Printer Friendly

Humean intentions.

It is almost platitudinous that intentions are distinct from desires. A recovering alcoholic feels a strong desire to have some scotch, but has no intention of acting upon that desire. A soldier desires to sneak away before the coming battle, but intends to stay. A rough characterization of this distinction is that intentions involve an element of commitment that desires lack. This familiar distinction is an important one. Intentions provide provisionally fixed points in our deliberation; if I intend to have pasta for dinner then I can quit worrying about whether to have Chinese and begin thinking about what kind of pasta to fix.(1) Merely desiring to have pasta would not provide such a fixed point, since I might desire to have pasta but not be committed to doing so, and instead intend to have Chinese. More generally, the mere fact that I desire to do something does not entail that I am committed to doing it in the way that the fact that I intend to do something does. For this reason, intentions can also play a distinctive role in enhancing our coordination with others; if my dinner partner knows that I intend to fix pasta then she knows to bring wine instead of beer. Finally, intentions and desires are subject to different normative constraints. For example, if I intend to have pasta but do not purchase any while realizing that doing so is the only way I could have some, then I am to some degree irrational, while simply wanting pasta does not commit me to taking the necessary means to having some.

Many hold that the differences between intentions and desires are so significant that, not only can we not identify intentions with desires simpliciter, but that intentions are irreducible to any subclass of desires. Indeed, some philosophers go on to embrace the stronger thesis that intentions are not reducible to any complex of belief(s) and/or desire(s). Alfred Mele observes that, "recent years have seen mounting evidence for the thesis that intention ... is not reducible to complexes of beliefs and desires." (Mere 1992a: 300) Bratman 1987, Brand 1984, Hertzberg 1995, McCann 1986, and Pink 1996 also endorse this view. Though not usually distinguished, the irreducibility thesis can come in weaker and stronger versions. Strong irreducibility is the thesis that intentions are not even partially constituted by beliefs and/or desires. Weak irreducibility only holds that intentions are not fully constituted by beliefs and/or desires.

Were either of these theses true, it would be bad news for the otherwise attractive Humean view that a full, rationalizing, explanation of an agent's intentional actions always can be had without appealing to anything other than beliefs and desires, since intentions can clearly be an essential part of a rationalizing explanation of an agent's actions. My main aim is to explain why we should reject the irreducibility thesis in troth forms, thereby defending the Humean view of action explanation. Arguments for the irreducibility thesis tend to be elimination arguments. McCann, for example, argues that an intention to [Phi] cannot be identified with a desire to [Phi], a strong desire to [Phi], nor a strong desire to [Phi] plus the belief that one will [Phi], and therefore infers that it is "unlikely that intention can be reduced to other mental states" (McCann 1986: 205). Elimination arguments are only sound if they consider all the plausible contenders, and the arguments in the literature tend to shortchange the reductionist on this count.

I have two primary aims in this paper. First, I defend a particular reductive analysis of intentions. Second, and more importantly, I argue that the irreducibility thesis is false. It should be emphasized that although I go to some length to defend my own analysis, my argument against the irreducibility thesis does not depend on its being exactly right. Even if the details of my analysis of intentions are not quite right, the dialectic illustrates that it would be wrong to infer that no such analysis is to be had. I argue that the reductionist has a sort of recipe for generating an analysis that will meet any objection worth taking seriously. Indeed, my own analysis could be seen primarily as an expository device, meant to illustrate this recipe "in action," although I hope my use of this recipe yields a result that is itself palatable.


It was beautiful and simple as all truly

great swindles are.

--O. Henry

Although it is quite clear that intentions cannot be identified with desires simpliciter, that is, although the supposition that all desires are intentions and that all intentions just are desires is clearly false, it is useful to see just why that simplistic identification cannot be right. For lack of a better name, call the partisan of this view the Desirist. A sound diagnosis of why the Desirist's identification fails should make vivid just what seems to be so distinctive about intentions. Further, it is not as if there is nothing to recommend this view. One might think, for example, that what makes behavior intentional is its goal-directedness, and that its goal-directedness is explained by its being caused in the right way by a desire. Supposing that an action is intentional just in case it is caused by an intention--one might infer as an initial hypothesis that intentions just are desires.

There are two problems with this identification. First, intentions involve an element of commitment that is not essential to desires. As my description of a recovering alcoholic and a frightened but disciplined soldier in the introduction illustrate, it is easy to generate cases in which someone wants to do something but does not intend to do it.

The second, and related, problem is that the normative constraints governing intentions are different from those governing desires. The relation between this problem and the first one is that this difference in normative constraints supervenes on the descriptive difference which underwrites the first problem--that intention involves an element of commitment that desire does not. This problem can be seen by focusing on a principle similar to Kant's Hypothetical Imperative. Kant holds that "Whoever wills the end, so far as reason has decisive influence on his action, wills also the indispensibly necessary means to it that lie in his power" (Kant 1785: 27 [417]) is analytic. Although not equivalent, the following also may be an analytic principle of practical reason:

IHI (Intention-based Hypothetical

Imperative): If A intends to [Phi] and

believes that [Psi]ing

is a necessary means to [Phi], then she intends

to [Psi] in so far as she is rational.

The Desirist is forced, by her insistence that intentions just are desires, to conclude that the following principle just is another way of expressing IHI:

DHI (Desire-based Hypothetical

Imperative): If A desires to [Phi] and

believes that [Psi]ing is a necessary

means to [Phi], then she desires to [Psi]

insofar as she is rational.

The Desirist faces a dilemma. We can plausibly understand desires in one of two ways. On the first way of understanding desires, DHI comes out to be true, but it is impossible for an agent to fail to meet it, unless she fails to make a very particular sort of inference. Since an agent can rather clearly flout IHI without failing to make this sort of inference, this way of understanding entails that DHI is not just another way of expressing IHI. Contra the Desirist, intentions end up being governed by different norms than desires. On the only other apparently plausible way of understanding desires, DHI, unlike IHI is vulnerable to counterexamples.

The Desirist, like most Humeans, might characterize desires dispositionally:

A desires to [Phi] if and only if for any

[Psi], if A believes that [Psi] is a

necessary means to [Phi] then A is

disposed to [Psi].

Given this characterization, it might seem like DHI simply cannot be flouted, since the clearest way to flout it would be to desire something but simply not be disposed to take what one believes to be the necessary means to it. However, given this conception of desires, one just does not count as desiring something unless one is disposed to take what one believes to be the necessary means to it.

Yet an agent can flout DHI, even on this characterization of desires. To do so, an agent must desire to [Phi], believe that [Psi] is a necessary means to [Phi], and yet not desire to [Psi]. As we have seen, on the Humean characterization of desires, to desire to [Phi] simply is to be disposed to take what one takes to be the necessary means to [Phi], so any agent that desires to [Phi] and believes that [Psi] is a necessary means to [Phi] must at least be disposed to [Psi]. Being disposed to [Psi] is not quite enough to count as desiring to [Psi], though; one must be disposed to take what one believes to be the necessary means to [Psi] as well. So here is the only way an agent could fail to live up to DHI. She could desire to [Phi] and believe that [Psi] is a necessary means to [Psi], but nonetheless not desire to [Psi], in spite of her necessarily being disposed to [Psi]. She could fail to desire to [Psi] in virtue of her not being disposed to take what she believes to be the necessary means to [Psi]. So she might believe that a is a necessary means to [Psi] yet not be disposed to [Alpha], and hence not count as desiring to [Psi].

