-ing raises the

probability of E-ing: to the extent that there is a greater difference of the probability, conditional

on one’s M-ing, that one E’s less the probability that one E’s. But this view implies, oddly, that

the more likely one is to M, the less reason there is for one to M. This is because the more likely

7  

one is to satisfy the relevant condition (i.e., M-ing), the more the conditional probability

approaches the corresponding unconditional probability.

This problem is avoided by a second view, which says that more reason is transmitted to

the extent that M-ing rather than not-M-ing raises the probability of E-ing: to the extent that

there is a greater difference of the probability, conditional on one’s M-ing, that one E’s less the

probability, conditional on one’s not M-ing, that one E’s. But this view too has odd results.

Suppose the only relevant end is keeping the patient alive. Giving drug A alone improves the

patient’s chances of living by 98 percentage points, and giving drug B in addition improves them

by a further percentage point. Intuitively, the doctor has no less reason to give both A and B,

which has a 99% chance exactly of keeping the patient alive, than to give the patient at least A,

which has at most a 99% chance. However, giving both A and B rather than not giving both A

and B (assuming the doctor is very likely to give A but not B, if he does not give A and B) raises

the probability by something approaching only one percentage point, whereas giving at least A

rather than not giving at least A raises the probability by something approaching 98 percentage

points. So the current proposal implies, counterintuitively, that the doctor has less reason to give

both A and B than to give at least A.

The problems with these two construals suggest to me that the relevant notion of

“probabilizing” the end is not a comparative one at all: making the end more probable than some

alternative. So I propose, instead, that some of the reason to E is transmitted to M-ing if there is

positive probability, conditional on M-ing, of E-ing, that more of the reason is transmitted as this

probability increases, and that all of the reason is transmitted if the probability is one. This

avoids the two problems with the comparative construals that we just discussed. Being more or

less likely to M has no effect on a probability conditional on M-ing, so being more likely to M

8  

does not reduce one’s reason to M. And the probability, conditional on giving A and B, of

saving the patient’s life, is higher than the probability, conditional on giving at least A, of doing

so, so we get the intuitive result.

Admittedly, it may sound odd that to say, when the probability is very small, that one has

reason to M. If the end is winning a million-to-one lottery, do I have reason to buy a ticket? But

it is crucial to bear in mind, first, that the reason transmitted to the means is only pro tanto. It

may, and often will, be outweighed; it may well not be the case that I ought, or have conclusive

reason, to take the means, and it may well be the case that I ought not, or have conclusive reason

not, to take the means. Second, while it can sound odd to say that one has reason to do

something when this reason is very weak, or vastly outweighed, there is a plausible pragmatic

explanation of this. As Schroeder 2004, 2005, and 2007 observes, it is a violation of the Maxim

of Relation of Grice 1989 to say that there is a reason, when, as one knows, it is very weak or

vastly outweighed. So long as one bears this in mind, it does seem true that one has some reason

to buy a ticket. At least this much can be said for doing so, whatever else can be said against it:

there is some chance, if one does, of winning a lot of money as a result.

How should we understand this conditional probability? Very crudely, I suggest that we

understand it as the proportion of the “relevant” worlds at which one M’s where one also E’s. I

realize that this talk of “worlds” risks alienating both of my potential readerships. The ethicists

are liable to see it as introducing needless technicality (and betraying a lack of sub-disciplinary

self-confidence!). Whereas the non-ethicists will think that it introduces insufficient technicality

(and betrays a kind of wooly-headed dilettantism!). For one thing, this way of calculating the

conditional probability requires the set of relevant worlds to be finite! To the non-ethicists, I

apologize that, while this is oversimplified, it should suffice for the points I want to make. To

9  

the ethicists, I apologize that speaking of what is happening at a given world simply makes it

much easier to describe clearly and accurately the somewhat involved examples that you will be

asked to endure. Please don’t be put off by it; it’s mainly an accounting device.

What are the “relevant” worlds? I prefer to see them as (i) the epistemically possible

worlds: the worlds compatible with the relevant body of information. I should stress that the

relevant body of information need not be that of the agent at the time of acting. It may be

instead the information, for example, of an advisor, or an onlooker (as a limiting case, an

omniscient observer). More generally, it will be contextually specified.11 But there are other

frameworks, which are compatible with most of what I go on to say. The relevant worlds might

be (ii) the historically possible worlds: the worlds compatible with the history of the actual world

up until the relevant time, or (iii) the worlds at which one M’s that are counterfactually “closest”

to the actual world (in which case the probabilities will often be either zero or one).12

2. Desideratum: means must be effective

So far I have suggested that reason for E-ing transmits to M-ing if the probability, conditional on

one’s M-ing, that one’s E’s is positive. In two ways, however, “that one E’s” is too simple. To

begin with, it matters whether one’s M-ing is effective with respect to one’s E-ing: whether one’s

M-ing helps to bring about or to make it the case that one E’s.

Suppose a boxer’s end is landing a punch. Whenever he decides to throw a punch, he

“telegraphs” his intention with a conspicuous facial twitch. Nevertheless, there are some worlds

                                                                                                               11 As for the question of how context determines the relevant body of information, I favor the view of Kolodny and MacFarlane ms., that the truth of an occurrence of an ‘ought’ or ‘reason’ sentence depends on the information relevant at a context of assessment: to a first approximation, the information of the person—agent, advisor, onlooker, etc.—who is considering the occurrence. 12 Why not also: (iv) the actual world alone? That would mean that reason transmits to M-ing only if one actually does M. Otherwise the probability is not positive (because undefined). But it seems that one can have reason to M even though one does not M.

10  

where, in spite of inadvertently warning his opponent, he connects. So, the probability that he

connects conditional on telegraphing is positive. But it doesn’t seem that any reason transmits to

telegraphing. This is because at no world in which he connects does telegraphing “help to bring

it about” or “help to make it the case”; at no world is it effective to that end. Instead,

telegraphing is merely a by-product of something—deciding to throw the punch—that does help

to bring it about. Thus, the relevant probability should be something more like that of one’s E-

ing and one’s M-ing helping to bring it about that one E’s conditional on one’s M-ing.

By “helping to bring about” or “helping to make it the case,” I have in mind not simply

causing, but also constituting, satisfying preconditions of, as well as helping to cause, constitute,

or satisfy preconditions of. And there may be still other ways of helping to bring about an end.

But for our purposes, we need not fully analyze the relation of helping to bring about. We only

need to say enough about when putative means do and do not help to bring about ends—for

example, that means that help to cause the end help to bring it about—to evaluate candidate

principles of instrumental transmission.

3. Desideratum: means must not be superfluous

But it is still not specific enough to focus on the probability, conditional on one’s M-ing, that one

E’s and that one’s M-ing helps to bring about one’s E-ing. This is because one’s M-ing might be

superfluous with respect to one’s E-ing.

Suppose the end is relieving the patient’s pain. Drug 1 alone will do this for sure, as will

Drug 2 alone. Moreover, if both are given, they at first neutralize one another, but then combine

in the patient’s bloodstream to become Drug 3, which also relieves the patient’s pain for sure. It

is (epistemically, historically, etc.) necessary (i.e., true at every relevant world) that Dr.

Twoways gives the patient Drug 1. It is necessary Dr. Oneway does not give Drug 1.

11  

Intuitively, Dr. Twoways has less reason than Dr. Oneway has to give Drug 2. Indeed, I think

that Dr. Twoways has no reason at all to give Drug 2. After all, any cost to giving Drug 2, no

matter how slight, would outweigh whatever there is to be said for giving it—which would seem

to imply that there is nothing to be said for it.

But our account so far does not explain why Twoways has less reason than Oneway to

give Drug 2, let alone why he has no reason. For both doctors, the probability, conditional on

giving Drug 2, that the pain is relieved and that giving Drug 2 helps to bring this about, is one.

In every world in which they give Drug 2, their doing so helps to cause the relief of pain. After

all, in every such case, Drug 3 is the proximate cause of the relief of pain, and the presence of

Drug 3 is a causal consequence of having administered Drug 2.

The difference is that for Twoways, in every world in which Drug 2 helps to bring it

about that the pain is relieved, Drug 2 brings this about superfluously, whereas for Oneway, in no

possible case in which Drug 2 helps to bring it about that the pain is relieved does Drug 2 do so

superfluously. This suggests that the relevant probability is that of one’s E-ing and one’s M-ing

helping to bring about one’s E-ing in a nonsuperfluous way conditional on one’s M-ing.

Defining the relevant sense of “superfluity” is not easy. A necessary condition of one’s

M-ing helping to bring about one’s E-ing in a superfluous way at a world is that the

counterfactual is true, at that world, that if one had not M-ed, one would still have E-ed. But this

is not a sufficient condition. Suppose that in every world in which I enter by the front door (and

so not by the back door), it is true that if I had not entered by the front door, I would have entered

by the back door, and vice-versa. If the above necessary condition were sufficient, then no

reason would transmit to entering by the front door and no reason would transmit to entering by

12  

the back door, even though I have reason to enter and even though (entering either by the front or

by the back) is necessary and sufficient for entering.

