Ethical Theory

In S. F. Delaney, Michael J. Loux, Gary Gutting, W. David Solomon, Synoptic Vision: Essays on the Philosophy of Wilfrid Sellars (University of Notre Dame Press, 1977).

Ethical Theory
W. David Solomon
Wilfrid Sellars' writings on ethical theory have received less attention than any other aspect of his work. This is doubly unfortunate since these views are an essential part of the comprehensive philosophical system Sellars has been developing for the last twenty-five years and an understanding of them is required for an understanding of that system. But, also, these views are interesting and important in their own right. At least I will argue in this essay that Sellars' views on ethical theory are both profound and original; whether they are "true" is, of course, another question.
Having lamented the neglect Sellars' ethical views have encountered, however, one must also admit that Sellars is not altogether blameless in the affair. There are a number of reasons, I suspect, why moral philosophers have not given Sellars the attention he deserves. First, his ethical writings are, of all of his writings, the most painfully dialectical in form. He almost always develops his own positions in opposition to various views that he draws from the writings of other moral philosophers, both classical and modern. In trying to spin truth from many-stranded error, however, Sellars frequently leaves the reader unclear about the exact nature of the issues under discussion. A second reason for the neglect of his work derives from his tendency to integrate his ethical theory with his broader views in philosophy of mind and philosophy of language as he goes along. Sellars, perhaps more than any other recent philosopher who has written on ethical theory, has tried to develop an ethical theory while drawing out its full implications for other areas of philosophical inquiry. Indeed, some might argue (and not without plausibility) that Sellars is at least as interested in showing how moral discourse and moral thought might fit within his broadly empiricist philosophy of mind as he is in developing positions within ethical theory for their own sake. Be this as it may, the reader who is unwilling to follow Sellars down various side paths (as it may seem) into issues that may appear to belong more properly to philosophy of mind, action theory, or metaphysics will be unlikely to persevere long in the study of his ethical writings.
The comprehensiveness of Sellars' aims in his ethical writings, while doubtless creating difficulties for the reader, also provides much of the interest that his writings have. While in much of recent ethical theory one has the impression that views in philosophy of mind, metaphysics, and philosophy of language are invoked in an ad hoc fashion to suit the ethical-theoretical purposes of the writer, one never has that feeling in reading Sellars. Indeed, for Sellars, the tendency of an ethical theory to fit within the constraints of an adequate philosophical "system" seems as important a criterion for its acceptability as others more purely "ethical" criteria.
These difficulties in approaching Sellars' ethical writings have largely dictated the strategy I adopt in this essay. I attempt here to give an exposition of Sellars' ethical views that will make them available to a larger audience and tempt others to the critical work that I must leave largely undone. This general strategy involves two tactical devices: first, I have tried, insofar as is possible, to set out Sellars' main arguments and conclusions freed from the dialectical garb in which he clothes them; second, I have downplayed, again insofar as is possible, the dependence of Sellars' ethical views on his views in other areas of philosophical inquiry. There is a danger in each of these maneuvers, since neither the dialectical nor systematic aspect of Sellars' philosophical enterprise is an accidental feature of it. For this reason I present my exposition below as merely a way into Sellars ethical thought; I present it frankly as a reconstruction and not as a substitute for grappling with Sellars on his own terrain.
Sellars' work in ethical theory moves largely within two distinguishable contexts of discussion. At one level, he works within the classic metaethical tradition in attempting to give an account of the logical structure of moral judgments, in particular ought-judgments. His conclusions here constitute an account of the logic of 'ought', which, he alleges, combines the strengths of the three classical metaethical views: naturalism, intuitionism, and noncognitivism. At another (partly overlapping) level, however, Sellars deals with the set of issues that have classically divided deontological and teleological (or Kantian and utilitarian) ethical theories. Here he concludes that deontologists were largely correct in their account of the motivational base of morals while teleologists were more nearly correct in their account of the ultimate justification of moral principles.
In the discussion below I try to separate, insofar as this is possible, these two contexts of discussion. In section II suggest a model for understanding the general problematic of classical metaethics and locate Sellars' own views on metaethics within this model. In the following four sections I discuss various strands of his metaethical view as follows: in III I discuss his arguments against emotivist and imperativist views of the Harean variety; in IV I discuss his general account of practical reason and the role that expressions of intention play within that account; in V I show how his analysis of ought-judgments as complicated expressions of intention can be reconciled with the universalizability of moral judgments: and in VI I explore his account of the intersubjectivity of moral judgments. In VII, I turn to the second context of discussion mentioned above and locate Sellars' views on the justification of moral judgments as a "compromise" between deontological and teleological views.
As the discussion proceeds, it will become evident that the distinction between what I have called the two contexts of discussion is more than a little artificial. Even if this distinction must be transcended at the end of this essay, however, I hope that it will serve, at least for a while, as a device for helping us to see some things in Sellars more clearly than would be possible without it.
II
Sellars' views on ethics, like his other philosophical views, grow out of a profound grasp of the positions and controversies within twentieth-century analytic philosophy. For this reason, I want to locate his general stance within a picture of metaethics painted with fairly broad strokes. While it may seem that we are moving quite far from Sellars in this section, I think that some such characterization of his place in the larger disputes of twentieth-century ethics is essential for understanding him. Indeed, the difficulty in locating Sellars within the standard taxonomy of ethical views is one more reason why his work has attracted so little attention from other moral philosophers.
There is a certain common view of the problematic of ethical theory which is largely shared by Anglo-American moral philosophers from Moore to Hare. I will call the tradition that embodies this view classical metaethics (CM) and those philosophers who work within this tradition classical metaethicists (CMists). CM is continuous with earlier ethical theory in that it conceives of the two fundamental problems in ethics as what I shall call the problem of justification and the problem of motivation. The problem of justification can be broadly characterized as the problem of determining the nature of the evidence that is relevant to the truth or falsity of moral judgment and of explaining the relation between this evidence and a moral judgment when the moral judgment is justified. (Of course, a possible view here is that there is no evidence relevant to the truth or falsity of moral judgments for the very good reason that there is no significant sense in which moral judgment can be true or false.) The problem of motivation can be broadly characterized as the problem of determining how assent to moral judgments, or commitment to moral principles or ideals, or the susceptibility to moral feelings (possession of a moral sense, in an earlier tradition) "relates to" one's behavior or dispositions to behavior.
If the recognition of these two problems as fundamental problems in ethics shows the continuity of CM with earlier ethical theory, the particular way in which these problems were construed within the tradition points to its distinctive character. Both problems were construed with reference to the following trileveled model:¹

L1) statements of fact,
↓
L2) moral judgments,
↓
L3) actions.