Notice, however, what an agent must do to fail to live up to DHI. She must believe that [Alpha] is a necessary means to [Psi], and believe that [Psi] is a necessary means to [Phi], and yet not believe that [Alpha] is a necessary means to [Phi]. Otherwise, given our present characterization of desires, she would be disposed to a in virtue of desiring to [Phi], but by stipulation she is not disposed to [Alpha]. She must, in other words, not infer what follows from her beliefs and the transitivity of the "is a necessary means to" relation. So long as one does not fail to make such inferences, one cannot fail to satisfy DHI. However, an agent can fail to satisfy IHI without failing to make such inferences, so DHI and IHI are distinct. For example, one might intend to get a Ph.D., believe that doubling one's research is a necessary means to getting a Ph.D., and even desire to double one's work load, and hence have made the relevant inferences, but still not intend to double one's work load. In particular, one's intentions might not be informed by one's beliefs--one's belief that doubling one's work load was needed might have all along been tacit and unconscious, and hence one might simply not have noticed that one needed to commit oneself to such a doubling.

Moreover, there is a second way IHI could be flouted that does not involve the failure of inference required to flout DHI. One might intend to [Phi], realize that [Psi] is a necessary means to [Phi], but be unwilling, on independent grounds perhaps having to do with what one takes to be the nature or consequences of [Psi]ing, to intend to [Psi]. Indeed, one might have another intention which would be undermined by one's [Psi]ing, which explains one's unwillingness to form an intention to [Psi]. For all that has been said so far, though, one might nonetheless irrationally hold on to one's intention to [Psi]. So long as one's beliefs respect the transitivity of "is a necessary means to" one will still desire to [Psi], but this does not yet mean that one intends to [Psi]. IHI and DHI are different normative constraints.

So far, desires have been analyzed simply in terms of dispositions to act, but this conception may be too narrow. It may be that being disposed to feel pleasure upon realizing that [Phi] is sufficient for counting as desiring [Phi]. Perhaps we therefore should see desires as dispositions either to act or react appropriately. On this account, someone desires [Phi] if she is disposed, say, to feel pleasure upon realizing that [Phi]. Desires may not be essentially phenomenological, but a disposition to have certain phenomenological reactions may be sufficient for having a certain desire. This suggests the following characterization:

A desires to [Phi] if and only if either (a)

for any [Psi], if A believes that [Psi] is a

necessary means to [Phi] then A is

disposed to [Psi], or (b) A is disposed to

experience pleasure upon coming to

believe that [Phi].

This account avoids the main problem of our earlier characterization; given this account, one can flout DHI without failing to make any obvious inferences. One might desire to [Phi] in virtue of being disposed to feel pleasure upon realizing that [Phi], but still not desire what I take to be the necessary means to [Phi]. We avoid this problem because (b) does not make taking pleasure in the realization that a necessary means to [Phi] has been taken necessary for desiring [Phi]. However, this account makes DHI is subject to counterexamples.

Consider the following case. I desire to beat Keith in chess, in virtue of criterion (b); if come to think I beat him I will be pleased. I believe, however, that cheating is the only way I can win. Nonetheless, I am not at all disposed to cheat, nor would I react with pleasure upon discovering that I had cheated; so in no sense do I desire to cheat.(2) DHI convicts me of irrationality. However, there is simply nothing irrational about this. I am disposed to feel pleasure if I discover that I win, and believe that the only way I will win is to cheat. Perhaps it will be insisted that I exhibit some sort of incoherence if I do not either desire to cheat or give up my desire to win. I need exhibit no such incoherence, though. For one thing, I realize that though I fully believe that cheating is the only way I can beat Keith, I am fallible; I could be mistaken in this belief. Given this possibility, why should I not remain disposed to be pleased if I am mistaken, and it turns out that I do beat him fairly? Suppose I do not rationalize my desire in this way. The only putative incoherence here is my being disposed to take pleasure in some state of affairs and yet not being disposed to do or take pleasure in what I believe is necessary to bring that state of affairs about. It is unclear just what is supposed to be incoherent or irrational about this, though. Why think being disposed to take pleasure in something rationally commits me to doing or taking pleasure in what I believe is a necessary means to it? Perhaps my condition is analogous to that of a person who desires the impossible, since my not being disposed to do what by my own lights is necessary to bring about what I desire amounts to my making it the case, insofar as my beliefs are true, that what I long for is impossible. However, this analogy does not suggest that I am irrational, as there is nothing obviously irrational about desiring what you believe to be impossible. So on this construal of desires, DHI is vulnerable to decisive counterexample. As IHI is not so vulnerable, IHI is distinct from DHI on this understanding of desires as well. IHI and DHI are distinct principles, contra the Desirist.


Those who restrain Desire, do so

because theirs is weak enough to be


--William Blake

Intention involves an element of commitment that desire does not necessarily involve, and intentions are governed by different normative constraints than desires. This leaves it open that intentions might nonetheless be (partially) constituted by some special subclass of desires. Before going on to consider specific versions of this idea, I should explain why intentions must be at least partially constituted by desires, in the Humean sense of desires as dispositions to act (the first conception of desires canvassed in the preceding section). The general Humean thought is that a desire is a mental state that the world must fit. Less metaphorically, it is a mental state with a potentially belief-mediated disposition to bring about its content. Whatever else we say about intentions to distinguish them from desires more generally, they must be potentially belief-mediated dispositions to bring about their contents.(3) Otherwise, it is hard to see how they could explain actions in a rationalizing way. Returning to a familiar example, it is hard to see what it could mean to say that I intend to fix pasta if I am not at all disposed to take what I believe are the necessary means to fixing pasta.

So intentions must, in part, be constituted by desires, though not all desires are intentions. A natural way to capture the element of commitment distinctive of intending is to stipulate that to intend to [Phi] one must most want to [Phi], since there is an intuitive sense in which if one wants to do one thing more than another then one is more committed to that first thing than to the second. To make this really intuitive, we need to understand one's most wanting to [Phi] not as meaning the desire to [Phi] is one's strongest desire simpliciter, since this would entail that an agent can have no more than one intention. Rather, the idea must be that to have a strongest, or dominant desire to [Phi], there must be no desire [Psi] such that the agent views [Psi] as a real competitor with [Phi], and the agent's desire to [Psi] is stronger than her desire to [Phi]. Also, we do not intend to do what is beyond our power, so we must also stipulate that the agent does not believe [Phi]ing to be beyond her power.

Talk of the strength of a desire must be understood in this context as talk about the desire's motivational strength--how strongly it disposes the agent to bring about its content, rather than as talk about its normative strength--how much of a normative reason it gives the agent to bring about its content. Strasser 1992 argues that we can understand the notion of a desire's motivational strength in one of two ways, neither of which is friendly to the reductionist's project. On the first understanding it is a tautology that one always performs the action that one thinks will satisfy one's dominant desire. This yields the unhappy consequence that one can never fail to do what one intends. However, we need not understand what it is for a desire to be an agent's dominant one in this way. Following Carruthers, we might distinguish active and dormant mental states:

... we need to draw a further distinction,

between a belief or desire that is activated

(engaged in current mental processes) and

one that is dormant (possessed, but not

currently being used). Most of our beliefs and

desires lie dormant for most of the time. I

continue to have beliefs about my parents'

birthdays, for example, throughout both my

waking and sleeping life. But only

occasionally do these beliefs become

activated, as when I fill out a passport renewal

form and have to complete a section headed

`Dates of birth of parents'. (Carruthers 1992: 173)

The obvious point is that one's dominant desire might be to study, but if this desire is not activated, because one is sleeping or inattentive, then it will not lead one to study.

Though Strasser does not explore this distinction, he allows for this kind of account--the second understanding of a dominant desire he considers is meant to capture all such accounts that do not make it tautologous that an agent will pursue whatever her strongest desire is. Strasser argues that on this understanding, it cannot be that an agent's dominant desire simply is an intention on the grounds that, "a normal person can have incompatible desires, but not incompatible intentions" (Strasser 1992: 15). Of course, Strasser's real point is that a normal person can have incompatible dominant desires (where "incompatible dominant desires" means desires which are equally strong and stronger than all competitors, but which are incompatible), but cannot have incompatible intentions. Strasser might be read as arguing that talk of incompatible intentions is incoherent, but this cannot be right, as one paradigm of an irrational agent is one who has incompatible intentions.