The crucial difference between entering by the front door, which is not superfluous at any

world, and Twoways’s giving drug 2, which is superfluous at every world, seems to be this. At

every world at which I enter by the front door, I do not also enter by the back door, whereas at

every world at which Twoways gives Drug 2, Twoways also gives gives Drug 1. Hence, we

might say that M-ing is superfluous toward E-ing at world just when at that world both (1) if one

had not M-ed, one would still have E-ed and (2) for some M*:

(i) one M*’s and one’s M*-ing helps to bring about one’s E-ing,

(ii) if one (had not M-ed but had still M*-ed), then (one would still have E-ed and one’s

M*-ing would have helped to have brought about one’s E-ing), and

(iii) it is metaphysically possible that one (M’s and does not M*).

This condition is fulfilled for Drug 2 as M, with Drug 1 serving as the relevant M*. But this

condition is not fulfilled for front door as M, with back door serving as the relevant M*. At no

world at which one enters by the front does one also enter by the back, so (i) fails. Now, it might

still seem that we could show that entering by the front door is superfluous by citing as the

relevant M* either entering by the back door or entering by the front door. At every world at

which one enters by the front, one (enters either by the front or by the back)—this satisfies (i)—

and (entering either by the front or by the back) suffices, even without entering by the front, to

bring about entering—this satisfies (ii). But this choice of M* is ruled out by condition (iii). It is

not metaphysically possible that one enters by the front door but does not enter either by the

13  

front or by the back.13 I am not sure whether this is a fully adequate account of the notion of

superfluity. But, again, for our purposes such an account is not needed. We need only to say

enough about when means are and are not superfluous—for example, that if one would not have

E-ed had one not M-ed, then M-ing is not superfluous to E-ing—to be able to evaluate candidate

principles of instrumental transmission.

Nevertheless, it is important to distinguish superfluous means from what we might call

“excessive” means. Consider Dr. Fourchoice, who might possibly give neither drug, or give

Drug 1 alone, or give Drug 2 alone, or give both (both simultaneously, if you like, although

that’s not essential). By the above definition, his giving both is not superfluous at any relevant

world. Indeed, as insofar as relieving the patient’s pain is concerned, there is just as much reason

to give both drugs as there is to give exactly one: doing so would relieve the patient’s pain. If

the cost of giving both drugs is the same as the cost of giving exactly one, then he ought either

(to give exactly one or to give both). It is only if the cost of giving both is higher that he ought

to give exactly one. This suggests that there is just as much reason to give both as there is to

give one. In the case of Twoways, by contrast, any cost incurred by taking the superfluous

(rather than excessive) means of giving Drug 2 outweighs the reason to give it, which suggests

that there is no reason at all to do so.

4. The positive account: General Transmission

Putting these suggestions together, I propose:

                                                                                                               13 Why does (iii) invoke metaphysical, rather than historical or epistemic possibility? Because, in the drug case, it is not historically or epistemically possible that Twoways M’s (i.e., gives Drug 2) but does not M* (i.e., give Drug 1). If (iii) were put in terms of historical or epistemic possibility, reason would transmit to giving Drug 2.

14  

General Transmission: Relative to a fixed context,14 if there is reason for one to E, where

this reason is not explained by an application of General Transmission to reason for

some distinct E’, and there is positive probability, conditional on one’s M-ing, that (one

E’s and one’s M-ing helps to bring about one’s E-ing in a nonsuperflous way15), there is

reason for one to M, more reason the higher this probability, and at least as much reason

as there is for one to E if the probability is one[, so long as the reason against one’s M-

ing is not sufficiently more weighty than the reason that there is, or would otherwise be,

for one’s M-ing].

This just assembles the three desiderata identified earlier, save for the three italicized provisos.

The first proviso, “relative to a fixed context,” needs little explanation. I assume that the

truth of claims about what one has reason to do may be relative to certain features of context,

such as a specified time (as in the historical framework) or a body of information (as in the

epistemic framework). It is no argument against General Transmission, therefore, that there is

reason for one to E relative to one context, which invokes, say, one body of information, whereas

there is not reason for one to M relative to a different context, which invokes a radically different

body of information. I will assume that a similar proviso applies to all of the other transmission

principles that we will discuss, except for a complication that arises with transmission principles

                                                                                                               14 Or pair of context of occurrence and context of assessment. See note 11. I’m being lax here and throughout about the use-mention distinction. Strictly speaking, we should say “an occurrence of the sentence ‘There is reason for one to E’ is true at a context” rather than “there is reason for one to E at a context.” 15 Some may worry about the intelligibility of a conditional probability of something’s helping to bring something else about in a nonsuperfluous way, since this would be a conditional probability of certain counterfactual and causal states of affairs. But I think that we can make sense of this. Suppose that we can say, of any given world, whether the relevant counterfactual or causal claim is true: say, by considering the worlds surrounding it. Then we can calculate the conditional probability by, so to speak, counting the worlds, in the set of relevant worlds at which one M’s, at which the relevant causal or counterfactual claims are true.

15  

for ‘ought’ claims in section 5. There, shifts in the contextually specified set of “relevant

alternatives” may be standard as we move from the question “Ought one to E?” to the question

“Ought one to M?” So this shift in context, but only this shift, should be built into transmission

principles for ‘ought’ claims.

The second proviso arises from a problem explored by Millsap ms. a (who credits it to

Kenny Easwaran) and Bedke 2009 679 n. 12 (who credits it to Jamie Dreier). In Millsap’s

example, Kenny has reason to prepare a display on the life of Marie Antoinette, which requires

that he bake a cake and a loaf of bread. We can assume that there’s a positive probability,

conditional on baking a cake, that he prepares the display and that his baking the cake helps to

bring this about in a nonsuperfluous way. So General Transmission implies that Kenny has

reason to bake a loaf of bread. We can also assume that there’s positive probability, conditional

on baking a mega-loaf, which uses all of the available flour, that he bakes a loaf of bread and that

baking a mega-loaf helps to bring this about in a nonsuperfluous way. So General Transmission

implies that Kenny has reason to bake a mega-loaf. But baking a mega-loaf is not simply

“excessive” but also self-defeating. It prevents Kenny from baking a cake, and so prevents him

from preparing the display—which was the point of baking a loaf in the first place. A reason to

do something has led to a reason to prevent himself from doing it.

The problem here, one wants to say, is that Kenny does not have reason to bake a loaf,

period. Instead, he has a reason to bake a loaf in a way that helps him to achieve the end of

preparing the display. And baking a mega-loaf, while a means to baking a loaf, is not a means to

that: to baking a loaf in a way that helps Kenny achieve the end of preparing the display. In

other words, reason for a means to a means is transmitted directly from the “ultimate” end, not

from the means to which it is a means. This is why General Transmission reads: “If there is

16  

reason for one to E that is not explained by an application of General Transmission to reason for

some distinct E’…”16

The third proviso, which I bracket because it seems to me optional, arises from a

“silencing” objection that is often raised against other transmission principles.17 Suppose the end

is slightly improving college policy, which, at least at first glance, is something that I have

reason to do. The only roadblock is the kindly old don who is sure to veto any proposed change.

Suppose there’s a positive probability, conditional on poisoning the don, that I change college

policy and poisoning the don helps to bring this about in a nonsuperfluous way. General

Transmission implies that I some reason to poison the don—although obviously overridden by

the reasons against doing so. But some will say that I don’t even have overridden reason for the

means; reason for the means is “silenced.” This is the argument for appending at the end of

General Transmission: “… so long as the reason against one’s M-ing is not sufficiently more

weighty than the reason that there is, or would otherwise be, for one’s M-ing.”

However, this proviso seems to me optional. I grant that it is odd to say that there is

reason to poison the don. And I grant that a fully virtuous agent would not treat the

improvement to college policy as a reason for poisoning the don. That thought would not even

arise in his deliberations.18 But I doubt that we should conclude from these observations that

there is no such reason.19 As Millsap ms. a, Raz 2005b 3, and Bedke 2009 684–686, observe,

                                                                                                               16 Bedke opts for more or less the same solution. Millsap’s solution differs. It is worth noting that, if we add this “intransitivity” restriction, then we need not worry about the problem of “explosion”—that if there is reason for something, then there is reason for anything—that Millsap ms. b notes is a consequence of certain transmission principles. 17 Setiya 2005 raises it against something like Weak Necessity, whereas Broome 2005 raises it against something like Weak Sufficiency and Raz’s Facilitative Principle. 18 See McDowell 1998a, b, and c; Price 2008 12 n. 9, 184 n. 72; and Setiya 2005. 19 I am not here questioning the “particularist” idea that a consideration that is a reason in one context may not be a reason in another. For example, that something would bring one pleasure is

17  

Schroeder’s pragmatic point (discussed in section 2) seems to explain why it is odd to say. And,

to entertain the thought, at least in the context of live deliberation, does indeed indicate a vice.20

(Doesn’t it go without saying, or thinking, that the agent shouldn’t kill the don, in which case

entertaining thoughts about the reason for doing so is idle? So why is he entertaining thoughts

about them? Is the verdict somehow not obvious to him? Or is he somehow tempted to defy it?)