The problem of justification and the problem of motivation are seen respectively in CM as the problem of the nature of the arrow that connects (L1) to (L2) and the nature of the arrow that connects (L2) to (L3). That there is some sort of connection, however weak, at both levels has seemed obvious to almost everyone; even the most perfunctory reflection on the place of moral thought and talk in our everyday life shows both that we normally appeal to facts in support of our moral claims and that we normally feel moved to act in accord with moral judgments.
If there has been broad agreement that there is some connection between the factual beliefs a man holds, the moral views he espouses, and his actions, however, there has been constant and deep-seated disagreement among CMists about the nature of these connections. At the most general level the disagreement has been over whether the transitions at either point have the character of necessity or whether they are merely contingent. One might hold, for example, that certain facts entail certain moral judgments and that real assent to a moral judgment logically guarantees that one will act (or be moved to act) in accord with the judgment, thus making the connections at both points matters of logical necessity. One might hold to the contrary, however, that the connections between both facts and moral judgments and between moral judgments and actions can vary from person to person and that they depend not on some feature of moral judgments that can be grasped by logical or semantical principles but rather on the peculiar psychological makeup of individual moral agents. Thus one might hold that whether a certain set of facts stands in an evidentiary relation to a moral judgment for some person depends upon that person's particular aims, goals, or aspirations. In the same way, one might hold that whether some person's assent to a moral judgment "moves" him to action will depend on that person's attitude toward morality -- i.e., it will depend on whether he is "for" or "against" morality.
What has been most characteristic of CM has been its adherence to two not unrelated methodological points with respect to this debate. First, CMists have presupposed that the nature of the two points of transition is to be determined by an analysis of the meaning of so-called ethical terms -- usually 'good', 'right', and 'ought'. Second, they have presupposed that the crucial question about the nature of the points of transition has been whether these transitions involve entailment relations.
With this broad characterization of the problematic of CM before us, it is relatively simple to set out the positions of the standard metaethical views. Naturalists have argued that the meaning of ethical terms is exhausted by the empirical criteria of their application and, hence, that there is an entailment relation tying (L1) to (L2). They denied, however, that any strong logical tie holds between (L2) and (L3).² Noncognitivists, on the other hand, have rejected such an entailment relation between (L1) and (L2); in searching for their reasons for rejecting such an entailment, we have to turn to their treatment of the transition from (L2) to (L3). It was at this point that noncognitivists found the strong logical connection that they had failed to find between (L1) and (L2). They were struck by the close connection between the assertion of moral judgments and actions, and this provided them with a reason, they supposed, for rejecting the entailment relation between (L1) and (L2).³ The precise explanation of why a strong logical relation between (L2) and (L3) should be thought to preclude such a relation between (L1) and (L2) is a complicated and lengthy matter, but the fundamental reason was that it was felt that there could not be entailment relations both between (L1) and (L2) and between (L2) and (L3). It seems clear that such an entailment cannot hold, however, since if it did, two persons who shared identical factual beliefs could not, without some logical fault, disagree about what actions to perform in similar situations. Most philosophers have accepted as a datum of moral philosophy, however, the view that agreement on factual beliefs need not entail agreement on ideals, aspirations, or choice of lifestyle.
The battle lines between naturalists and noncognitivists are drawn, then, between those ethical theorists who claim that there is an entailment relation between (L1) and (L2) and deny that such a relation holds between (L2) and (L3) and those who claim that there is an entailment relation between (L2) and (L3) but deny that there is one between (L1) and (L2). Each view is in the unhappy position of having to deny what seems from the point of view of the opposing position obvious: naturalists have to deny that moral judgments unqualifiedly commit their users to action, while noncognitivists have to deny that intersubjective standards of factual evidence for moral judgments can be found.
Fitting the intuitionists into this debate is a complicated matter in that it is not at all clear whether intuitionism was determined to have it both ways or whether it was content to have it neither. Seen in one way, intuitionism seems to render mysterious both the connection between facts and moral judgments and the connection between moral judgments and actions. In trying to hold onto the intersubjective character of moral discourse while emphasizing its uniqueness and nonreducibility, intuitionists were forced to postulate special moral (nonnatural) facts in virtue of which moral judgments are true and cognitively meaningful. But intuitionists were never able to clarify the relation between "ordinary" facts and nonnatural facts, thus rendering the move from (L1) to (L2) mysterious; nor were they able to make perspicuous the connection between the apprehension of their peculiar moral facts and action in accord with them, thus rendering the move from (L2) to (L3) mysterious.
If in one sense, however, intuitionism must be considered a failure, there is another sense in which some have thought that it succeeded at least insofar as it managed to keep its heart in the right place. When intuitionism was propounded and defended by its most sophisticated proponent, Prichard it seemed to have at least a sound grasp of the weakness of opposing metaethical views and a dogged determination to avoid all of these weaknesses. Thus, Prichard was determined to fashion an account of the logic of 'ought' that would account for (1) the unique and irreducible character of moral discourse (pace naturalism); (2) the intersubjective character of moral discourse (pace noncognitivism); and (3) the action-guiding character of moral discourse (pace naturalism and certain other forms of intuitionism).⁴
If we now return to Sellars from this excursion through the intricacies of CM, we can best characterize his position within this tradition as one that attempts to retain Prichard's fundamental commitment to the uniqueness, intersubjectivity, and action-guidingness of moral discourse while removing some of the mystery about how these features of moral discourse can cohabit. Sellars says of Prichard's attempt to defend an account of moral judgments that emphasizes both their "fact-stating" character and their motivational force.

My difficulty with Prichard on this point (as with the many philosophers who have insisted that moral thinking qua thinking is "conative", no "mere blend" of thinking and "emotion") is that he offers not even the beginning of a satisfactory analysis of this phenomenological insight, an analysis which would account for the fact that moral thinking differs from, but resembles, other forms of thinking by relating both to the fundamental categories of an adequate philosophy of mind. Prichard's grasp of the distinctive traits of moral thinking exhibits the combination of thinness and acuteness which is characteristic of his philosophy. (IIO, 161.)