Hence Strasser is more charitably read as restricting his attention to agents who are, in his terms, "normal"--that is, agents who are rational. So the crucial premise of the argument must really be that there is nothing irrational about having incompatible dominant desires, but it is irrational to have incompatible intentions. This assumption, however, is question-begging. While it may be relatively obvious that one can have incompatible desires without being irrational, it is not so obvious that one can have incompatible dominant desires without being, to some degree, irrational.(4)

While Strasser's objection to the dominant desire account of intentions is not fatal, that account is not quite right. Bratman provides the following potent objection:

... even if I now have a predominant desire to

go to Tanner ... I might still not see the issue

as settled: I might be disposed to continue to

give serious consideration to the possibility

of taking the afternoon off and going to a

concert. But if I were to intend to go to

Tanner, I would be disposed not to continue

to deliberate in this way: this is what is

involved in the resistance to reconsideration

characteristic of intention ... my predominant

desire to go to Tanner at noon does not

guarantee that when I see that it is noon that

my desire will control my conduct. I might still

be disposed to deliberate about what to do;

for I might not see the issue as settled.

(Bratman 1987: 18-19)

Bratman's point is well taken. An important function of intentions is to serve as provisionally fixed points in our practical reasoning. If intending to [Phi] did not involve seeing the question of whether to [Phi] as provisionally settled, intentions could not play this role. In the next section, I develop my own analysis in a way that is sensitive to this point.

III. Humean Intentions

What is hard today is to censor one's

own thoughts.

--Arthur Waley

Bratman rightly argues that intentions cannot simply be predominant desires, since one can have a dominant desire to [Phi] without seeing oneself as settled upon [Phi]ing. However, Bratman himself unwittingly provides the appropriate antidote to this problem. What is it to see some practical question as provisionally settled? It is to be disposed not to deliberate about it, at least unless new, relevant information comes to light. Moreover, this disposition is presumably one that is potentially belief-mediated; if I come to believe that I am deliberating about something I have formed an intention to do when no new, relevant information has come to light, I typically will stop deliberating. So seeing an issue as provisionally settled just is having a potentially belief-mediated disposition not to deliberate about it. However, on the broadly dispositional Humean understanding of desires, a potentially belief-mediated disposition not to deliberate about something just is a desire not to deliberate about it! Presumably, Bratman's unwillingness to endorse this conclusion, stems from his sympathy for Platts' claim that "nothing but muddle (and boredom) comes from treating desire as a mental catch-all" (Platts 1979: 256). Nonetheless, Bratman's account suggests that we can avoid his criticism of the predominant desire view if we analyze intentions in the following way:

A intends to if [Phi] and only if (a) A has a

desire to [Phi], (b) A does not believe that

[Phi]ing is beyond her control, (c) A's desire to

[Phi] is a predominant one, which is just to

say that there is no desire [Psi], such that

A does not believe [Psi]ing is beyond her

control, she desires to [Psi] as much as or

more than she desires to

[Phi], and she believes that a necessary

means to her [Phi]ing is that she refrain

from [Psi]ing, (d) A has a desire not to

deliberate any more about whether to [Phi]

unless new, relevant information comes to


Before discussing the merits of this analysis, it should be mentioned that the notion of predominance employed here is different from the notion of dominance that was in play in the preceding section's discussion of Strasser. In the sense of "predominant" isolated in (c), an agent cannot have two predominant desires which she believes are incompatible with one another; in other words, this conception of predominance rules out "ties." I favor this interpretation for reasons that go beyond the scope of the present essay. It might seem that this interpretation entails that an agent can never have incompatible intentions, running afoul of the common sense thought that while having such intentions is typically irrational, it is not impossible. In fact, however, my account does allow for an agent to have incompatible intentions in at least two senses. First, an agent might intend to do two things which are in fact incompatible but which he has no way of knowing actually are incompatible. This will not be of too much help, though, since it is not clear what would be irrational about this. Fortunately, there is a second and more interesting sense in which my account can accommodate the possibility of an agent's having incompatible intentions. An agent might intend two results which he could (perhaps rather easily) see are incompatible but which he nonetheless has not yet seen are incompatible. An example may help illustrate this second, more interesting possibility.

Suppose I want to attend a wedding on April Fool's Day, and I want to spend all day at home working on the first Tuesday in April. I also believe that April Fool's Day is the first Tuesday in April, and that I cannot both stay at home all day and attend a wedding on that day. My beliefs entail that a necessary means to satisfying each of these desires is that I refrain from satisfying the other. However, I might fail to draw this inference. Hence I might not believe that not attending that wedding is a necessary means to staying home all day on that Tuesday, or that not staying home all day on that Tuesday is a necessary means to attending that wedding. In that case, both of these desires could still count as predominant on my analysis. Recall that on my characterization, for a desire [Phi] to be predominant, it must simply be the case that there is no desire [Psi], such that A does not believe [Psi]ing is beyond her control, she desires to [Psi] as much as, or more than she desires to [Phi], and she believes that a necessary means to her [Phi]ing is that she refrain from [Psi]ing. Since an agent might fail to infer that not satisfying one desire is a necessary means to satisfying the other, even when her other beliefs entail this conclusion, an agent might have predominant intentions which are incompatible relative to her beliefs, in that they are not mutually satisfiable if all her beliefs are true. Now all we need to add to my story is that the agent has the relevant desires not to deliberate and lacks the belief that either of these things are beyond her control, and she will count as intending both to go to the wedding and to stay at home all day. It is also not too hard to see why such an agent is irrational--she has not drawn an obvious and useful inference.

I do think this is a plausible account of what an intention is. However, it cannot be overemphasized that even if this particular account is not just right, the strategy I have mobilized in generating it is perfectly general, and speaks in favor of the possibility of a plausible Humean reductive analysis of intentions, and against the irreducibility thesis. For any proposed reduction, anti-Humeans might complain either that it has still not captured the element of commitment so distinctive of intentions, or that it is not subject to the same normative constraints as intentions. Suppose they take the first strategy, emphasizing commitment. It will not do for the anti-reductionist simply to make this charge without explaining just why the proposed account has not fully captured the elusive element of commitment; after all, we will have at least given prima facie reason for thinking we have captured it. As soon as the anti-reductionist tells us what in particular we are missing, though, we can add that to the analysis. Importantly, whatever we are missing is likely to be a disposition to behave or a disposition not to deliberate, since to be committed to some end intuitively just is to be strongly disposed to bring that end about, or to be disposed not to rethink that disposition lightly. Since these dispositions to behave and dispositions not to deliberate will presumably be potentially belief-mediated, to capture the way in which our intentions are sensitive to new information, they will simply be more desires, on the broad Humean understanding of desires. Finally, even if the Humean could not fully capture this element of commitment, it is not obvious why she cannot capture part of that element, in which case, we will still at least have good reason to reject the strong version of the irreducibility thesis.

Suppose instead that the anti-reductionist argues that a proposed reduction does not yield a mental state that is subject to the normative constraints that govern intentions. Given the supervenience of the normative on the descriptive, the anti-reductionist will owe us a story about what descriptive feature(s) is (are) missing from the mental states picked out by our account, such that if that (those) feature(s) were present, our account would avoid this worry. The most obvious candidate is the element of commitment distinctive of intentions, but if the argument of the preceding paragraph is correct, this will not help the anti-reductionist. Alternatively, they might argue that the normative constraints supervene in part on our having the right sort of representation of the world. They might, for example, argue that the reason it is irrational to have inconsistent intentions is because to do so one must have inconsistent beliefs, because to intend to [Phi] is, in part to believe that one will [Phi]. Indeed, such a strong belief criterion has been defended by Audi 1973, Beardsley 1978, and Harman 1986. The arguments for this criterion raise important and difficult questions that deserve a paper of their own, so I will not deal with them here.(6) Obviously, though, the requirement of a belief is no threat to the Humean reductionist. However, these two features--the element of commitment and some representational element like a belief that one will succeed--seem to exhaust all the relevant logical space. As soon as the anti-reductionist gives a precise, plausible characterization of what a reductive analysis is missing and why it should not be missing it, we have reason to think that the Humean can capture it via beliefs or desires.