But the viciousness of entertaining the thought is compatible with its truth. It similarly indicates

a vice of one kind to entertain lascivious but true thoughts, during wedding vows, that sex with

the officiant would be pleasurable, and a vice of another kind to entertain distracting but true

thoughts, during oral argument, that it is time to treat oneself to a new judicial robe.

5. The positive account: The transmission of ‘ought’s

General Transmission speaks only of the instrumental transmission of reason, not of ‘ought.’

This is as it should be, since I do not believe that there are any interesting generalities about the

instrumental transmission of ‘ought’s.

The most plausible link between reason and ‘ought,’ to my mind, is roughly that one

ought to do the option, from among the contextually specified relevant alternatives,21 that one

                                                                                                                                                                                                                                                                                                                                                                     often a reason to do it, but not when the something is causing an innocent to suffer. I am only questioning the claim that within a fixed context, there might be reason for the end without there being reason for probabilizing, effective, nonsuperfluous means to it. 20 It is worth noting that McDowell sometimes characterizes silencing in this way: in terms of the virtuous agent’s not treating the consideration as a reason instead of in terms of its not being a reason. Of course, on some views (and perhaps McDowell’s own), there is little or no difference between the two; a reason just is a consideration that a virtuous agent would treat in a certain way. 21 I leave it largely open how “alternatives” should be understood. However, first, in the epistemic framework, alternatives should probably be thought of as descriptions of actions (or types of action), with the consequence that two distinct alternatives might, in fact, be one and the same action (or type of action). Second, alternatives should not be thought of as sets of relevant worlds at which a given action takes place. General Transmission requires that what one has reason for be something that can stand in the “helps to make it the case” relation. A set of relevant worlds seems too coarse-grained for that. The set of relevant worlds at which the boxer

18  

has most reason for, or, if several alternatives are “tied” for what one has most reason for, their

disjunction.22 Notice that his actually implies that when E and M are distinct alternatives, it

cannot be the case, relative to a fixed context, that one ought to E and one ought to M. Since

many necessary means, and many sufficient means, to an end are distinct from that end, one

might first complain that this falsifies Ought Necessity and Ought Sufficiency straightaway,

short-circuiting any interesting case against them. Second, one might complain that it is simply

implausible. Don’t we sometimes consistently think both that we ought to achieve an end and

that we ought to take some means to it?

The response to the second complaint is that a shift in the contextually specified set of

relevant alternatives is standard in such cases. “Ought I to E?” is one question, which invokes a

certain set of relevant alternatives, by default: {E, refrain from E-ing}. “Ought I to M?” is

another question, which invokes a different set, by default: {M, refrain from M-ing}. With this

shift taken into account, both claims can be true. This, in turn, suggests a way to respond to the

first complaint: that is, simply to view this shift as built into Ought Necessity and Ought

Sufficiency. For example, we might have Ought Necessity say that if M-ing is a necessary

means to E-ing, and relative to {E, refrain from E-ing} one ought to E, then—provided there is

no other shift in context (i.e., in the relevant information)—relative to {M, refrain from M-ing}

one ought to M. This would not prejudge Ought Necessity.

However, I believe that, even if we allow this shift in the set of relevant alternatives, we

still will not find any interesting generalities about the instrumental transmission of ‘ought.’

Suppose one ought to E. Then one has reason to E, and more reason than one has to refrain from

                                                                                                                                                                                                                                                                                                                                                                     throws the punch might be the set of relevant worlds at which the boxer telegraphs it, but only his throwing the punch helps to bring it about that he lands it. 22 This approach is heavily indebted to Jackson 1985 and Cariani 2009, among others. I discuss a rival “quantificational” semantics for ‘ought’ in note 33.

19  

E-ing. According to General Transmission, some of the reason for E-ing will transmit to any

probabilizing, effective, nonsuperfluous means, M-ing. But why, in general, should this reason

outweigh the reason to refrain from M-ing? As we will see, it very often does not, even when

M-ing is a necessary means.

6. Against other principles: Incompleteness, disunity, and uninformativeness

I turn now to criticism of other principles that have been advanced in the literature: the Necessity

and Sufficiency Principles, and Raz’s Facilitative Principle. In this section, I briefly note that,

even if these principles were correct, they still would not make for a very satisfying theory.

These principles are at best incomplete, in the sense that there are means to which none of

them applies. Suppose an aimless teenager has not adopted any plan that contains taking the

SAT as a step. Nevertheless, a parent may know that an aimless teenager has reason to take the

SAT exam as a means to a college education (perhaps partly because the parent knows that, in

time, the teenager will begin to take his future seriously and take other steps, such as filling out

application forms). While there is reason for the teenager to take the SAT, it is neither a

necessary means (the ACT is another option), nor a sufficient means. Nor does it count as a

“facilitating plan.” For one thing, if taking the SAT did count as a facilitating plan toward the

end of getting a college education, then other partial means, such as mailing in the application,

would also count as facilitating plans to that end. But then the “but only one” restriction of the

Facilitative Principle would imply that the teenager does not have reason (at least as far as

getting a college education is concerned) to both take the SAT and mail the application. Nor,

finally, is it covered by Raz’s supplemental claim that:

20  

we have only conditional reason to take [steps within a plan], the condition being that we

have adopted and are pursuing the plan (and that it is still reasonably likely to facilitate

what it is meant to facilitate) (2005a 6).

The teenager has not adopted, and is not pursuing, any plan that has taking the SAT as a step.23

By contrast, General Transmission explains this reason, on the assumption that there is a positive

probability that, conditional on taking the SAT, the teenager goes to college and his taking the

SAT helps to bring this about in a nonsuperfluous way. (And if we don’t assume this, then it is

no longer clear that he does have reason to take the SAT.)

Second, once we see that any one of these principles would at best need to be

supplemented by other principles, the natural question for any pluralistic theory arises: Is there

some single principle that underlies and explains the several principles that the pluralistic theory

invokes? General Transmission, by contrast, aspires to be a unified account: one principle that

explains all the rest.

Finally, some of these principles, such as Weak Necessity and the Facilitative Principle,

are not informative, in the sense that they do not say how much reason is transmitted. This turns

out to be problematic, since, as we will see, these are the most defensible of these principles. So

the appeal to these principles ends up being uninformative, in a way that General Transmission

tries to remedy.24

                                                                                                               23 The point isn’t that Raz denies that reason can transmit to actions that “keep options open” for not yet adopted plans. Indeed, he appeals to precisely this phenomenon (2005b 8). The point is just the pedantic one that the letter of his formulations doesn’t capture their spirit. 24 Two other principles are Bedke’s 2009 678 “Instrumental Principle”: “One has reason to take the means to what one has ultimate reason to do,” and Schroeder’s 2009 246 “General Reason Transmit”:

If X has… reason to do A and X’s doing B would facilitate her doing A, then X has… reason to do B of weight at least proportional to X’s… reason to do A, and to how well her doing B would facilitate her doing A.

21  

7. Against Strong and Ought Necessity: low probabilization

The more serious problem with these principles, however, is that, apart from the exception of

Weak Necessity, they are false. Here is why the Necessity principles seem plausible. If one does

fails to take a necessary means, then there is no chance of achieving the end. So, surely, if one

has reason to achieve the end, one has reason to take the necessary means! The problem is that it

does not follow from this fact—that there is no chance of achieving the end if one does not take

the necessary means—that there is a chance of achieving the end even if one does take the

necessary means. And if there is no chance of achieving the end even if one does take the

necessary means, then why should any reason for achieving the end transmit to taking the

necessary means? As Anscombe observes, “I do P, so that Q,” can be “contradicted” by saying,

“But Q won’t happen, even if you do P.”

Actually, I think that this is too quick. If there is no chance of E-ing even if one takes the

necessary means of M-ing, then there is no chance of E-ing at all. And if there is no chance of

E-ing, then it is not clear that one has reason to E in the first place. Moreover, if there is no

chance of E-ing, then it seems to follow that everything is, trivially, a necessary means to E-ing.

If no matter what one won’t E, then one can say of anything that, unless one does that, one won’t

E. The notion of a necessary means loses its content. It seems that we should define “necessary

means” in such a way as to rule out this possibility. Let us say, then, that one’s M-ing is a

necessary means to one’s E-ing iff (i) at every relevant world at which one E’s, one M’s and

one’s M-ing helps to bring about one’s E-ing and (ii) there is some relevant world at which one

                                                                                                                                                                                                                                                                                                                                                                     Both seem to be unified and complete, covering means such as the teenager’s taking the SAT. However, Bedke’s is entirely silent on how much reason is transmitted, and Schroeder’s, while saying more on this front, leaves it largely open how “how well” is to be understood. Moreover, since Bedke takes his principle to entail Weak Sufficiency, it seems vulnerable to problems of superfluity, and since Schroeder takes General Reason Transmit to entail Strong Necessity, his principle seems vulnerable to problems of low probabilization.