Sellars' attempt to put flesh on the bare bones of Prichard's insights leads him to construct a comprehensive model for practical reasoning within which he locates moral thought and talk. Sellars is concerned, that is, to identify moral judgments as one form of practical judgment and to explore the relationship between them and other practical judgments. By locating moral judgments within the context of practical reasoning, Sellars assures that these judgments will have practical force. In order to assure the uniqueness and intersubjectivity of moral judgment, Sellars has to characterize the special form that practical thinking has when it is moral thinking. I will explore the special character that Sellars attributes to moral thinking toward the end of this essay; in the next few sections I will concentrate on his attempt to integrate moral thinking into a comprehensive view of practical reason.
Ill
Sellars, as we have seen, wants an account of moral judgments that will do justice to their intimate relation to action. The contingent connection to action that naturalistic and certain intuitionistic views (e.g., Ross') allow is too weak, he thinks.⁵ In searching for an account of moral judgments that does justice to their practical character, it is only natural that Sellars should turn first to the noncognitivist views developed by emotivists like Ayer and Stevenson and to the imperativist analysis of moral judgments defended in recent years by R. M. Hare. Each of these varieties of noncognitivism has as its leading idea the view that the primary role of moral language (and of evaluative language generally) is to "influence" behavior and dispositions to behavior. Emotivists have argued broadly that moral judgments are best seen as primarily expressions of attitude that, from the point of view of the speaker, appear as symptoms of the presence of certain attitudes and, from the point of view of the audience, appear as stimuli eliciting or reinforcing certain attitudes in others. If moral judgments are thus tied intimately to the attitudes of their users, the necessary connection between the uses of these judgments and action seems assured, since attitudes are just, in one sense, dispositions to act in certain ways.
While emotivism is attractive in doing justice to the link between moral judgment and action, however, Sellars feels that in this case the price for justice is set too high.⁶ Emotivism preserves the practical force of moral judgments by removing moral discourse from the conceptual realm altogether. The constant recourse of emotivists to grunts and groans, boos and yeas, and expletives and interjections as the primary analogues of moral judgment signals their commitment to the broadly nonconceptual character of moral discourse.
But how do we show that moral discourse does belong within the conceptual order? Perhaps the emotivists were right in treating moral discourse as though it had a completely different character -- i.e., a nonconceptual one -- from fact-stating discourse. In response to this query, Sellars invokes a criterion that he has used in other places to mark off those bits of language, broadly conceived, that are conceptual from those that are not. "The criterion l propose is that a word stands for a concept when there are good arguments in which it is essentially involved." (SE,408.) One might be inclined to think at first that this criterion as used against the emotivists is question-begging, since one of their central claims was that there really are no good arguments in ethics in the same sense as there are good arguments in, say, science and mathematics. But when emotivists denigrated (or reinterpreted) the role of inference and argument in ethics, they had in mind arguments of this form:

O is P₁, P₂, P₃ (where P₁-P₃ are empirical properties).
Therefore, O is E (where E is an ethical "property").
And if Sellars needs to suppose that arguments of this sort are good ones in order to use his criterion, he would indeed be begging the question. There are many arguments, however, that involve moral terms but are not of this character. Consider the following two arguments:

Jones ought to do A and B.
Therefore, Jones ought to do A.
or

All objects of kind K are good.
This is an object of kind K.
Therefore, this is good.
Each of these arguments involves essential uses of moral terms and only someone determined to hold onto emotivism at all costs would deny that they are good ones. More important, they do not violate the basic strictures of emotivism against "fact-value" arguments and hence are not question-begging in the way the first argument mentioned above is. If we accept Sellars' criterion of membership in the conceptual order (and it certainly seems plausible), emotivism must be rejected on the grounds that it cannot do justice to the conceptual character of moral terms and moral discourse generally.
Of course, this line of criticism as developed by Sellars against emotivism is not original with him. From its inception, emotivism was confronted with the charge that it could not do justice to the role argument and inference play in ordinary moral thought and talk. The evermore complicated recipes that emotivists concocted for integrating the "descriptive" and "emotive" meaning of moral judgments testifies to the extent to which even they were moved by this objection. This general line of argument was developed in an especially powerful way by R. M. Hare, who wanted to preserve what seemed right in emotivism while exorcising its antirationalist tendencies. Sellars, indeed, has been strongly influenced by Hare, and he says in one place:
The neo-imperativist approach of R. M. Hare is, in my opinion, sufficiently close to the truth to be a useful point of departure for the ideas I wish to develop. While I think that something like his account of the concept of ought is true, I do not think that it will do as it stands. Indeed, I think that its very closeness to the truth has enabled it to obscure essential points about the concept of obligation. (IIO, 162.)
By setting out Hare's approach to the action-guidingness of moral judgments and then examining Sellars' objections to Hare's account, we are led quickly to the heart of his theory.
Hare argued in The Language of Morals that while the emotivists were right in taking the action-guiding force of moral language to be its central feature, they went wrong in confusing the notion of "getting someone to do something" with ' 'telling someone to do something''. As he says there, "The process of telling someone to do something, and getting him to do it, are quite distinct, logically, from each other".⁷ More generally, Hare thought that the emphasis emotivists placed on the tendency of moral judgments to affect causally the behavior of hearers distracted attention from the complicated nexus of logical relations in which they were caught up.
Hare's own view of the proper analysis of moral judgments involved an attempt to retain the action-guidingness of moral judgments in a fairly strong sense but also to preserve the logical connections among moral judgments and between moral judgments and nonmoral judgments. The key notion in his analysis is that of the, imperative. and his central claim is that all properly evaluative judgments (including, of course, moral judgments) either are imperative or entail them.
If Hare's account is to handle both the action-guiding character of moral discourse and its conceptual (inference-allowing) character, he must show (1) that imperatives are action-guiding and (2) that there are patterns of inference involving imperatives. The first task is fairly straightforward; if there are any action-guiding pieces of discourse, surely imperatives must belong to that class. Hare makes this clear beyond any question by characterizing the assent conditions for an imperative in such a way that to assent candidly to an imperative is (physical and psychological impediments aside) to act in accord with it.⁸ The second task above is not so straightforward, however. In particular, since imperatives are neither true nor false, it has seemed to many persons that it is improper to talk of inference here at all. Hare develops a sophisticated apparatus for understanding the deep structure of imperatives, however, which makes his claim that imperatives can play the role of either premise or conclusion in inferences a good deal more plausible than it might at first seem.
I do not have the space here to set out Hare's full defense of the notion of imperative inference, but I can briefly note its leading idea. Hare construed both imperatives and indicatives as having a structure that divides roughly into two parts: a content part and an operator that signals what the speaker is "doing" with the content. Thus, the two sentences
(1) Jones shut the door, (said of Jones)
(2) Jones, shut the door, (said to Jones)
have the depth structures respectively of
(1') (Jones shutting the door), yes.
(2') (Jones shutting the door), please.
where the 'yes' in (1') signals that the content is being asserted while the 'please' in (2') signals that the content is being commanded. Hare argued that inference involves relations among the phrastics (as he called the content) of sentences and that an imperative neustic (as he called the operator) does not absolve its phrastic from its logical relations.
Whatever one may think ultimately of Hare's particular explanation of how imperative inference is possible, however, one can hardly fail to be impressed by the examples that he proposes as embodying such inferences. Thus, one of his examples goes as follows:⁹
(1) Take all of the boxes to the station.
(2) This is one of the boxes.
(3) Therefore, take this to the station.
The conclusion of this argument surely seems to follow from the premises in as straightforwardly a logical way as the conclusion of the following purely indicative argument:
(1) All of the boxes are at the station.
(2) This is one of the boxes.
(3) Therefore, this is at the station.
But if these two arguments equally involve inference, then Hare has made good his claim that there is imperative inference.
It is at this point, however that Sellars parts company with Hare. He objects, that is, to the central role Hare assigns to imperatives in his account of the logic of moral judgments. He develops two distinguishable but intertwined arguments against Hare's position. He argues first that if the imperatives that Hare places at the center of his account are construed as commands, the account fails because commanding as a performance presupposes principles of obligation. If Hare avoids this objection by construing the use of imperatives in a narrower way that does not involve the explicit invocation of the performance of commanding, Sellars argues that the imperatives so construed can play no role in inference.¹⁰
Let us look first at the argument against commanding as the primary analogue of moral judgments. This argument involves the following three claims:

Commands are performances that can only "come off" in certain circumstances.
Included among the circumstances that are the necessary conditions of a command's coming off are certain relationships between the person who issues the command and the person to whom it is addressed.
These special relationships are defined by principles that involve 'ought'-talk.
The thought behind this argument is the Austinian one that commands as performances can only be successfully carried off if the commander in some sense appropriately (by right) issues the command. Looking at the same situation from the other end, we can see that the person to whom the command is addressed is really commanded to do something only if, again in some sense to be specified, he ought to do whatever he is commanded to do. But if the performance of commanding is to shed any light on the logic of ought-judgments in particular and normative language generally, it is necessary that an understanding of commanding performances not involve the use of normative language. Since commanding performances do "depend on" norms that are acknowledged by the performers in the commanding situation, it would be a mistake to construe the use of moral judgments as just a "special kind" of commanding.
Even if one agrees with this argument, however, one might claim that it really does not touch Hare's position, since Hare need not understand imperatives as full-blown commands. There are many uses of sentences in the imperative mood, after all, that do not constitute commands. Directions in cookbooks, for example, are frequently couched in the imperative mood, but it would be absurd to suppose that for this reason we must think of the author of the book as issuing commands.
In order to counter this "thin" interpretation of the imperativist position, Sellars develops the second argument that we mentioned above. This argument depends on a distinction that he draws between three ways in which a speaker's intentions relative to the actions of another person might become known to that person (Cf. IIO, 166). Thus, in order of strengthening involvement with one's audience, we have:
(1) Merely intending out loud,
(2) Telling of one's intentions,
(3) Telling someone to do something.
Examples of these three ways in which one's intentions become public would be:
(1') Jones simply overhearing Smith say, "Jones shall do A,"
(2') Smith telling Jones, "I intend that you do A" (said in a merely autobiographical way),
(3') Smith telling Jones, "Do A".
Of these three ways of "expressing" one's intentions, (3) seems to capture most closely, Sellars argues, what the thin interpretation of imperativism would place at the center of an analysis of ought-judgments. That is, with (3) we have a bare telling-to stripped of the performative trappings in which commanding is clothed.
Sellars claims, however, that even though tellings-to need not be as encumbered with performative overtones as commandings, they are, nevertheless, performances of a sort. For one thing, tellings-to require an audience, and he goes on to argue that this feature is all that is required to show that they cannot figure in inference either as premises or as conclusions.¹¹
Sellars' argument that no telling-to can be the conclusion of a reasoning is summed up in the following passage:
It makes sense to suppose that an expressed reasoning could have occurred without being expressed; and if so, there cannot be such things as expressed reasonings the premises or conclusions of which are promisings, tellings to, or tellings that. In particular, there is no such thing as imperative inference. (IIO, 171.)
It may seem puzzling initially why Sellars should require that all reasonings, or inferrings, be capable of being rehearsed without an audience. This requirement may seem ad hoc and without support. Certainly Sellars spends little time defending it; but a little reflection shows, I think, that it really is as obvious as he takes it to be.
Consider what it might be about some linguistic episode that could make it necessary that it occur only in public, i.e., in interaction between two language users. The only reason sufficient to explain this would seem to be that the appropriateness of that linguistic episode must depend partly at least on the situations of the speaker and the audience. It is in virtue of my wants or intentions or my assessment of your submissiveness, perhaps, that renders it reasonable for me to tell you to do something. Tellings-to, to put it a different way, are actions that can be deliberate or hasty, justified or unjustified, successes or failures. And, of course, it makes sense to speak of these actions, i.e., tellings-to, as being supported by reasons. But for an action, even of the telling-to sort, to be supported by reasons is not for it to be the conclusion of a reasoning. Indeed, since a telling-to, insofar as it is an action, is something that can be done for reasons, it cannot itself be a conclusion of an inference. To think otherwise would be to confuse the decision or intention to tell someone something with the telling-to performance itself.
In the end, then, it is the fact that tellings-to, like commands, are actions that make the notion of imperative inference unacceptable to Sellars. But in disposing of Hare's prescriptivist account of the logic of moral judgment, Sellars may appear to have exhausted the ways in which contemporary ethical theory can account for the action-guidingness of moral judgments. He has rejected Prichard's view because of its essentially ad hoc character; he has rejected emotivist accounts on the grounds that they remove moral discourse from the conceptual order altogether; and now he rejects Hare's view on the ground that there can he. no imperative inference. Where can he turn for an account that will do justice to the practical force of moral judgments but that will avoid the difficulties of the available views?
Sellars resolves this difficulty by claiming that the fundamental conceptual item out of which moral judgments are constructed is the expression of intention. By giving an account of expressions of intention that shows both how they play a role in inference and how they are connected necessarily to action (through their relation to volitions). Sellars accomplishes, he thinks, what Hare set out to do with imperatives. In the next section I will set out Sellars' account of practical reason and examine his notion of an expression of intention as it is defined by the role it plays within this structure.
IV
It is impossible to understand Sellars' account of the logic of moral judgments without setting out in some detail his general view of the nature and structure of practical reason.¹² In particular, his account of practical reason is partially determined by his goal of fitting such an account within his substantive views on the philosophy of mind, according to which mental acts are construed on the model of overt utterances for which they are seen as the causal antecedents. Also, he relies heavily here on certain views about the nature of implication that he has developed elsewhere, including the controversial view that causal laws are to be treated as material rules of inference. Having admitted the difficulties here, I will proceed nevertheless to set out briefly his account of practical reasoning.
The most important feature of Sellars' account of practical thought is his commitment to the view that there is such a thing as practical reasoning. Unlike many other philosophers who have been attracted to a broadly empiricist philosophy of mind, Sellars believes that the ordinary processes of deliberation and practical thinking do involve reasoning that is worthy of the name. Sellars' account of practical reasoning encompasses three main topics: (1) intentions; (2) volitions; and (3) the processes of practical inference. I will take up these topics in that order. The central practical statement, according to Sellars, is the expression of intention. Examples of Els would be the following:
(1) I will go to town in ten minutes.
(2) Jones shall pay me the money.
(3) You shall bring it to me.
Sellars has developed a technical use of 'shall' that, regardless of person, signals an expression of intention. The function of 'shall' as an operator that converts indicative sentences into Els is brought out by the following translation of (1) through (3):
(1) Shall (I will go to town in ten minutes)
(2) Shall (Jones will pay me the money)
(3) Shall (You will bring it to me)