Having seen why even if my account is not quite right I have still made a strong case against the irreducibility thesis, we should return to my proposed account. Consider once more the anti-Humean worries. The first anti-Humean worry is that any reduction will not capture the element of commitment distinctive of intentions. While many such putative reductions are vulnerable to this worry, we would need an independent argument that the particular analysis being proposed here is likewise vulnerable. For the requirement that one's desire be predominant and that one have a desire not to deliberate any more does prima facie nicely capture this notion of commitment. It entails an unwillingness of the agent to reconsider her options unless new information comes along, and it also gives us good grounds to expect that she will in fact pursue the object of her desire, should the opportunity arise. It is unclear what more the anti-Humean might reasonably demand by way of commitment.

The second anti-Humean worry is that any reduction of intentions to desires will not fit with the difference in normative constraints governing intentions as opposed to desires. I have only considered one such difference, that IHI is a very different constraint than DHI.(7) That difference clearly is captured by the present analysis. Only the Desirist would be forced to think that DHI and IHI were the same normative constraint. The proposed analysis also allows one to explain how an agent can flout IHI in the ways emphasized in section I. Recall that the problem with DHI was that an agent could not fail to meet it unless she overlooked the transitivity of the "is a necessary means to" relation, but this was not true of IHI. The present analysis can explain this difference. An agent might have a predominant desire to [Phi], not believe that tiny is beyond her control, and have a desire not to deliberate about whether to [Phi] unless new relevant information comes to light, believe that [Psi] is a necessary means to [Phi], but still not intend to [Psi] without overlooking the transitivity of the "is a necessary means to" relation. In particular, such an agent might desire to [Psi] but that desire might nonetheless not be predominant, and the agent would therefore not intend to [Psi], in virtue of failing condition (b) of my analysis. The Humean analysis of desires leaves room for a mismatch between the motivational strength of your desire to [Phi] and your desire to take what you believe to be a necessary means to [Phi]. The proposed account uses this logical space to explain the range of mistakes we make in flouting IHI as opposed to DHI. Finally, the account provides a second possible explanation of this phenomenon. An agent might intend to [Phi] and believe that [Psi] is a necessary means to [Phi], and yet not intend to [Psi] in virtue of not having any desire not to deliberate about whether to [Psi] unless new, relevant information comes to light.

IV. Objections

If there is a meaning, it is doubtless


--British Board of Film Censors

Having spelled out and defended my own particular analysis as well as a reason for thinking some such reductive analysis must be correct even if my own is not, it is worth considering what I take to be the most potent objections to my proposed account. A guiding idea behind this account is that we can capture the element of commitment that is distinctive of intentions in terms of certain desires, in the Humean sense of "desires." Since in the Humean sense of "desire" a desire just is a kind of disposition, the idea is ultimately to analyze commitment in terms of dispositions. It is often thought, however, that dispositions have bases. For example, the base of a vase's fragility may be its crystalline structure. This might lead one to wonder why we should not instead identify intentions with the base of the dispositions I have isolated, rather than the dispositions themselves. Assuming that dispositions are distinct from their bases, this analysis would be a competitor with my own.(8) It is worth pausing to see what would be wrong with such an account.

Any given disposition could have a wide variety of bases. For example, a wide variety of crystalline structures might, even in the actual world, underwrite a vase's fragility. Furthermore, insofar as it is conceptually possible for the laws of physics to be different than they actually are, there is a very wide variety of conceptually possible bases for any given disposition. Suppose for the sake of argument that the proposed alternative account is right--intentions are constitutively analyzed in terms of the bases of the dispositions I have isolated rather than the dispositions themselves. Now travel to the conceptually possible world in which we have the dispositions, etc., that on my account are sufficient for someone's having an intention, but where these dispositions have a radically different kind of base. On the proposed alternative account, such a person would not have the intention in question. However, this conclusion is unmotivated. Such a person will presumably think and behave in precisely the same ways as her twin whose dispositions are differently based. It is unclear why it should matter what the base of those dispositions is. Indeed, insofar as my argument to this point is sound, it is clear that it is being disposed in the appropriate way that is essential to being committed to an end; how those dispositions are grounded is not in any obvious way relevant to whether one is so committed.

Next, consider the conceptually possible world in which we have someone with what in the actual world is the base of all the relevant dispositions, but in this possible world the laws of physics are different in such a way that it does not ground any such dispositions. On the proposed alternative account, such a person would have the intention in question. Again, this conclusion seems unmotivated. Why should we think that someone, say, who is not at all disposed to wash his car nonetheless intends to wash it? Indeed, insofar as my arguments that such dispositions are necessary for intending are sound, such a person does not have the intention in question. So unless the partisan of this alternative account has some independent argument against thinking that (a)-(d) are individually necessary and jointly sufficient for intending, we may reasonably presume my account to be more plausible than this proposed alternative.(9)

Another alternative would be to maintain that the desire not to deliberate isolated in condition (d) is a causal consequence of intending rather than being a constitutive part of the intention.(10) However, given that causes and effects are "distinct existences" (on many views, including the Humean one), this entails that we could have an intention without the desire isolated in (d). Again, insofar as my arguments for thinking that (d) is necessary for intending are sound, we have reason to reject this proposal. In particular, unless we have some explanation of how someone could be settled upon something without being disposed not to deliberate any more about it, we have reason to reject such a proposal. These two alternatives depend for their plausibility upon some independent challenge to the claim that (a)-(d) are individually necessary and jointly sufficient for intending. At this point it is worth considering what such a challenge might be.

Consider the following thought experiment offered by Mele as a counterexample to the thesis that an intention to [Phi] just is a predominant desire to [Phi]:

Alan now wants to humiliate Carl at the

dean's party tomorrow more than he

wants to do anything that he takes to be

incompatible with his doing this; that is, his

preponderant motivation falls on the side of

his humiliating Carl at the party, and he

realizes that this is so. However, largely

because Alan thinks that it would be

morally wrong to humiliate Carl at the

party, he has decided to resist his desire to

do so. Though he feels that he must go to

the party, he plans to do everything in his

power to avoid humiliating Carl there.

(Mere 1992b: 154-55)(11)

Mele argues that even though Alan has a predominant desire to humiliate Carl, he does not have any intention to do so. Part of the force of Mele's argument emerges from a comparison with Bob, who also is predominantly motivated to humiliate Carl, and decides against resisting his desire to humiliate him, in spite of his moral qualms about doing so. The suggestion is that those who are not "already in the grip of a theory" will be strongly inclined to conclude that Alan does not intend to humiliate Carl, while Bob does so intend.

So far, this is no direct threat to my own account, since Alan does not necessarily meet the four conditions of my analysis I am not yet forced to conclude that he does intend to humiliate Carl. Let us add that Alan does not believe that humiliating Carl is beyond his power. Then the only condition he might not meet would be (d). Let us stipulate that Alan also meets that condition, and let us stipulate that he meets it in a way that will cast as much doubt as possible on my analysis. In particular, suppose that Alan comes to desire not to deliberate any more about whether to humiliate Carl simply because Martin, who Alan believes to be a mind-reader, has convincingly threatened to kill Alan if he deliberates any more about whether to humiliate Carl.(12) Alan would thereby count as intending to humiliate Carl on my account, but it might seem as if Alan has no such intention, particularly when we consider the etiology of his dispositions and when we compare him with Bob, who clearly intends to humiliate poor Carl. One might complain that Alan at most is settled upon not deliberating further about whether to humiliate Carl, but hold that being settled upon that does not mean he is settled upon humiliating him. If this complaint were just, the claim that (a)-(d) are jointly sufficient for intending would be false.