22  

E’s. Clause (ii) ensures that necessary means probabilize the end to some extent, and so blocks

the line of reasoning above to the effect that there might be a necessary means that inherited

none of the reason for the end. Thus Weak Necessity seems secure. And, indeed, given this

definition, General Transmission entails Weak Necessity.25

So far, so good. The question is why we should expect Strong or Ought Necessity to

hold. Yes, since necessary means probabilize the end to some extent, they inherit some of the

reason. But necessary means might probabilize the end only to some minute degree. Why then

should the necessary means inherit, as it were, all of the reason for the end (so that there is at

least as much reason for the means), or the conclusive reason for the end (so that one ought to

take the means)?

Consider Lucky and Unlucky, both of whom have the same reason to get a Ph.D. A

necessary means to this is to apply to graduate school. There are no side-benefits to applying;

the only reason to apply, if there is any, transmits from the reason to get a Ph.D. However, in

Lucky’s universe, there is little competition, so he is extremely likely to be admitted, if he

applies. In Unlucky’s universe, by contrast, there is fierce competition, so he is extremely

unlikely to be admitted, even if he applies. It seems that Lucky has more reason to apply than

                                                                                                               25 At least if we assume that for every relevant world, the closest worlds at which one does not M are themselves relevant worlds. What we need to show is that, if M is a necessary means to E, then there is some positive probability, conditional on one’s M-ing, that one E’s and one’s M-ing helps to bring this about nonsuperfluously. If M is a necessary means to E, then by definition there are some “successful” relevant worlds in which one M’s and one E’s and one’s M-ing helps to bring about one’s E-ing. So all we need to show is that at some “successful” world, one’s M-ing is not superfluous. From the assumption, it follows that at every “successful” world if one had not M-ed, one would not have E-ed. (Take such a world, W. By the assumption, the closest worlds to W at which one does not M are themselves relevant worlds. Because they are relevant, and because M-ing is a necessary means, they are worlds at which if one does not M, one does not E. So it is true at W that if one had not M-ed, one would not have E-ed.) Thus, the necessary condition of superfluity noted in section 3 does not obtain at the “successful” worlds, which means that, at every “successful” world, M-ing is nonsuperfluous.

23  

Unlucky has. To bring this out, imagine that each has to pay the same application fee. If the fee

is high enough, but not too high, then we might advise Unlucky, but not Lucky, that he ought not

to apply. Given that they both have the same reason against applying, it is hard to see what else

would explain this difference in what they ought to do, if not that Unlucky has less reason than

Lucky to apply. According to Strong Necessity, however, each should have the same reason:

namely, as much reason as there is to get a Ph.D.26

                                                                                                               26 “How can Strong Necessity be false?” one might object. “After all, it is just the analogue to a perfectly plausible principle of the transmission of reasons for belief:

Theoretical Strong Necessity: If there is reason to believe P, and it is epistemically necessary that (P ⊃ Q), then there is at least as much reason to believe Q.”

As Darwall 1983, 47–48 writes: “[T]he force of reasons is, as it were, transferred back and forth along the line connecting an end and its necessary means in the same way that the rational force of a deductive argument is transferred between premises and conclusion.”

So where and why does the analogy break down? One way of answering this question is to treat reasons for believing P as just like reasons for action, except directed exclusively at bringing about a distinctive kind of “end”: namely, believing P truly. Then we can actually deduce from General Transmission a version of Theoretical Strong Necessity (that is, a version that adds the further conditions (A) that the epistemic probabilities P and Q are independent of whether one believes them, and (B) that there is no more reason to believe P truly than to believe Q truly). Along the way, we can identify two crucial differences that explain why we cannot similarly deduce Strong Necessity from General Transmission.

Applying General Transmission to the idea that reasons for believing something are transmitted from reasons for the “end” of believing it truly, we get that:

the reason for believing P varies positively with (i) the reason to believe P truly and (ii) the epistemic probability, conditional on believing P, that (one believes P truly and believing P helps to bring it about in a nonsuperfluous way).

Since believing P is necessary for the end of believing P truly, believing P is never superfluous for that end. Moreover, so long as P is true, believing P is sufficient for and helps to bring about (by constituting) its end: namely, believing P truly. This is the first crucial difference. M-ing need not be sufficient (given some standing condition) for its end: namely, E-ing. Indeed, it isn’t in the Unlucky or Procrastinate cases. Continuing the derivation, (ii) becomes:

(ii’) the epistemic probability, conditional on believing P, that (one believes P and P). Now assume (A). Since P is independent of whether one believes P, (ii’) becomes simply:

(ii’’) the epistemic probability that P. Thus we have:

the reason for believing P varies positively with (i) the reason to believe P truly and (ii’’) the epistemic probability that P,

and similarly:

24  

A natural, if somewhat inchoate, worry about this example is that Lucky and Unlucky

don’t really have reason for the end of getting a Ph.D, because it somehow isn’t yet “up to” or

“available” to them. After all, the admissions committee still has to weigh in on the matter. To

bypass this worry, whatever its merits, take Professor Procrastinate, a character from Jackson

1985 and Jackson and Pargetter 1986. He and Professor Dispatch have equally strong reason to

review a book. A necessary means to this is accepting the commission to review it. Reviewing

the book is “up to” each of them in any ordinary sense. Each can review the book simply by

accepting the commission and then writing the review. The problem is that while Dispatch is

sure to write the review, if he accepts, Procrastinate is extremely unlikely to write the review, if

he accepts. According to Strong Necessity, Dispatch and Procrastinate have the same reason to

accept: namely, as much reason as they have to write the review. (Assume that there are no side-

benefits to accepting.) But intuitively Dispatch has more reason to accept than Procrastinate has.

Note that this is not to say that the reason to do something depends on how likely one is to do

that very thing (a claim that we found unacceptable in section 1). It is rather to say, as seems

undeniable, that one’s reason to do something depends on what else is likely to happen, where

this “what else” may be something brought about by, or even consist in, other things that one

does.

                                                                                                                                                                                                                                                                                                                                                                     the reason for believing Q varies positively with (i) the reason to believe Q truly and (ii’’) the epistemic probability that Q.

Now assume (B). Since there is at least as much reason to believe Q truly as there is to believe P truly, there is at least as much reason to believe Q as there is to believe P so long as the epistemic probability that Q is at least as high as the epistemic probability that P. And it is, if the antecedent of Theoretical Strong Necessity is fulfilled: i.e., if it is epistemically necessary that (P ⊃ Q). This is the other crucial difference. Whereas Strong Necessity assumes a necessary connection between the responses for which one might have reason (e.g., one’s M-ing), Theoretical Strong Necessity assumes a necessary connection between their contents (e.g. that P). A truer theoretical analogue to Strong Necessity would read not “… (P ⊃ Q)…” but instead: “…(one believes P ⊃ one believes Q and one’s believing P helps to bring it about that one believes Q)…” And that principle is not at all plausible.

25  

“Procrastinate actually has just as much reason to accept,” one might object. “It just

sounds odd to say that he has as much reason as Dispatch to accept, because Procrastinate has so

much more reason than Dispatch against accepting—just as it seems odd to say that one has

reason to kill the don of section 4, because one has so much reason against doing so. Suppose

Procrastinate’s accepting prevents someone else from writing the review. So—if you like—there

is positive probability, conditional on Procrastinate’s accepting, that his doing so brings it about

nonsuperfluously that no one writes the review. That certainly seems like a reason against

accepting. By contrast, the comparable probability is lower for Dispatch, since even though his

accepting also prevents someone else from writing the review, Dispatch is so likely, if he

accepts, to write the review himself. So Procrastinate has much more reason against accepting

than Dispatch has.” To control for this, consider a case in which accepting but not writing would

not prevent anyone else from writing the review. Imagine that Procrastinate and Dispatch

inhabit different universes in which each is the only qualified reviewer, so that no one else will

review the book if he doesn’t. More generally, imagine that the only reason against accepting is

the cost of replying to the review editor (who has written, “You only need to reply if you

accept”). This cost is not affected by the probability of writing. So both Procrastinate and

Dispatch have the same reason against accepting, since apart from the probability of writing,

their situations are identical. Hence, the pragmatic explanation no longer applies. All the same,

it still seems that Procrastinate has less reason to accept. If the cost of replying were high

enough, but not too high, then we would advise Procrastinate, but not Dispatch, that he ought not

accept.