Two further points will make the notion of expression of intention clearer. First, Els are to be distinguished from biographical statements describing or reporting intentions (SM, 185). Such biographical statements are true just in case the appropriate person has the intention imputed to him. Expressions of intention on the other hand are neither true nor false (which is not to say, of course, that they cannot be criticized in various ways for being defective, e.g., on the grounds that they are less than candid). Second, expressions of intention always express the speaker's intentions; although I can report your intention, I cannot express it (SM, 184). When I intend that Jones go to town, I am intending to do whatever may be necessary (with standard qualifications) to bring about Jones' going to town. When I report that Jones intends to go to town, I am, of course, expressing no intention at all. This point suggests that we use subscripts on the shall-operator to indicate whose intentions are being expressed. Thus,
(4) Shall (Jones will go to town) (said by Jones)
(5) Shall (Jones will go to town) (said by Smith)
will become, respectively,
(4') Shall_Jones (Jones will go to town)
(5') Shall_Smith (Jones will go to town)

Intentions are practical in that they inevitably (changes of mind and so forth aside) turn into volitions. A volition, on Sellars' view, is an intention whose time has come. Whereas expressions of intention manifest my commitment to act at some future time (perhaps precisely datable, perhaps not), volitions are commitments on my part to act here and now. It is, Sellars argues, part of the conceptual framework of practice that intentions ripen into volitions and that volitions issue in actions. The ordinary sequence of events would proceed as follows:
(1) Shall_Jones (I will leave in three minutes) (said at 2:00)
(2) Shall_Jones (I will leave in two minutes) (said at 2:01)
(3) Shall_Jones (I will leave in one minute) (said at 2:02)
(4) Shall_Jones (I will leave now) (said at 2:03)
(5) Jones leaves.
The volition "occurs" at (4) and issues in action at (5).
The details of the connections here need not detain us, but it is important to see that the connections (ceteris paribus clauses suitably understood) are necessary ones: to learn the language of intentional expressions is to learn the transitions represented in this sequence. The transitions involved in the moves from (1) to (4) are intralinguistic moves, licensed by rules for the proper use of the intentional 'shall'.¹³ When we move from (4) to (5), however, we move within a language to an action. Sellars argues that there are linguistic rules that license not only moves within languages -- intralinguistic moves -- but also moves from extralinguistic items to linguistic items -- what he calls language-entry transitions -- and from linguistic items back to extralinguistic items -- what he calls language-exit transitions. Thus, some concepts (e.g., 'or') may be completely defined by rules licensing intralinguistic moves; other concepts (e.g., 'red' and 'this') may be partially understood in terms of language-entry devices. A 'shall' thought that expresses a volition is a paradigm case of a language-exit device (cf. SRLG, 329-30).
Volitions are seen on Sellars' view then as being, on the one hand, here and now intentions and, on the other hand, as causal antecedents of actions. He is not unaware, of course, of the criticism that volitions have encountered in contemporary philosophy, and he spends considerable energy responding to these objections (cf. TA, 108). Again, we must pass over the detail of his response, but we can note that he thinks that many of these objections have been based on two misunderstandings of volitions: it has been thought by many critics, first, that volitions are themselves actions and, second, that they cause agents to act. Sellars, on the contrary, argues that volitions are not actions but rather mental acts in the classical sense; thus the infinite regress these critics see in the offing is avoided. Against the second claim, he argues that volitions can cause actions without causing the agent to perform the action. Thus, on his view, the deterministic implications of this view are avoided.¹⁴
If we can identify expressions of intention and volitions as the main ingredients in practical reasoning, the next step is to show how they are tied together in patterns of inference -- or pieces of practical reasoning. That certain intentions imply other intentions seems initially plausible. If I intend some action A' and if some other action on my part, A", is a necessary condition of my performance of A', then it would seem plausible to say that my intention to do A' implies an intention to do A".¹⁵ Or, put schematically:
(1) Shall_Jones (Jones will do A')
(2) (Jones will do A') implies (Jones will do A")
(3) Therefore, Shall_Jones (Jones will do A")
One difficulty with this argument immediately leaps to the eye, however. Many persons, through lack of will power or ignorance, notoriously do not intend what is implied by their intentions, just as many persons do not believe what is implied by their beliefs. This objection counts against Sellars' views, however, only if he construes the implication relation as holding between acts of intending. But he need not, and in fact does not, hold this. Just as we can distinguish between some person's act of believing and the content of his belief, we can also distinguish between some person's act of intending and the content of his intention (cf. SM, 182-83). That a person intends a certain action, does not imply, of course, that he will intend the implications of that intention. But even to say this presupposes that it makes sense to talk about the implications of what is intended by that person. As Sellars says, "An ideally rational being would intend the implications of his intentions just as he would believe the implications of his beliefs." (SM, 183.) If we recognize then that Sellars is claiming that implication relations hold between contents of intentions and not between acts of intending, the objection above does not tell against his account.
Sellars argues that there are only two rules that govern implications among Els:¹⁶