It is relatively clear that my account commits me to supposing Alan does intend to humiliate Carl. So if I am to respond successfully to this putative counterexample, I will have to show that insofar as it seems to be a genuine counterexample, our intuitions are being pumped the wrong way. Fortunately, there is good reason to think our intuitions are easily pumped the wrong way by these kinds of cases. A crucial feature of the case is that Alan has "decided to resist his desire" to humiliate Carl. Importantly, my account can make sense of this claim in a variety of ways, all of which are compatible with thinking that Alan nonetheless intends to humiliate Carl. This is important because the intuition that there is a very important difference between Alan and Bob is reasonable enough. Unlike Alan, Bob has decided not to resist his predominant desire to humiliate Carl. The question is whether the best interpretation of that difference is that Bob, but not Alan intends to humiliate Carl. There are, in fact, a number of alternative interpretations of this difference, each of which is better than this one.

One crucial difference between Alan and Bob is in what they decide to do. On the basis of his moral views, Alan decides to "resist his desire" to humiliate Carl, whereas Bob decides not to resist his desire to do so, and decides instead to humiliate Carl. Mele's suggestion is that this difference means that Bob, but not Alan, intends to humiliate Carl. However, given that Alan's decision is to resist his desire, one might instead conclude that though Alan intends to humiliate Carl, he disapproves of this intention and desires to get rid of it. This suggests a quite general strategy for dealing with such putative counterexamples. What initially seems like a difference in first-order intentions turns out, upon further reflection, to be a difference in the agents' second-order evaluation of their first-order intention and/or a difference in their second-order desires regarding their first-order intentions. Finally, there is even room on my account for the difference to be one in the agents' second-order intentions. So it may be that (a) Alan, but not Bob, disapproves, all-things-considered, of his intention to humiliate Carl, or (b) Alan, but not Bob, wants to get rid of his intention to humiliate Carl, or (c) Alan, but not Bob, intends to get rid of his intention to humiliate Carl, or any combination of (a)-(c). So long as we are clear on the distinctions between first-order intentions, and second-order desires, second-order intentions, and evaluations (both first and second-order) such seeming counterexamples can be dealt with. Moreover, this way of dealing with such putative counterexamples is not ad hoc--once these alternative interpretations are suggested they seem natural and intuitive enough. Nor should we be left with any lingering suspicion that Alan does not intend to humiliate Carl. We have a thorough explanation of the ways in which Alan is averse to humiliating Carl which is perfectly compatible with his intending to humiliate him, and the fact that he has a predominant desire to humiliate him and a desire not to deliberate any more about whether to humiliate strongly suggests that he is committed to humiliating him. Indeed, it is not clear in what else an agent's commitment could consist.

One seemingly counterintuitive upshot of this account is that it seems to imply that when we engage in self-control we are trying to control our "unruly intentions" rather than our "unruly desires." The first thing to notice is that on my account this is a false dichotomy; aiming to control one's intentions is aiming to control one's desires. Second, it is not clear how much weight we should put on the residual intuition that in engaging in self-control we aim at whipping our unruly desires, but not our unruly intentions into line. It seems that this residual intuition can be seen as the result of a failure to distinguish evaluations from intentions--in ordinary discourse we often assume that if someone intends to [Phi] she must at some level think she ought to [Phi].

So far I have focused on the ways in which my account can make sense of Alan's having decided to resist his desire to humiliate Carl while nonetheless having an intention to humiliate Carl. On each of these interpretations, what Alan is doing is forming some second-order desire, intention, or evaluation, or perhaps some first-order evaluation, that is, in some sense, at odds with his first-order intention to humiliate Carl. There is, however, another interpretation of what Mele means in saying that Alan has "decided to resist" his desire to humiliate Carl.(13) It might instead mean that he has formed an intention not to humiliate Carl and eliminated any antecedent intention he might have had to do so. If this is what is meant, then on my account we must conclude that Alan no longer has a predominant desire to humiliate Carl, or no longer desires not to deliberate any more until new and relevant information comes to light, or comes to believe that humiliating Carl is beyond his control. The assumption that we can keep all those facts fixed and still have Alan not intending to humiliate Carl is one that is ruled out by my account; I must maintain that this interpretation of the case is incoherent. However, it is not at all obvious that it is possible to keep all those facts fixed and have Alan not intending to humiliate Carl. Once we have seen the ways in which we can make sense of our pretheoretic intuitions about the case via second-order desires, intentions, and evaluations (and perhaps first-order evaluations), we need some independent argument that this interpretation of the case is as much as coherent. It will no longer be sufficient to rely upon our common sense reaction to the case as presented by Mele. Once we have accommodated the common sense view that Alan may in some sense be settled upon resisting his desire to humiliate Carl while still satisfying the criteria of my analysis, we seem to have done all that is required by way of accommodating common sense.

Furthermore, there are strong grounds for thinking that this interpretation is incoherent. Consider how we might fill in the details of Alan's case:

Alan tries to avoid Carl at the party, but

Carl approaches him and makes some

especially annoying comments. Alan then

offers Carl several glasses of wine,

knowing that Carl has a weakness for

wine and is quick to get drunk and

embarrass himself if he drinks too much;

typically, Alan never offers people drinks

at parties. Alan then gets Carl to follow

him into another room where he knows

several people who Carl cares a great

deal to impress are. Then, Alan carefully

"pushes all of Carl's buttons"--he says

things that he knows will set Carl, in his

inebriated state, into a very embarrassing

rage. He then reveals some of Carl's

most intimate secrets and loudly insults

Carl. In all of this, Alan is guided only by

his predominant desire to humiliate Carl,

though he is also constantly saying to

himself things like, "You really shouldn't

humiliate Carl," "Control yourself!" and

"Why are you doing this? You decided

not to humiliate him!"

If Alan tried to excuse his behavior by claiming that he had no intention to humiliate Carl--that he didn't do it intentionally--I submit that his defense would fall on deaf ears, and rightly so. It is not that we do not think his not having had any intention to do it wouldn't to some degree excuse him--we would I assume grant this normative assumption. Rather, it is that we do not take seriously his claim not to have had any such intention. After all, he knew what he was doing and he went through elaborate measures to do it well. By stipulation, though, he was guided in all of this only by his predominant desire to humiliate Carl, with no intermediate mental states caused by that desire (even those who think intentions are sui generis allow that desires sometimes can cause actions without first causing an intermediate intention). Given that we would not buy Alan's claim that he had no intention to humiliate Carl, and given that his actions are explained by his predominant desire to humiliate him, it seems plausible to conclude that his predominant desire must at least partly constitute his intention to humiliate Carl. If, however, the predominant desire (partially) constitutes an intention in the case in which that desire issues in action, it would seem ad hoc to say that this very same desire does not (partially) constitute an intention in the case in which it does not issue in action, especially when the desire is accompanied by a desire not to deliberate any more about whether to perform the action unless new and relevant information comes to light. So in the end, it is not nearly so obvious as the objection presupposes that Alan does not all along have an intention to humiliate Carl, or that if, in deciding to resist his desires he comes not to intend to humiliate Carl, he must thereby no longer satisfy the criteria of my proposed analysis.(14) Common sense not only does not obviously speak heavily in favor of the objector's interpretation of the case; in fact, it speaks heavily against it.

One important feature of the case of Alan that I have so far not discussed enough is the way in which he comes to satisfy (d)--by being threatened by Martin the mind-reader with death if he deliberates any more about whether to humiliate Carl. This etiology might be thought to undermine our intuition that Alan is really settled upon humiliating Carl. After all, we do sometimes stop deliberating about whether to do something without reaching any decision about whether to do it. This is all fair enough, but we must be careful to distinguish decisions from intentions.(15) Decisions often issue in intentions, but this does not mean that all intentions are the result of decisions. For example, I might intend to avoid excruciatingly intense pains, but never have made a decision, conscious or otherwise, to do so.(16) Indeed, I take it to be a virtue of my account that it allows agents to have intentions that are not the result of any decision they have made. So I can allow that Alan did not decide to humiliate Carl, but still hold that he nonetheless intends to do so, in spite of morally disapproving of doing so, and perhaps in spite of intending to get rid of this intention. Of course, it should also be emphasized that Alan does not count as being settled upon humiliating Carl merely because he desires not to deliberate any more about whether to do so unless new and relevant information comes to light. He only counts as being settled upon humiliating Carl because he also has a predominant desire to do so.