A similar objection tells against Ought Necessity. We only need to suppose that

Procrastinate’s reasons to write the review are strong enough to make it the case that he ought to

26  

write, but the probability of writing if he accepts low enough and cost of replying high enough

that it is not the case that he ought to accept.27 Of course, this does not mean that, when

Procrastinate does not accept, we cannot criticize him. We can criticize him for having the vice

of procrastination. Moreover, we can even criticize him for failing to do something that he ought

to do: namely, for failing both to accept and to write. But all that is compatible with the answer

to his question, “Ought I to accept?” being “No.”

8. Against the Sufficiency Principles: superfluity

If the Necessity Principles have problems with low probabilization, the Sufficiency Principles

have problems with superfluity. Here’s why the Sufficiency principles seem plausible. If one

takes some sufficient means, then one is sure to achieve the end. So surely if one has reason to

achieve the end, one has some reason to take that sufficient means! The problem is that one may

be sure to achieve the end even if one does not take that sufficient means. In that case, one

seems to have no reason (at least as far as achieving the end is concerned) to take that sufficient

means. As Anscombe observes, “I do P, so that Q,” can be “contradicted” by saying, “But it will

happen whether you do P or not.”

Corresponding to our definition of “necessary means,” let us define M-ing as a sufficient

means to one’s E-ing iff (i) at every relevant world at which one M’s, one E’s and one’s M-ing

helps to bring this about, and (ii) there is some relevant world at which one M’s. Then, in our

terms, a sufficient means can be superfluous at every relevant world at which it is taken. Indeed,

this is exactly what happens with Dr. Twoways of section 3. For Twoways no less than Oneway,

Drug 2 is a sufficient means to relieving the patient’s pain. So, according to Weak Sufficiency,

Twoways does have reason to give Drug 2. However, for Twoways, unlike Oneway, Drug 2 is

                                                                                                               27 Way 2010 225 n. 32 reports a similar objection to Ought Necessity from John Broome.

27  

superfluous toward the end of relieving pain in every relevant world where she gives Drug 2. So

intuitively Twoways has no reason to give Drug 2 (at least not as far as this end is concerned). It

seems even clearer that Twoways has less reason than Oneway—which tells against Strong

Sufficiency—and that while Twoways ought to relieve the patient’s pain (as he will do by giving

Drug 1), it is not the case that he ought to give Drug 2—which tells against Ought Sufficiency.28

It is in an effort to ban transmission to superfluous means, I suspect, that Raz includes the

restriction, “but only one,” in his Facilitative Principle.29 So understood, the restriction says that

if at all relevant worlds, one takes some facilitating plan A, then no reason is transmitted to

taking a second facilitating plan B. The problem is that this restriction bans more than merely

superfluous means. The condition might be met even though B is not superfluous. B might be a

failsafe, in case A doesn’t come to fruition. Intuitively, in such a case, it seems that at least some

reason to take B is transmitted from the end. So, unless more is said about what a facilitating

plan is, the “but only one” restriction seems too strong.

                                                                                                               28 One might protest that since Twoways is already sure to relieve the patient’s pain (since he is sure to give Drug 1), there is no longer reason to relieve the patient’s pain. Thus, the fact that he has no reason to give Drug 2 is no counterexample to any of the Sufficiency Principles. This protest assumes, as it were, that “ought” implies “might not,” which is far from obvious. In any event, similar counterexamples would still tell against Strong and Ought Sufficiency, if not also Weak Sufficiency. Suppose that Drug 1 fails in one out of every hundred cases, whereas Drug 2 works in every case. Twoways is not already sure to relieve the pain, but there is much weaker reason than there is for Oneways to take the sufficient means of giving Drug 2, because in 99 out of a hundred cases, it is superfluous. 29 Alternatively, the restriction might be understood as a ban on “excessive,” rather than “superfluous,” means, in the sense distinguished at the end of section 3. It would then say that if A and B are each facilitating plans, no reason is transmitted to the facilitating plan consisting of the conjunction of A and B. But the case of Dr. Fourchoice suggests that reason should be transmitted, and thus that this restriction would be misplaced. Moreover, I suspect that Raz would agree, since he stresses, in other contexts, that reason transmits to means with essentially the same “excessive” character. For example, he observes (to my mind, rightly) that reason for the end of alleviating one’s hunger transmits even to the “excessively costly” means of killing oneself (2005b 2–4). (See the end of section 4 for a response to the objection, from Broome 2005, that reason for such means should be “silenced.”)

28  

It might first be replied that by “facilitating plan,” Raz means “sufficient means.” Taking

a second sufficient means, when one has already taken one sufficient means, is always

superfluous. However, Raz writes that the Facilitative Principle is only “roughly speaking”

about sufficient means to an end or “what we may crudely and inaccurately describe as means

sufficient for its realization” (2005b 9).

It might next be replied that by “facilitating plan,” Raz means “unimprovable means.”

Unimprovable means need not be sufficient; they may fail to achieve the end. But they fail to

achieve the end only when this couldn’t have been helped: only when taking additional means

would not have achieved it either. More rigorously, a means, M-ing, to E is unimprovable iff

there is no distinct means M* such that if one M’s at every relevant world, there is some relevant

world at which one E’s and one’s M*-ing helps to bring about one’s E-ing in a nonsuperfluous

way. This would validate the “but only one” restriction. If one is sure to take an unimprovable

means, then any further (putative) means are (at best) superfluous at every relevant world. The

problem is that if facilitative plans are identified with unimprovable means, then the Facilitative

Principle transmits reason to facilitating plans that are superfluous. Suppose that, at all relevant

worlds, one takes some means, M*-ing, that is improvable and so not a facilitating plan. M*-ing

makes taking some further means, M-ing, which is unimprovable and so a facilitating plan,

superfluous at every relevant world. If one additionally M’s, one will still be taking only one

facilitative plan overall. So the Facilitative Principle allows reason to transmit to additionally M-

ing, even though it is superfluous at every relevant world.30 No doubt, further qualifications

                                                                                                               30 For an illustration, suppose that at every relevant world, one gives the patient Drug A (our M*). Drug A will cure the patient if the patient has Disease 1, as is likely. However, administering Drug A is not unimprovable, and so not a facilitating plan. If one additionally gives the patient Drug B, this will not interfere with the effect of Drug A and will cure the patient on the off chance that the patient has Disease 2. If one additionally gives the patient Drug C (our

29  

could be brought in to avoid such cases. But the effect of these qualifications would be to bring

the revised Facilitative Principle closer to General Transmission. This suggests, to my mind, that

General Transmission better captures the ideas that animate the Facilitative Principle.

9. Against Ought Sufficiency and Ought Necessity: costly means

The Weak and Strong principles say only that the means inherits reason for the end, not that this

reason outweighs whatever reason there may be against the means. So the fact that the reason

for some means is outweighed by their cost is no objection to the Weak and Strong principles.

But the Ought principles say that the means inherits an ‘ought’ from the end, which seems to

imply that the reason inherited by the means does outweigh whatever reason there may be

against the means.31 So the fact that reason for means may be outweighed by their cost is

relevant to the Ought principles.

Indeed, although it will surprise few to say, cases of costly means refute Ought

Sufficiency. Suppose we ought to visit my folks for Thanksgiving. Options A and B are

sufficient means in the sense defined. But A involves a 36-hour, $5,000 flight with six layovers,

where B involves only a six-hour, $150, direct flight. It is not the case that we ought to take A.

This is so, it seems, because visiting my folks doesn’t require A. There’s also B.

                                                                                                                                                                                                                                                                                                                                                                     M), this will combine with Drug A to make Drug D, which has exactly the same effect as Drug A. However, giving Drug C is unimprovable, let us suppose, since it precludes Drug B from having any effect. So giving Drug C is a facilitating plan. According to the Facilitative Principle, under the present interpretation, one has reason to give Drug C, in addition to A. But this seems wrong, since giving Drug C is superfluous at every relevant world. 31 At least if we accept the principle that if one ought to X, then it is not the case that there is more reason against X-ing than there is for X-ing. Although this principle seems independently plausible, it is worth noting that it follows from the combination of (i) the semantics sketched in section 5, (ii) the proposal later in this section that reason against X-ing just is reason to refrain from X-ing, and (iii) the assumption that when X-ing is a relevant alternative, so is refraining from X-ing.

30  

By contrast, it may seem that Ought Necessity avoids this same problem with costly

means, because necessary means are required for the end. Suppose I decide that we ought to

visit my parents for Thanksgiving. I might then learn that the expense of buying plane tickets

there, which is the only way to get there in time, is prohibitive: that even the cheapest ones

would overdraw on our bank account.32 In other words, I might learn that it is not the case that

we ought to buy tickets to there, despite my earlier judgment that we ought to visit. “But,” it will

be said, “this is hardly a counterexample to Ought Necessity. What you have learned is not only

that it is not the case that you ought to buy tickets, but also that it is not the case that you ought to

visit after all. Just as the fact that buying tickets involves a cost is a reason against buying the

tickets that outweighs the reason to buy them, so too the fact that visiting requires this cost is a

reason against visiting that outweighs the reason to visit. If it is not the case that you ought to

pursue the end in the first place, then it is no objection to Ought Necessity that it is not the case

that you ought to take the necessary means.”