(R1) No EI can be derived from anything other than another EI.
(R2) All implications among EIs are warranted by the following schema:
'A' implies 'B' iff shall (A) implies shall (B)

The first rule parallels Poincare's rule that no imperative sentence can be derived from an indicative: the second rule suggests that there is no "special" logic of intentions but rather that implication relations among intentions are parasitic upon implication relation among the statements that express the content that is intended in an intention.
In order to see how these rules apply in an actual case of practical inference, we can examine Sellars' treatment of the inferential character of what Hare took to be a case of imperative inference. As noted above, Hare suggested the following argument as an instance of imperative inference:
(1) Take all the boxes to the station.
(2) This is one of the boxes.
(3) Therefore, take this to the station.
As we have seen, since Sellars rejects imperative inference altogether, he must reject this argument as an instance of imperative inference. But one cannot escape the feeling that there is something like inference going on here. Sellars suggests that we can account for that feeling by seeing how this argument can be treated not as an inference from an imperative plus an indicative to another imperative but as an inference from one El to another making use of certain "factual" beliefs. Thus, Jones might reason as follows:¹⁷

Shall_Jones (Jones will take all the boxes to the station)
This is one of the boxes.
(Jones will take all of the boxes to the station) implies (Jones will take this box to the station)
Shall_Jones (Jones will take all of the boxes to the station) implies Shall_Jones (Jones will take this box to the station)
Shall_Jones (Jones will take this box to the station)
In this piece of practical reasoning, Jones comes to see that his original intention to take all of the boxes to the station (which from the point of view of this piece of reasoning is underived) implies the derivative intention to take this box to the station. The steps in the argument are justified as follows: (1) is the intention that starts the whole process going; (2) is a belief of Jones; (3) follows from (2); (4) follows from (3) in accord with Sellars' second rule for deriving intentions: (5) follows from (1) and (4) by modus ponens.
Sellars suggests that all practical reasoning can be reconstructed on this model. One notices immediately that one strength of this model is that there are no inferences from indicatives to intentions (or from "facts" to "values"). Thus, this model will not encounter the difficulties raised by the so-called is[ought problem. Also, this structure allows practical reasoning to be really practical in that the Els that are the conclusions of such reasoning will grow (in the fullness of time and changes of mind, physical and psychological impediments aside) into volitions and thus action. But this practical force is not bought at the expense of reducing practical thoughts and expressions to nonconceptual ejaculations or interjections. Finally, since Els do not require an audience and are not actions, this account of practical reasoning does not encounter the difficulties that we have seen that an imperativist account succumbs to.
We next have to show how Sellars can develop from this view of practical reasoning his claim that moral judgments are essentially practical. If all instances of practical reasoning issue in Els and if moral reasoning is practical reasoning, then moral judgments must be construed as special forms of Els; and this is Sellars' final conclusion. Indeed, Sellars' positive views on the logic of moral judgments consist in proposing complications in the structure of simple Els in order to suit them as models for certain basic kinds of moral judgments.
In this regard there are two features of moral judgments that Sellars has to build into Els if they are to capture what we mean when we use moral judgments. First, he has to account for the universal character of moral judgments. If moral judgments are Els, that is, we must show how intentions can be universal in the way that we think of moral judgments as being universal. Second, Sellars must account for the intersubjectivity that we ordinarily take to characterize moral judgments. When I claim that Jones morally ought to do A and you claim that Jones morally ought to refrain from A, we ordinarily think that one of us must be wrong. That is, we think that moral judgments can clash in a way that would make it appropriate to say that the one contradicts the other. But Els seem to lack this element of intersubjectivity. While Els can certainly clash, their essential egocentricity (each person can express only his own intentions) would seem to insure that contradiction can occur among them only in a metaphorical sense. Thus, since Sellars fully agrees that ordinary moral discourse exhibits intersubiectivity, he must show how this feature can be accommodated within a view that construes moral judgments as Els.
In the next section, we will take up Sellars' treatment of universalizability, and in the section following we will turn to his views on intersubjectivity.
V
Although Sellars thinks that moral judgments are to be understood on the model of Els, it is clear that they are not just EIs. To judge the action of some person morally is not just a matter of expressing one's intentions with respect to that person's actions. For one thing, as virtually all moral philosophers have noticed, moral, judgments are implicitly universal. The point is often made by saying that when one assents to a particular moral judgment -- e.g., "Jones morally ought to join the resistance" -- one must be prepared to support it with a universal principle -- e.g., "Anyone who is in a situation relevantly similar to Jones ought to join the resistance." If the person in question is not prepared to make the principle truly universal (if, for example, he wants to remove himself from its scope), one would be justified in concluding that he is not expressing a moral judgment at all. We might capture the central point here in terms of moral approval or disapproval in the following principle: If P morally (dis)approves of q's doing A in S, P morally (dis)approves of anyone's doing an action of type A in a situation relevantly similar to S.
Needless to say, this formulation of the principle of universalizability is not unproblematic. Much remains to be said about how "relevantly similar circumstances" are to be defined, and the notion of an action-type has to be given a clear sense. Also, one has to avoid suggesting that the claim that particular moral judgments presuppose universal principles implies that anyone who makes a moral judgment must have a relevant moral principle already formulated and at hand. The point rather is that one must not bring forward a patently nonuniversal "principle" in support of one's moral judgment, and one must believe that there is a universal principle applying to the action in question even if one cannot, in the moment, formulate it (cf. IIO, 163).