One might object that Alan is not settled upon humiliating Carl simply because he believes, truly let us suppose, that there is some chance that he will avoid doing so. However, this would only count as an objection if we accepted the following thesis: Someone counts as settled upon [Phi]ing only if she does not think there is some chance she will avoid [Phi]ing. However, this thesis is rather clearly false, since we often intend to do things we are not absolutely certain we will do, or even that we are not absolutely certain we will try to do when the time comes. People make mistakes, have bad luck, and lose their nerve, and they know that about themselves, but still manage to count as having intentions. For example, I might now intend to confront a friend about something tomorrow even though I know that I might forget to do it, or not see the friend tomorrow, and that even if I do see her and do not forget, I might lose my nerve.

In focusing on the case of Carl, I have been focusing on the worry that (a)-(d) are not jointly sufficient for intending. Similar moves, however, will be helpful against the worry that some of those conditions are not necessary for intending. For example, Mele argues that having a predominant desire to [Phi] is not necessary for intending to [Phi]. Consider the following cases: (a) a man on a third-story porch of a burning building who intends to jump but who is shocked to find he lacks the motivation to jump, (b) a boy who judges it would be best to cut his hand as part of a "bloodbrother" ceremony but who stops short because of an aversion whose force he finds surprising (Mere 1992b: 165).(17) If these descriptions are coherent, it would seem that my account's claim that predominant motivation is necessary for intending is false. Of course, it might be that the man in the burning building really did intend to jump but once he got out on the ledge the intention evaporated. However, more interesting interpretations of such cases are available. Just as before, an appeal to higher-order states will be helpful. Perhaps the man in the burning building judges he ought to jump, judges he ought to intend to jump, desires to intend to jump, and intends to intend to jump, even though he does not yet intend to jump. He may even have falsely thought that he intended to jump prior to the "moment of truth," as we can sometimes be deluded about our intentions. Insofar as such moves are plausible we can explain why cases like (a) and (b) seem to describe real possibilities even though, given my account, they do not.(18)


Considering the alternative, its not too bad

at all.

--Maurice Chevalier

Having considered some important worries about my account, it will now be helpful to consider how it differs from some of its important competitors. I will consider Bratman's putatively non-reductionist view(19) and then Audi's reductionist view. Bratman's view is that A intends to [Phi] just in case:

(a) A has a "conduct-controlling"

disposition to [Phi], (b) A is disposed not

to deliberate any more about whether to

[Phi] unless new and relevant information

comes to light, and (c) A is disposed to

deliberate about intended means or

preliminary steps to [Phi] as well as about

more specific intentions, as when one

reasons from an intention to take a bus to

the library and information on a bus

schedule to the intention to take a

particular bus.

Though there are a number of important differences between Bratman's view and my own, the most interesting one is that where I rely upon the notion of a predominant desire to [Phi], Bratman relies upon a "conduct-controlling" disposition to [Phi]. This is one way in which he attempts to illustrate the contrast between intentions and desires. On his account, ordinary desires "are merely potential influencers of action" (Bratman 1987: 16). By contrast, an intention is a conduct-controlling proattitude. What this means is that, "if my future-directed intention manages to survive until the time of action, and I see that the time has arrived and nothing interferes, it will control my action then" (Bratman 1987: 16). There are, however, at least two ways of understanding what Bratman means by "I see that the time has arrived." Suppose I intend to call my friend at noon, and I see that it is noon. On one reading, I would thereby see that the time has arrived, and so either will call my friend or have lost my intention. This reading is too strong. It is perfectly intuitive to say in such a case that, though I never gave up my intention to make the call, because I am so absent-minded I did not remember to call even when I became aware that it was noon.

This suggests a second reading of "I see the time has arrived." Bratman might be understood as requiring that I am aware not only that it is noon, but that it is time to make my call. If this is the content of my belief, then it is fair to question my still having the intention if I do not go on to make the call. However, if this is all there is to having a conduct-controlling disposition to [Phi], it is too weak. Suppose that I have the belief that it is noon but do not have the belief that it is time to fulfill my intention. My intending to [Phi] should still dispose me to some degree in this case to [Phi], even if, as I have suggested, it is too strong to require that it guarantees that I will [Phi]. The predominant desire condition of my account is able to navigate between these two extremes. Unlike Bratman's account on the second reading, it is not too weak because it can explain why my intention, e.g., to make a call at noon, can lead me to make the call when I realize it is noon even if I do not realize it is time to make the call (even if I do not have a belief whose content is: "It is now time to make the call"). On the other hand, it is not too strong, like Bratman's account on the first reading, because it does not guarantee that I will make the call just because I am aware that is noon--I might not remember that I had intended to make the call at noon, so my predominant desire might remain dormant.

Unlike Bratman, Audi explicitly analyzes intentions into beliefs and desires. On his account, an agent intends to [Phi] just in case:

(a) she believes that she will (or probably

will) [Phi], (b) she wants to [Phi], and has not

temporarily forgotten that she wants to

[Phi], and (c) she either has no equally

strong or stronger incompatible want (or

set of incompatible wants whose

combined strength is at least as great), or

if she does have such a want or set of

wants, she has temporarily forgotten

that she wants the object(s) in question.

or does not believe she wants the

object(s), or has temporarily forgotten her

belief that she cannot realize both. (Audi

1973: 395, paraphrased)(20)

The most important difference between Audi's position and mine is that he embraces a strong belief condition, but as I have already noted, addressing that question requires another paper altogether. There is, however, another important difference between my account and Audi's. Audi does not require that your desire to [Phi] be predominant; you may have another desire that you realize is incompatible with [Phi]ing so long as you have "temporarily forgotten" that you want it, or do not believe that you want it, or have forgotten that you cannot realize the object of both desires. In this respect, Audi's account does not capture the element of commitment distinctive of intentions. Suppose I have a very strong desire to call my sick grandmother this evening (it is her birthday), and I have a much weaker desire to go camping today. These desires are incompatible, and I know them to be incompatible, but I have temporarily forgotten about my desire to call my grandmother, being absent-minded. Further, I do not have any other desires that I see as competitors with my desire to go camping, and I have not forgotten my desire to go camping, so I do go camping (believing that I will; I also believe, inconsistently, that I will call my grandmother, although this belief is both unconscious and not engaged in current mental processing). On Audi's account, I intend to go camping but do not intend to call my grandmother, but this does not square with a natural description of the case. One wants to say that although I intended all along to call my grandmother, I forgot to do so. Indeed, this seems to be at least part(21) of the explanation of why I will see myself as having acted irrationally in going camping--I never changed my mind about calling my grandmother, I still intended to do so, but I went on to do something I knew to be incompatible with doing so. The inclination to give this description reflects the tight connection between commitment and intention--since my desire to call my grandmother was extremely predominant, we are more inclined to say that I intended to call her than we are to say that I intended to go camping (although I may intend both).(22) By making one's predominant desire criterial of what one intends to do, regardless of whether one has forgotten that desire, my own account escapes this worry. Finally, Audi's analysis does not require a disposition not to deliberate unless new, relevant information comes to light, which is another respect in which it does not capture this element of commitment.


What is the natural expression of an

intention?--Look at a cat when it stalks a

bird; or a beast when it wants to escape.

--Ludwig Wittgenstein

Nonhuman animals and infants present a problem for my analysis.(23) The trouble is that Wittgenstein seems right--one cannot help seeing a cat stalking a bird as in some sense expressing an intention. However, on my account it is quite dubious whether nonhuman animals, or young infants, could have intentions. For to have an intention an agent must have a desire not to deliberate any more about something unless new and relevant information comes to light. To have such a desire, it might plausibly be argued, an agent must have the concept of deliberation.(24) Since it is doubtful that infants and most nonhuman animals have this concept, it is doubtful on my account that they have any intentions.