Once we take into account the fact that a necessary means to an end can probabilize it to

varying degrees, however, this reply is no longer so convincing. Why is the fact that visiting

requires buying a reason against visiting? Suppose, as seems plausible, that reason against X-ing

just is reason to refrain from X-ing. So the question is: Why is reason to refrain from buying the

tickets reason to refrain from visiting? The reply between the quotation marks in the last

paragraph needs some explanation of this. The most natural explanation, to my mind, is that

refraining from visiting is a means to refraining from buying. By refraining from visiting, we

refrain from buying.

                                                                                                               32 The problem is not, as with Procrastinate, that the cost of the means is too high because of the low probability of achieving the end conditional on taking the means, but instead that the cost of the means is too high even if we are certain to achieve the end conditional on taking the means.

31  

Suppose, as seems plausible, that one’s refraining from X-ing is one’s bringing it about

that it is not the case that one X’s. Then while some ways of refraining from visiting—i.e.,

bringing it about that we do not do it—help to bring it about that we refrain from buying the

tickets, other ways don’t. Refraining from calling the airline until it is too late does help to bring

it about that we refrain from buying the tickets. But refraining from driving to the airport, after

buying the tickets, doesn’t help to bring it about that we refrain from buying them. So how much

refraining from visiting probabilizes refraining from buying depends on in what way we are

likely to refrain from visiting, if we do refrain. To be sure, refraining from visiting by refraining

from calling the airline until it is too late does significantly probabilize refraining from buying.

The point is that refraining from visiting simpliciter, which encompasses all the ways we might

bring it about that we do not visit, need not much probabilize refraining from buying, depending

on how we are likely to refrain from visiting. (If in a spasm of filial guilt, I am very likely to buy

the tickets akratically, then refraining from visiting does not much probabilize refraining from

buying. If we refrain from visiting, we are almost sure do so in some way that does not help to

bring it about that we refrain from buying.) In such a case, the reason to refrain from buying the

tickets remains unchanged, but there is less reason to refrain from visiting. Thus, while it is still

not the case that we ought to buy the tickets, it can be the case that we ought to visit: that we

ought to pursue to the end, even though it is not the case that we ought to take the necessary

means.33

                                                                                                               33 Although the arguments of sections 7 and 9 against Ought Necessity are consistent with the relevant alternatives semantics of section 5, they do not assume that semantics. Instead, they turn on intuitive judgments about what the imagined agents ought to do: that is, on the data for such a semantics. So they are non-question-begging arguments against the standard “quantificational” semantics for ‘ought.’ This semantics, which says that what ought to be the case is what is the case at all “ideal” worlds, validates (provided all ideal worlds are relevant worlds):

32  

10. Implications: No reasons to be instrumentally rational

                                                                                                                                                                                                                                                                                                                                                                     Inheritance: If one ought to E, and at every relevant world (one E’s ⊃ one M’s), then one ought to M.

And Inheritance entails Ought Necessity. (I take the label from Cariani 2009. For defense of something like Inheritance, see Wedgwood 2006.)

This is not so surprising, since there are other counterexamples to Inheritance besides the counterexamples to Ought Necessity. There’s Ross’s 1941 famous paradox: that, according to Inheritance, “One ought (to post the letter)” entails “One ought (to post the letter or to burn it).” And there’s our telegraphing boxer. If in every relevant world in which he throws the punch he telegraphs it, and if he ought to throw it, then, according to Inheritance, he ought to telegraph it.

In a concessive spirit, however, we might grant that the quantificational approach correctly captures one, “report-of-the-ideal,” use of ‘ought,’ but note that this use is simply different from the use that has been our focus. Our focus has been on the use of ‘ought’ that characteristically guides the deliberating agent to decision and that—by extension—is offered in advice, by those who seek to help the deliberating agent reach a decision. One might call this the “deliberative-advisory” use. (Of course, claims using ‘ought’ and ‘reason’ in this way can be made, or assessed, by parties who are neither deliberating, nor advising. The point is simply that this use is distinguished by its capacity to play this role in deliberation and advice.) By contrast, in the report-of-the-ideal use, “One ought to X” is heard as something like “If one were deontically perfect, one would X.” For example, if one were deontically perfect, and so posted the letter, one would make it the case that one either posted it, or burned it. Likewise, if Procrastinate were deontically perfect, and so wrote the review, then Procrastinate would have accepted the commission. To the extent that we are using ‘ought’ in this way, we are no longer engaged in deliberation or advice; we are not attempting to guide the agent’s decision. Instead, we are merely reporting what the world would be like if the agent were deontically perfect.

(It should be intuitively clear what “deontically perfect” is getting at, but it can also be more explicitly related to the deliberative-advisory use. If A were deontically perfect, A would X iff there is some Y such that (i) at all relevant worlds where A Y’s, A X’s, (ii) A ought (in the deliberative-advisory use) to Y and (iii) if the set of relevant worlds were contracted to include only worlds in which A in fact Y’s, there would be no Z such that, relative to that contracted set, A ought (in the deliberative-advisory use) to Z, but at some world in that set A does not Z. Hence, if Procrastinate were deontically perfect, he would accept (=X), because he ought to accept and write (=Y), and relative to a contracted set where he accepts and writes at all worlds in the set, it is not the case that he ought not to accept (=Z).)

From the foothold thus conceded, however, the quantificational semanticist might then take the offensive and argue that, semantically speaking, there is only the report-of-the-ideal use with the appearance of a deliberative-advisory use being purely pragmatic. For example, someone might argue: “It’s clear why we don’t tell Professor Procrastinate, ‘You ought to accept,’ i.e.: ‘If you were deontically perfect, you would accept.’ We fear that this will lead him to accept, from which no good will come.”

But this seems a very hard row to hoe. What pragmatic explanation can be given of Procrastinate’s thinking, as he surely might think: “Even though, if I were deontically perfect, I would accept and write, and so would accept, I ought to not to accept, because I am not deontically perfect”? And as Cariani 2009 shows, even Ross’s Paradox, which at first glance seems tailor-made for pragmatic explanation, turns out to resist it rather stubbornly.

33  

Suppose, then, that Strong and Ought Necessity, and the Sufficiency principles are incorrect for

the reasons given. What becomes of familiar appeals to these principles of instrumental

transmission in the literature on the normativity of instrumental rationality? One of the main

appeals has been to show that “narrow-scope” reason or ‘ought’ for the consequent F “detaches”

from “wide-scope” reason or ‘ought’ for a material conditional (one E’s ⊃ one F’s). Detachment

threatens the claim of Broome 1999 that, in general, one ought to be instrumentally rational: that

is, that one ought to see to it that (one intends at t to E and believes at t that intending at t to M is

a necessary means to E-ing ⊃ one intends at t to M). For it scarcely seems to follow from the

facts that one intends to start World War III and that one believes that intending to launch this

missile is necessary for starting World War III, that one has “detached,” “narrow-scope” reason

to intend to launch the missile, let alone that one ought to!

Setiya 2007 and Schroeder 2009, revisiting an observation of Greenspan 1975, advocate:

Ought Detachment: If one ought to make it the case that (one E’s ⊃ one F’s) and one

cannot alter the fact that one E’s, then one ought to F.

According to Ought Detachment, Broome’s 1999 claim implies that if one could not alter the

intention and means-end belief above, one ought to intend to launch the missile. However,

Setiya’s and Schroeder’s arguments for Ought Detachment rely on the rejected Ought Necessity,

so we lack grounds for accepting Ought Detachment.

To a similar purpose, Raz 2005a argues for:

Weak Detachment: If one has reason to make it the case that (one E’s ⊃ one F’s), then

one has reason to F,

which he takes to follow not from the Facilitative Principle, but instead from the more basic and

general idea that: “People have reason to do what will bring them into conformity with the

34  

reasons that apply to them” (2005b, 3). Bedke 2009 argues that something similar follows from

his “Instrumental Principle.” If they are right about this, then we need no auxiliary assumptions,

about unalterable intentions or anything else, to refute Broome 1999. If there were reason to be

instrumentally rational, there would always be reason to intend to M, for any M whatsoever!

But, first, the case for Weak Detachment rests on questionable principles. Since Bedke’s

Instrumental Principle entails Weak Sufficiency, it would seem to be falsified by the

phenomenon of superfluity. The “more basic” idea of Raz’s quoted above also seems falsified

by superfluity. Presumably we do not have reason to do what superfluously brings us into

conformity with what we have reason to do.

Second, there seems to be a compelling case against Weak Detachment. Rippon

forthcoming notes that Weak Detachment seems to prove far too much. After all, some wide-

scope reasons seem perfectly plausible: for example, that I have reason (I am in France ⊃ I speak

French). Now suppose that I’m not in France; I’m in the USA, surrounded by English

monolinguists. Weak Detachment implies that I have reason to speak French. But intuitively I

have no reason at all to do this.