Sellars agrees that moral judgments are universalizable, but he has approached this topic in a different way than have most recent ethical theorists. Instead of investigating universalizability as a logical property of moral judgments, Sellars focuses his attention on the notion of action on principle. That is, he is more interested in universalizability as it might be seen as characterizing the action and practical reasoning of a moral agent than as simply a logical feature of moral judgments.¹⁸
Sellars' account of action on principle is designed to handle two different sorts of difficulties that arise with it: first, he attempts to make explicit the manner in which the universal performance of some action is intended or desired in action on principle; second, he wants to account for the role that principles play in the reasoning that leads to action done on principle.¹⁹ We will look first at Sellars' account of the difficulties in these two areas and then turn to his positive account.
To act on principle is to be committed in some sense to everyone's doing an action of a certain sort. But one can be committed to the universal performance of an action for quite different reasons; Sellars suggests four such reasons (cf. IIO, 182-83). Consider someone wishing that everyone keep promises. He might wish this because (1) he thinks that the consequences of the joint action of everyone's keeping his promises would be valuable; (2) he thinks that the consequences of each person's keeping a promise would be valuable; (3) he thinks that the universal keeping of promises would be valuable for its own sake; or (4) he thinks that each instance of promise keeping is valuable for its own sake. Each of these reasons might justify one in intending (or wishing for) universal conformity with the moral rule against promise breaking. And also, of course, each of these would justify some person in thinking that he should conform to the moral rule prohibiting promise breaking.
The question arises, however, as to whether the actions of a person who acted in accord with any of these reasons would constitute action on principle. Sellars thinks there is good reason to doubt it in each case. When we say that a person acts on principle we imply, among other things, that he does an action for its own sake, i.e., that his motivation is of a nonconsequentialist sort. But this suggests that neither (1) nor (2) will do. Neither do we think, however, that a person who acts on principle is aiming at a merely universal accord with some principle for its own sake as one would in (3). As Sellars says, "action on principle is not silly as it would be if it were a matter of wishing that I keep this promise as a logically necessary condition of a world-long and world-wide keeping of promises." (IIO, 184.)
This leaves the fourth reason above, but it too seems not to measure up. We think that in action on principle a decision to do some particular action follows, in some sense, from a commitment to the universal performance of the action. But in the fourth reason above this mode of reasoning seems inappropriate. The fourth reason suggests rather that my commitment to the universal performance of an action follows from my commitment to the particular actions in accord with the policy.
The first problem for Sellars' account of action on principle is set then by the fact that it is difficult to specify the grounds on which the universal performance of an action is intended in action on principle. The second problem is suggested by his remark that "approving on principle is not just the same as being disposed to 'just like' each item which one comes to think of as being of a certain kind." (IIO, 188-89.) the thought behind this remark is that action on principle is more than mere consistency in acting and approving of the actions of others. Consider for example the approvings of three quite different sorts of men -- Jones, Smith, and Roberts.
Jones' approvings are purely random. He finds himself approving of one thing one day because it is F and another thing the next day because it is not-F. Needless to say, Jones will have great difficulty formulating policies or principles; and those of us who live with him will be incapable of predicting what he will approve of next or explaining why he approves of what he does turn out to approve of. Jones is a man with neither settled practical principles nor a firm character.
Smith, on the other hand, has approving with much more structure than Jones'. His approvings are regular and, like the darling of the precinct worker, he is a man whose vote can be counted on. We can, given sufficient familiarity with Smith's behavior and the ordinary canons of induction, formulate general principles with which his approvings conform and that we can in turn use both to predict and explain his approvings. Smith is a man who might seem to have both settled principles and firmness of character. Sellars invites us to consider, however, that a man like Smith might be the sort of person who, though regular in his approvings and his behavior in accord with those approvings, does not, in approving, act out of regard for any principles of action at all. Smith, that is, might be the sort of man who buys red cars whenever he buys cars, paints his house red whenever he paints his house, and buys strawberry ice cream whenever he buys ice cream, but who never has the thought that red cars, red houses, or red ice cream are good because they are red. Smith's approvings, while regular, predictable, and explainable, need not be reasoned.
We finally come to Roberts whose approvings are the furthest advanced on the road to self-consciousness. His approvings are regular, predictable, and explainable and in addition are accompanied by thoughts on his part of the "If anything is F, would that it be defended" type. To say that his approvings are reasoned, then, is to say that they are accompanied by thoughts of certain sorts that partly explain his approving. While Smith's approvings can be explained on merely associationist principles, Robert's approvings require some reference to the form of reasoning from which they issue in order to explain them.
It seems clear that only Roberts of the three men considered here has the kind of approvings that we associate with action on principle. Any account of acting on principle, then, must take account of not only the peculiar sense in which the universal performance of an action is intended (or wished for) in such action but also the reasoned character of the decision to act on that principle. To put the same point the other way around: not just any wish for the universal performance of an action nor just any kind of consistency in acting or approving on the part of some person will justify the claim that he is acting on principle.
With some understanding of the complexities that Sellars thinks any adequate account of action on principle must confront, let us turn to his own account. Sellars' basic view is that to act on principle is to be disposed to have certain underivative intentions. If, for example, Jones performs an action type A (because he is in circumstances C) on principle, then he must be disposed to reason about every other person when that person is likely to be in C.