There is, of course, no problem with a nonhuman animal or an infant meeting (a)-(c) of my analysis--those conditions are not conceptually demanding. At this point, we should return to our original motivation for introducing condition (d) into the proposed analysis. Bratman's complaint against identifying intentions with predominant desires is that someone might have a predominant desire to do something but not be settled upon it in the sense that she might still be unequivocally motivated to deliberate more about whether to do it even in the absence of any new and relevant information. The requirement that one have a desire not to deliberate any more unless new and relevant information comes to light is meant to address this complaint in a relatively obvious way. Perhaps, though, in the case of nonhuman animals and infants, Bratman's worry is not such a worry after all. We differ from nonhuman animals and our infant selves in that we, but not they, are habitual deliberators. We are constantly making plans, refining old plans, evaluating our ends, etc. Cats and six week old children are not quite like this. The thought would be that in the case of nonhuman animals and infants we could take on a second sense of "intention" which is just like my sense of intention except that this sense would drop condition (d). In their case, one might urge, (a)-(c) would suffice for being settled upon [Phi]ing.

There are two problems with this move. First, it is not obvious that nonhuman animals and infants, at least some of them, do not deliberate quite a bit, even if they deliberate less than us and even though, let us allow, they lack the concept of deliberation. In that case, Bratman's worry would remain. Second, even if neither brutes nor babes deliberated prior to having the concept of deliberation, it is still conceptually possible, so far as I can see, that there be an agent who does deliberate like crazy but who has no concept of deliberation. Surely if there were such an agent, she would be able to have intentions if, as we are allowing, nonhuman animals and infants can have intentions, given that she would presumably be more advanced, conceptually, than they are. However, it is obvious that neither my original analysis nor the one suggested in the preceding paragraph would explain how such an agent could have intentions in a way that evades Bratman's worry. She lacks the concept of deliberation, so she cannot satisfy my original analysis, and satisfying the alternative analysis is not sufficient for her counting as intending, given that it leaves it completely open whether she is unequivocally committed to deliberating further even in the absence of new and relevant information.

We do need a second sense of "intention" to handle such cases, but we will need some sort of proxy for condition (d) that an agent lacking the concept of deliberation could meet. Otherwise, this sense of intention will fall prey to Bratman's worry. What is needed is some condition that ensures that it is unlikely that the agent engage in further deliberation unless new and relevant information comes to light. I take it that when one deliberates is, in large part, a function of one's beliefs and desires. This suggests the following proxy for condition (d):

(d') A's set of beliefs and desires are

such that it is unlikely that A will

deliberate any more about whether to

[Phi] unless new information which A

takes to be relevant comes to light.

Nonhuman animals and infants will be able to satisfy (d') without having the concept of deliberation, and (d') allows us to meet Bratman's worry, so I suggest that we adopt a second sense of "intention" in which (d) is replaced by (d'). Of course, this leaves open just how unlikely further deliberation must be, as well as a more precise account of what is meant by saying that it is unlikely. There may be no precise answer to the first question, since it may simply be indeterminate how firm an intention has to be to count as an intention at all. A discussion of the second question is important, but would take us well beyond our present scope. For present purposes, the rough idea of something's being unlikely is intuitive enough.

One might wonder why I have insisted on having two senses of intention rather than junking my original one in favor of the one just presented, in light of the greater generality of the second account (it can explain how both we and cats and babies have intentions).(25) However, I think it is worth keeping the original sense as well because it isolates a particularly useful and straightforward way in which agents who do have a concept of deliberation can manage to keep themselves from constantly deliberating--having a desire not to deliberate any more unless new and relevant information comes to light. Also, the original sense draws a more precise distinction between having a very weak intention and having no intention at all. To have an intention at all, in that sense, one must have some desire, however weak, not to deliberate more unless new and relevant information comes to light. Assuming that one either has or does not have such a desire, there will be a straightforward and precise answer to the question, "Does she intend to [Phi]?" Whereas in the second sense of "intention," given the vagueness of "unlikely," we have no such precision.


Anti-reductionists argue that intentions are "not reducible to complexes of beliefs and desires" (Mere 1992a: 300) but are "on a par with desires and beliefs" (Bratman 1987: 22). They embrace the irreducibility thesis, which threatens the Humean account of action explanation, according to which an action can be fully explained in a rationalizing way via a belief/desire pair. The arguments for this thesis have been elimination arguments, and have emphasized how thinking of an intention to [Phi] simply as a belief that one will [Phi], or as a predominant desire to [Phi], or both, either cannot capture the element of commitment distinctive of intentions or is incompatible with the normative constraints on intentions. Elimination arguments are sound only if they consider all plausible options, and such arguments have underestimated the resources available to Humeans. I have defended my own analysis at some length, in the end drawing a distinction between two senses of "intention," to accommodate the thought that nonhuman animals and infants can have intentions. However, even if my particular account is lacking, we have reason to hope that the strategy used in generating it could produce an analysis that would not be so lacking. As soon as the anti-reductionist articulates what is missing and why it is important, the Humean should be able to fill this gap with a belief or a desire.(26)

Received September 24, 1997


(1.) More accurately, our knowledge of our intentions provides provisionally fixed points in our deliberation.

(2.) In some cases like this, the appropriate thing to say is that I only have a desire to beat him fairly. However, not all cases are like this, and I am stipulating that this is not such a case. So, for example, I might still take some pleasure in winning if I cheated, though I would also feel very guilty about that pleasure and what I had done. Whereas if I only had a desire to win fairly, I would presumably take no pleasure at all in having won through deception.

(3.) It should be emphasized that the dispositions in question are only potentially belief-mediated, meaning that they may, but need not necessarily, lead to action via the influence of a belief or beliefs. Hence a proximal intention, say, to raise my arm now, or to try to raise my arm now, might lead to action without the actual mediation of any belief whatsoever, for all I have said so far. All that I require is that this proximal intention be potentially belief-mediated, so that, for example, if I had believed that a necessary means to my raising my arm right now is my saying "Raise that arm, buddy!" to myself, then that belief might mediate my intention to raise my arm right now and lead me to speak to myself in this way in service of raising my arm now. Thanks to Alfred Mele for emphasizing the need to be clear about this point.

(4.) Though I am willing to allow, for present purposes at least, that one might have such desires without being, even to some degree irrational, since on my account one might have such desires but not have the corresponding incompatible intentions, for reasons that will become apparent in the next section. All I am claiming at this point is that this possibility is not nearly so obvious as the one to which Strasser helps himself, but it is this possibility he needs to establish if he is to defeat the dominant desirist.

(5.) A stronger condition than (b) would be that you actually believe that [Phi] is within your control, but it is not clear to me why this stronger condition is forced upon us; if someone simply suspends judgment on whether [Phi] is possible it is not obvious that she cannot nonetheless intend to [Phi].

(6.) Though I do take them up in some detail in a separate paper on the possibility and irrationality of incompatible intentions.

(7.) Elsewhere, I explore the differences in normative constraints governing intentions in more detail, focusing on what Bratman has called "Rational Agglomerativity" and "Strong Consistency."

(8.) Thanks to Alfred Mele for bringing this issue to my attention.

(9.) It should also be noted that the proposed alternative makes quite general metaphysical assumptions about the nature of dispositions and their putative bases that many find dubious, but a discussion of those important worries would take us too far beyond the scope of the present paper.

(10.) I am indebted to Alfred Mele for mentioning this possibility.

(11.) I have deliberately left out certain details of the case that are not relevant for present purposes, in particular that Alan believes he most likely will humiliate Carl (since on my analysis such a belief is neither necessary nor sufficient for an intention to humiliate him).

(12.) Thanks to Alfred Mele for suggesting the device of an alleged mind-reader's threat.