Nevertheless, I believe that we can derive from General Transmission analogues to Ought

and Weak Detachment that vindicate Setiya’s and Schroeder’s, and Raz’s and Bedke’s, lines of

argument. On the plausible assumption that one’s F-ing helps to bring it about that (one E’s ⊃

one F’s), General Transmission validates at least:

Strong Detachment: If one has reason to make it the case that (one E’s ⊃ one F’s), and, at

all relevant worlds, one E’s, then one has at least as much reason to F.34

                                                                                                               34 Whether General Transmission validates Ought Detachment is less clear, for reasons related to the costly means objection to Ought Necessity in section 9.

35  

This would imply that if at all relevant worlds one has the intention and means-end belief above,

and if one has reason to be instrumentally rational, then one has at least as much reason to intend

to launch the missile, which seems absurd enough.

General Transmission also validates:

Restricted Weak Detachment: If one has reason to make it the case that (one E’s ⊃ one

F’s), and there is positive probability, conditional on one’s F-ing, that one’s F-ing helps

to bring it about in a nonsuperfluous way that (one E’s ⊃ one F’s), then one has some

reason to F.

This seems to suffice for Raz’s and Bedke’s purposes. It entails that if intending to launch the

missile is not superfluous to being instrumentally rational, and if one has reason to be

instrumentally rational, then one has reason to intend it, which again seems absurd enough.

Moreover, Restricted Weak Detachment handles Rippon’s counterexample. When we

suppose, in entertaining Rippon’s counterexample, that I am not in France, we are in effect

shifting to a context in which there is no (epistemic) probability that I am in France.35 But, in

that case, the material conditional (I am in France ⊃ I speak French) is already satisfied, which

makes speaking French a superfluous means to satisfying it. And Restricted Weak Detachment

does not say that reason transmits to the consequent when satisfying the consequent is a

superfluous means to satisfying the conditional. So it does not follow from Restricted Weak

Detachment that, when I am not in France, I have reason to speak French.

11. Conclusion: Can instrumental transmission be explained?

I have tried to argue that General Transmission is true. But one might ask why General

Transmission is true. What explains it? My hunch is that nothing does: that General

                                                                                                               35 Compare Yalcin 2007 and Kolodny and MacFarlane 2010.

36  

Transmission is basic. I know of no proof that it cannot be successfully explained. But we can

consider some instructive failures.

One might first try to explain General Transmission as a kind of by-product of a value-

based decision theory. Throughout, we have been assuming that there are reasons for ends. But

what accounts for them? One suggestion would be:

Production for ends: If one’s X-ing is sure to realize some valuable state of affairs, S, in a

nonsuperfluous way, then there is reason for one to X, which depends positively on the

value of S.

For example, I have reason for the end of saving the whales, because this brings about a valuable

state of affairs: the whales’ survival. We might then explain my reason for taking a means to

that end—say, campaigning for a ban on whaling—by appeal to General Transmission in the

now familiar way. However, one might propose an alternative. Why not explain both the reason

for the end and the reason for the means with the following principle?

General Production: If there is positive probability, conditional on one’s X-ing, that

one’s X-ing brings about some valuable state of affairs, S, in a nonsuperfluous way, then

there is reason for one to X, which depends positively on the value of S and the

probability.

After all, my reason for the end is just that it has a certain probability (i.e., one) of bringing about

a valuable state of affairs. So why shouldn’t my reason to take the means be explained in just the

same way: that it has a certain (albeit lower) probability of bringing about that state of affairs?

Observe that whenever General Production says that there is reason for the end, it also says that

there is reason for the means, in the way in which General Transmission describes. So if all

37  

reasons for ends were provided by the value of states of affairs that they realize, General

Production would explain General Transmission.

There would be a problem, however. While General Transmission would be true, it

would also be explanatorily idle. Instead of explaining the reason for the means, General

Transmission would simply be an observation about the deliverances of what does explain the

reason for the means: namely, General Production. Whenever General Production gives us

reason for the end, General Transmission would observe, General Production also gives us

reasons for the means. But then General Transmission might just as well observe that whenever

General Production gives us reason for the means, it also gives us reason for the end. The idea

that we have reason for the means because we have reason for the end in a way in which we

don’t have reason for the end because we have reason for the means—in essence, the idea of

instrumental transmission itself—would be revealed as a kind of illusion. It is noteworthy (some

will say telling) that this problem can’t even arise for accounts of the normative significance of

instrumentality that take as their starting point rational activity, rather than from some structure

of reasons “already there,” as it were, prior to rational activity. On Korsgaard’s 1997 account,

for example, the means becomes normatively significant precisely because we have made the

end normatively significant—indeed, in the way that conduct becomes legally required precisely

because of an act of authoritative legislation. There’s simply no possibility that the normative

significance of the end might somehow “drop out” of the explanation of the normative

significance of the means—as it seems to when we try to explain General Transmission as a

corollary of General Production.

In light of this problem, friends of instrumental transmission should actually be relieved

to hear that General Production cannot explain General Transmission. It cannot—or so I think—

38  

because not all reasons for ends are provided by the value of states of affairs that they realize.

Some reasons for ends, for example, are to honor or respect something of value, independently of

whether doing so brings about a valuable state of affairs.36 With respect to these reasons, the

analogue to General Production would be:

General Recognition: If there is positive probability, conditional on one’s X-ing, that one

honors something of value, V, and one’s X-ing helps to bring it about in a nonsuperfluous

way that one honors V, then there is reason for one to X, which depends positively both

on the reason that one has to honor V, and on the probability.

And rather than explaining General Transmission, General Recognition seems just to be a special

case of it, where the reason to E is, specifically, a reason to honor something of value, V.

Taking another tack, one might suggest that we find it hard to explain General

Transmission only because we mistakenly treat means and end as stranded atoms, which

somehow need to be strung together. Once we recognize that action is unified—that the pursuit

of the end just is the taking of the means—then that itself is all the explanation we need.

“Suppose one has reason to E,” one might say. “Now, one’s E-ing isn’t just something that

happens; it is something that one brings about through one’s own causality, through one’s use of

means. In other words one’s E-ing is equivalent to one’s taking the means to E-ing. Thus, if one

has reason to E, then one has reason to take the means to E-ing.”

But this doesn’t take us very far.37 What we need to explain, if we are going to explain

any interesting transmission principle, is not only why reasons for E-ing transmit to the complex

                                                                                                               36 One might say that the state of affairs of honoring the thing of value is itself a valuable state of affairs, but this seems to me distorting, even if a useful fiction for some modeling purposes. 37 One also worries that it may even presuppose General Transmission. At least it helps itself to a principle that is a special case of General Transmission: namely, that if one has reason to E and

39  

of taking the means to E-ing but also why reasons for E-ing transmit to individual parts of that

complex—the particular means—to which E-ing is not equivalent. It may be that the complex of

Lucky’s applying to graduate school, and taking qualifying exams, and writing a dissertation,

etc. is equivalent to Lucky’s end of earning a doctorate. But it is not the case that any single

part, such as Lucky’s applying, is equivalent to Lucky’s earning a doctorate.

Here one might be tempted simply to invoke a “part-whole” principle to the effect that

where there is reason for the whole, there is reason for the parts. But I suspect that the needed

part-whole principle would just be General Transmission itself. For one thing, no such part-

whole principle can be said to follow from some general truth that normative-cum-evaluative

properties flow from whole to part. For there is no such general truth. This was the lesson of

Moore’s “organic unities.” A whole can be beautiful, for example, while its parts are not. For

another thing, any part-whole principle will have the wrong consequences, unless by “part” we

just mean probabilizing, effective, nonsuperfluous means. If Twoways, for example, gives Drug

2, then his giving Drug 2 is indeed “part” of his relieving the patient’s pain. Yet he does not

have reason for that “part.” And if by “part” we just mean probabilizing, effective,

nonsuperfluous means, then this part-whole principle just is General Transmission.

“Just as we suspected,” some will say. “Unless one roots the normative significance of

instrumentality, not in some independent structure of reasons, but instead in the nature of rational

agency itself, of course one will have to fall back on some unexplained principle!” Perhaps. But

I wonder whether one can avoid falling back on some such principle even if one does try to root

the normative significance of instrumentality in rationality.

                                                                                                                                                                                                                                                                                                                                                                     (at every relevant world) one’s E-ing is the same action as one’s F-ing, then one has the same reason to F.

40  

To illustrate what I have in mind, take the most serious attempt at such an explanation,

from the paper to which we are indebted for showing us that the normative significance of

instrumentality was even a problem:

To will an end is to give oneself a law, hence, to govern oneself. That law is not the

instrumental principle; it is some law of the form: realize this end. That of course is

equivalent to “Take the means to this end.” So willing an end is equivalent to

committing yourself, first personally, to taking the means to that end. In willing an end,

just as Kant says, your causality—the use of means—is already thought (Korsgaard 1997,

compare Korsgaard 2009a sec. 4.3.1).

Let’s grant that if I will E, then I must believe that E-ing is equivalent to “taking the means to E-

ing.” Moreover, let us grant that this means that when I will E, I thereby will “taking the means

to E-ing.” Why does it follow that I thereby will any individual means, any part of the complex

of “taking the means to E-ing”? This is more or less the problem that we have just encountered.

Well, it might be said, to will a conjunction is to will the conjuncts. So if Lucky believes

that earning a doctorate is achieved by and only by (gaining admission to a doctoral program and

completing its requirements), then in willing to earn a doctorate, Lucky thereby wills gaining

admission and he thereby wills completing the requirements. And the principle that to will a

conjunction is to will the conjuncts indeed seems more plausible than the principle that if there is

reason for a conjunction, there is reason for the conjuncts. So far, so good.

But what if Lucky believes that completing the requirements is achieved by and only by

(either writing a thesis in practical philosophy or writing a thesis outside of practical

philosophy)? When Lucky wills completing the requirements, does Lucky thereby will (writing

a thesis in practical philosophy)? Not necessarily. All the same, Lucky’s writing a thesis in

41  

practical philosophy is a means, albeit a non-necessary means, to achieving the end of earning a

doctorate. As such it seems to be normatively significant. If Lucky decides to write a thesis in

practical philosophy, this will make sense, and it will make sense because of its relation to the

end of earning a doctorate. This is something that an account of the normative significance of

instrumentality ought to account for.

That we find it difficult to account for by appeal to rationality alone is, on reflection, to

be expected. For even at the level of intuition, rationality is silent about non-necessary means.

As Kant’s classic formulation reflects, one’s refusal to intend a (believed) non-necessary means

to what one intends need not be irrational.38 Suppose all I know is that Lucky wills to complete

the requirements, that he believes that he completes them by and only by (either writing a thesis

in practical philosophy or writing a thesis outside of practical philosophy), and that he refuses to

will writing a thesis in practical philosophy. I don’t have enough to indict Lucky on a charge of

irrationality.

So we are left still needing to forge some normative link, which rationality seems unable

to supply, between end and non-necessary means. And, of course, this is what General

Transmission would do. Whether this point will move others, I can’t say. But it at least helps to

explain why I suspect what I do: that some such principle of the instrumental transmission of

reasons is fundamental.

References Bedke, Matthew, 2009: “The Iffiest Oughts,” Ethics 119: 672–698. Bratman, Michael, 1987: Intentions, Plans, and Practical Reason (Cambridge: Harvard).

                                                                                                               38 This is also freely acknowledged by “cognitivist” explanations of instrumental rationality, which reduce the incoherence of intending E, believing that intending M is necessary means to E-ing, and refusing to intend to M, to a conflict between beliefs: e.g., that will E and—since one will E only if one intends to M and does not intend to M—that one will not E. [Harman, Velleman, Wallace, Setiya.] On this point, compare Schroeder 2009.

42  

Bratman, Michael, 2009: “Intention, Practical Rationality, and Self-Governance,” Ethics 119 (April 2009): 411–443.

Broome, John, 2002: “Practical Reasoning,” in Reason and Nature: Essays in the Theory of Rationality, José Bermùdez and Alan Millar (eds.) (Oxford: Oxford University Press): pp. 85–111.

Broome, John, 2005: “Have We Reason to Do as Rationality Requires?—A Comment on Raz,” Journal of Ethics and Social Philosophy, Symposium 1:1–8.

Cariani, Fabrizio, 2009: The Semantics of ‘Ought’ and the Unity of Modal Discourse, Ph.D. Disseration, University of California, Berkeley.

Darwall, Stephen, 1983: Impartial Reason (Ithaca: Cornell University Press). Darwall, Stephen, 2006: The Second-Person Standpoint (Cambridge: Harvard University Press). Finlay, Stephen, 2009: “Oughts and Ends,” Philosophical Studies 143: 315–340. Finlay, Stephen, forthcoming: “What Ought Probably Means, and Why You Can’t Detach It,”

Synthese. Grice, H. Paul. 1989: “Logic and Conversation” in Studies in the Way of Words (Cambridge:

Harvard University Press), pp. 22–40. Hubin, Donald, 1999: “What’s Special about Humeanism,” Noûs 33: 30–45. Jackson, Frank, 1985: “On the Semantics and Logic of Obligation,” Mind 94: 177–195. Jackson, Frank and Robert Pargetter, 1986: “Oughts, Options, and Actualism,” The

Philosophical Review 95: 233–255. Kenny, Anthony, 1966: “Practical Inference,” Analysis 65–75. Kolodny, Niko, 2007: “How Does Coherence Matter?,” Proceedings of the Aristotelian Society

107: 229–263. Kolodny, Niko, 2008: “Why Be Disposed to be Coherent?,” Ethics 118: 437–463. Kolodny, Niko, forthcoming: “Aims as Reasons,” in Reasons and Recognition: Essays on the

Philosophy of T.M. Scanlon, edited by Samuel Freeman, Rahul Kumar, and R. Jay Wallace (New York: Oxford University Press, forthcoming).

Kolodny, Niko and John MacFarlane, 2010: “Ifs and Oughts,” Journal of Philosophy 107: 115–143.

Kolodny, Niko and John MacFarlane, ms.: “Ought: Between Objective and Subjective.” Korsgaard, Christine, 1997: “The Normativity of Instrumental Reason,” in Ethics and Practical

Reason, Garrett Cullity and Berys Gaut (eds.). Oxford: Oxford University Press, pp. 215–254.

Korsgaard, Christine, 2009a: Self-Constitution. New York: Oxford University Press. Korsgaard, Christine, 2009b: “The Activity of Reason,” Proceedings and Addresses of the

American Philosophical Association 83: 23–43. McDowell, John, 1998a: “The Role of Eudaimonia in Aristotle’s Ethics,” in Mind, Value, and

Reality (Cambridge: Harvard University Press), pp. 3–22 McDowell, John, 1998b: “Virtue and Reason,” in Mind, Value, and Reality (Cambridge: Harvard

University Press), pp. 50–73. McDowell, John, 1998c: “Are Moral Requirements Hypothetical Imperatives?” in Mind, Value,

and Reality (Cambridge: Harvard University Press), pp. 77–94. Millsap, Ryan, ms. a: “Reasons and Means.” Millsap, Ryan, ms. b: “Explosion.” Parfit, Derek ms.: “What We Together Do” Price, Anthony, 2008: Contextuality in Practical Reason. Oxford: Oxford University Press.

43  

Raz, Joseph, 2005a: “The Myth of Instrumental Rationality,” Journal of Ethics and Social Philosophy 1: 1–28.

Raz, Joseph, 2005b: “Instrumental Rationality: A Reprise,” Journal of Ethics and Social Philosophy, Symposium 1:1–19.

Ross, Alf, 1941: “Imperatives and Logic,” Theoria 7: 53–71. Scanlon, T.M., 2004: “Reasons: A Puzzling Duality?” in Reasons and Value: Themes from the

Moral Philosophy of Joseph Raz, R. Jay Wallace, Philip Pettit, Samuel Scheffler, and Michael Smith (eds.) (Oxford: Clarendon Press): pp. 231–246.

Scanlon, T.M., 2007: “Structural Irrationality” in Common Minds: Essays in Honor of Philip Pettit, Geoffrey Brennan, Robert Goodin, Frank Jackson, and Michael Smith (eds.) (Oxford: Oxford University Press): pp. 84–103.

Schroeder, Mark, 2004, “The Scope of Instrumental Reason,” Philosophical Perspectives 18: 337–64.

Schroeder, Mark, 2005: “Instrumental Mythology,” Journal of Ethics and Social Philosophy, Symposium 1: 1–12.

Schroeder, Mark, 2007: Slaves of the Passions (Oxford: Oxford University Press). Schroeder, Mark, 2009: “Means-End Coherence, Stringency, and Subjective Reason,”

Philosophical Studies 143: 223–248. Setiya, Kieran, 2005: “Is Efficiency a Vice?” American Philosophical Quarterly 42: 333–339. Setiya, Kieran, 2007: “Cognitivism about Instrumental Reason,” Ethics 117: 649–673. Vranas, Peter, ms.: “New Foundations for Imperative Logic II: Pure Imperative Inference.” Wallace, R. Jay, 2006: “Normativity, Commitment, and Instrumental Reason” (with postscript),

in Normativity and the Will (Oxford: Oxford University Press): pp. 82–120. Way, Jonathan: “Defending the Wide-Scope Approach to Instrumental Reason,” Philosophical

Studies 147 (2010): 213–33. Wedgewood, Ralph, 2006: “The Meaning of Ought” Oxford Studies in Metaethics, Russ Shafer-

Landau (ed.) 1: 127–160. Yalcin, Seth, 2007: “Epistemic Modals,” Mind 116: 983–1026.