(13.) In fact, Mele is committed to rejecting this interpretation, since, for him, a necessary condition of S's intending to [Phi] is that S lack the belief that he (probably) will not [Phi] (Mere 1992b: 146-48). Still, the position is worth considering.

(14.) One apparent problem with the argument in the text is that I seem to be assuming what is often called the "Simple View" of the relation between intentional actions and intentions--that anyone who intentionally [Phi]s must have had an intention to [Phi]. This is a problem not because the Simple View is obviously false (indeed, it has much to recommend it), but because it is obviously incompatible with my analysis and plausible Humean assumptions about the nature of intentional action. Following the standard Humean/causal line, I assume that an agent intentionally [Phi]s just in case her [Phi]ing is caused "in the right way" [ruling out the kind of "wayward causation" Davidson, Goldman and others have discussed in some way] by an appropriate belief/desire pair or (perhaps) by an appropriate desire. However, my analysis makes it clear that the mere fact that one intentionally [Phi]s, in this sense, is no guarantee that they have any intention, much less an intention to [Phi]. For one's [Phi]ing might be appropriately explained by a belief/desire pair but one might have no desire not to deliberate any more about anything, and hence have no intentions.

This is an important worry, and to accommodate it I will have to allow that "intention" goes ambiguous, having at least two senses (in section VI, I identify yet another sense). In the strict sense, someone does not intend to [Phi] unless she meets my criteria. In a much looser sense, someone intends to [Phi] if she intentionally [Phi]s. This sense is much looser--one need not even have a predominant desire to [Phi] to intend to [Phi]. Indeed, this is roughly the sense of intention that the Desirist had, so one obvious way in which it is weaker than the sense picked out by analysis is that the normative constraints distinctive of intentions in the strict sense do not apply to intentions in this much looser sense. Still, this is quite compatible with someone who intentionally [Phi]s intending to [Phi] in both the strong and weak sense, and my suggestion in the text is that we should understand Alan as intending in both senses. I admit, however, that the appeal to common sense only establishes, at most, that Alan intends in some sense to humiliate Carl, and the opponent of my analysis might insist that he only intends to do so in the very weak sense. Nonetheless, once it is allowed that in some sense Alan intends to humiliate Carl, it is not clear that the burden of proof is on me to show that he does not only intend to do so in some weak sense. If common sense unequivocally entailed that Alan in no sense intended to humiliate Carl then there would be a straightforward counterexample to my analysis; if instead common sense yields that in at least some sense he did intend to humiliate Carl, we have as yet been given no reason to think that he does not intend to do so in both a strong and weak sense. Insofar as the independent arguments given for my analysis are sound, it seems reasonable to assume that he does intend to do so in both a strong and weak sense.

(15.) We should also distinguish stopping one's deliberation from acquiring a desire not to deliberate any more unless new and relevant information comes to light. Since only the latter can make someone count as intending on my account, it is obvious that in at least one way my account allows for the possibility of someone stopping deliberating without having decided or without even having acquired any intention.

(16.) Alternatively, to take a relatively clear if incredibly fanciful case, I might come to have an intention because of an evil scientist's doing some weird brain surgery on me--such an intention would not be the result of any decision of mine. For further argument that intentions are not necessarily the result of decisions, see Mele 1992b: 140-146

(17.) Mele offers a range of other cases in this vein. For example his discussion of the person who intends to pet a snake at Mele 1992: 143, and his discussion of the smoker at Mele 1992b: 163.

(18.) One reason for being suspicious of these moves would be the thought that "what we are most motivated to do is not under our control to the same extent to which what we intend (or decide) to do is Evaluation does not have the same grip in both spheres" (Mere 1992b: 163). The picture seems to be that we have a great deal of control over our evaluations, and we thereby have a great deal of control over our intentions, but we do not thereby count as having a great deal of control over our desires. On this view, the man in the burning building can control his intentions--he really intends to jump--but he cannot control his desires--he lacks a predominant desire to jump. Unfortunately I lack the space in the present context to deal with the this worry in the detail it deserves. My suspicion is that the controversy will boil down to a disagreement about the nature of evaluative judgments and a related disagreement about what form, if any, of "judgment internalism" is true.

(19.) I say that Bratman's account is only putatively non-reductionist because his account is simply a dispositional one, and I take it that he would allow that those dispositions are potentially belief-mediated ones, and hence his account is in the end a reductionist one, despite his claims to the contrary.

(20.) Actually, this is a bit of an oversimplification. Audi distinguishes end-directed intentions and simple intentions, and gives each its own analysis, though these analyses are parallel. Though I have overlooked this distinction for expository purposes, it would be simple enough to integrate with my account.

(21.) Another part of the explanation might be that I judged that I ought to call her, but this is not essential to the case. I might have made a conscious decision to do what I thought I ought not to have done, in which case I will still have been akratic in not following through on that intention, as some of Mele's cases indicated.

(22.) Although in a very weak sense I may intend to go camping as well; see footnote 14.

(23.) Thanks to Alfred Mele for motivating me to discuss this important issue.

(24.) Although this is not obvious. See Harman 1993.

(25.) This, of course, is in addition to the third and very weak sense of "intention" isolated in footnote 14. Since that very weak sense of "intention" is not the sense in play in normative principles like IHI, I have in this section tried to find a stronger sense of "intention" so as not to prejudge whether nonhuman animals, etc., are bound by any principles of practical reason like IHI.

(26.) I am indebted to Nicholas Bull, Geoffery Sayre-McCord, Sean McKeever, Jon Tresan, and especially Michael Smith and Alfred Mele for very helpful comments on earlier drafts of this paper.


Audi, R. 1973. "Intending," Journal of Philosophy 70: 387-402.

Beardsley, M. 1978. "Intending," in Goldman, A. and Kim, J., eds., Values and Morals: Essays in Honor of William Frankena, Charles Stevenson, and Richard Brandt. London: D. Reidel Publishing Company.

Brand, M. 1984. Intending and Acting. Cambridge: MIT.

Bratman, M. 1987. Intentions, Plans, and Practical Reason. Cambridge: Harvard University Press.

Carruthers, P. 1992. The Animals Issue. Cambridge: Cambridge University Press.

Davidson, D. 1980. Essays on Actions and Events. Oxford: Clarendon Press.

Harman, G. 1993. "Desired Desires," in R. Frey and C. Morris, eds., Value, Welfare, and Morality. Cambridge: Cambridge University Press.

--. 1986. Change in View. Cambridge: MIT.

Hertzburg, L. 1995. "On Being Moved by Desire," Philosophical Investigations 18: 250-263.

Hume, D. 1888. A Treatise on Human Nature, second edition, L. A. Selby Bigge, ed.; P. H. Nidditch, rev. Oxford: Clarendon Press.

Kant, I. 1785. Grounding of the Metaphysics of Morals. Indianapolis: Hackett.

McCann, H. 1986. "Rationality and the Range of Intentions," in Midwest Studies in Philosophy X: 191-212.

Mele, A. 1992a. "Akrasia, Self-Control, and Second-Order Desires," Nous 26: 281-302.

--. 1992b. Springs of Action. New York and Oxford: Oxford University Press.

--. 1987. Irrationality. New York and Oxford: Oxford University Press.

Pink, T. 1996. The Psychology of Freedom. Cambridge: Cambridge University Press.

Platts, M. 1979. Ways of Meaning. Routledge and Kegan Paul.

Strasser, M. 1992. Agency, Free Will, and Moral Responsibility. Wakefield, NH: Hollowbrook Publishing.
COPYRIGHT 1998 North American Philosophical Association
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1998 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:David Hume
Author:Ridge, Michael
Publication:American Philosophical Quarterly
Date:Apr 1, 1998
Previous Article:The middle ground in moral semantics.
Next Article:Persons and psychological systems.

Terms of use | Privacy policy | Copyright © 2019 Farlex, Inc. | Feedback | For webmasters