X is (or will be) in C.
Shall_Jones (x will do A, if he is in C)
Therefore, Shall_Jones (X will do A)
Notice that in accounting for the universal character of Jones' intention, we do not characterize the practical reasoning involved as originating in an intention with a universal content. We do not suggest, that is, that the practical reasoning that issues in action on principle must begin with intentions of the following form:
Shall_Jones (X) (if X is in C, then X will do A)
To think of such universal intentions as occupying the place of premises in the practical thinking leading up to action on principle would be to confuse, Sellars argues, the premises of such reasoning with the rules that govern the reasoning (cf. IIO, 185). Sellars' basic insight here is that the universality implicit in action on principle is not to be explained with reference to the universal content of the intention involved but rather with reference to the disposition to reason in a certain way in all cases.²⁰
If Sellars is right, then we can see how the first complexity in action on principle is to be handled. It is simply not true that in action on principle we directly intend or wish for the universal performance of an action; rather we are disposed to have underivative intentions of the form
Shall (If X is in C, then X will do A)
whenever we think of someone being in C. Sellars invokes the notion of an axiom schema to further develop this point.
A more plausible approach, which has the advantage of interpreting the universality of moral principles in terms of each case rather than the totality of cases, draws on the concept of an axiom schema. According to it, instead of saying that
Shall (X)(X doing A, if X is a person in C)
is an axiom, we should rather say that every statement derived from the schema
Shall (X doing A, if X is a person in C)
by replacing X with the name of a person is an axiom or first principle of moral reasoning. (IIO, 192.)
We explain the universality of moral judgments, then, by characterizing their origin in practical thinking as being of a certain sort. Particular moral judgments are Els that issue from practical thinking that would, in relevantly similar situations, produce relevantly similar Els.
We can also see, I think, how this characterization of action on principle accommodates the reasoned character of such action. Action that issues from an El that is in turn derived from an axiom schema of the form Sellars proposes cannot be explained on merely associationist principles. This possibility is excluded, since in these Els the condition (the if-clause) is brought within the scope of the shall-operator. We are not saying, therefore, that the person in question is disposed to have certain intentions whenever he notices the relevant features of some person's situation. Rather, the underivative intentions he has are intentions that each person do the appropriate action because he is in a certain situation. Put another way, for Jones to have intentions that conform to Sellars' axiom schema is for Jones to intend that each person do A because that person is in C; it is not merely for Jones to be disposed to have the intention that a person do A whenever Jones notices that the appropriate person is in C.
There are many questions that could be raised about this account of action on principle. Indeed, Sellars has admitted that his most lengthy discussion of the topic merely "scratches the surface." (TA, 138.) Given the purposes in this essay, however, I must leave the topic here. It should at least be clear, nevertheless, that Sellars can give an account of the universalizability that characterizes moral judgments when these judgments are construed as having the force of Els. It is especially important to see that Sellars does not account for this universalizability by interpreting moral principles as having the force of Els with universal contents. This interpretation, which might have seemed the obvious one for him to take, is certainly implausible, and its implausibility has surely deterred others from modeling moral judgments on Els.
It is also worth remarking that Sellars' general approach to the topic of universalizability would seem to be worth the attention of others who have wrestled with this topic. His general thought that universalizability should be investigated as having its roots in the forms of practical reasoning associated with moral judgments seems sound. When universalizability is treated, as it usually is, as "merely" a logical feature of moral judgments, one is inclined to overlook the complexities of it to which Sellars' approach calls our attention.
VI
In his Ethics, G. E. Moore developed an argument against subjectivistic accounts of moral judgments that has continued to be very influential.²¹ Moore argued that any subjectivistic account of moral judgments is bound to be inadequate because it cannot account for the sense in which the moral judgments of two different persons can be real logical contradictories. His argument took the following form:

According to subjectivist ethical theories, if one person asserts that P ought to do A and another person asserts that it is not the case that P ought to do A, they are not contradicting each other, since what the first person is asserting is different from what the second person is denying.
But it is clear that in ordinary moral discourse we suppose that a conflict of this sort does involve contradiction in that both persons cannot be right.
Therefore, all subjectivist ethical theories must be mistaken as accounts of what we mean when we use moral judgments.
Some have argued that the second premise of this argument is question-begging in that it prematurely closes the question that is the whole point of subjectivist theories to open. Indeed subjectivists have frequently argued that all Moore's argument really does is force them to provide some account within their theories of why we feel that moral judgments can stand in a relation of contradiction.²²
It is an important feature of Sellars' approach to ethical theory, however, that he takes Moore's argument very seriously indeed. Sellars claims, in fact, that the most formidable objection to his attempt to model moral judgments on Els derives from our feeling that moral judgments are intersubjective in a way that Els cannot hope to be. He takes it as a criterion of success for any account of moral judgments that the account be able to accommodate the intersubjectivity of such judgments. The two features of Els that seem to render them inappropriate on this score are (1) the absence of an external negation in Els and (2) the conceptual impossibility of two persons having the same intention. We will explore each of these features of Els briefly and then turn to Sellars' attempt to resolve the difficulty.
That Els lack an external negation becomes clear if we reflect on the possible ways in which negation can enter into the structure of first person ought-judgments and then Els.²³ With ought-judgments we can distinguish the following four forms:
(1) P ought to A.
(2) P ought not to A.
(3) It is not the case that P ought to A.
(4) It is not the case that P ought not to A.
Each of these forms is basic in the sense that none can be reduced to any of the others. When we turn to Els, however, we find that only two forms are legitimate:
(5) Shall (P will A)
(6) Shall (P will not A)
The forms that are lacking are those that would be represented by
(7*) Not-Shall (P will A)
(8*) Not-Shall (P will not A)
Sellars has a lengthy argument to show that these forms are inadmissible, but it seems to me that the point is obvious enough (cf. IIO, 198). The only danger is that someone might suppose that certain biographical statements capture these forms. Thus, we can perfectly well say all of the following:
(9) Jones intends to do A.
(10) Jones intends not to do A.
(11) Jones does not intend to do A.
(12) Jones does not intend not to do A.
But it should be clear that (11) and (12) do not correspond to intentions that Jones has and that he could express in Els. They report, rather, the absence of intentions on Jones' part with respect to his either doing A or not doing A. The legitimacy of (11) and (12), then, should not lead us to suppose that a well-formed El could have an external negation.
The paucity of legitimate forms for Els leads to difficulties because in the absence of external negation it is not clear how Els can stand in the relation of logical contradiction in the way that Moore and Sellars think ought-judgments clearly can. Jones and Smith can straightforwardly contradict each other if Jones claims that Roberts ought to do A and Smith claims that it is not the case that Roberts ought to do A. The two ought judgments,
(13) Roberts ought to do A, and
(14) it is not the case that Roberts ought to do A,
stand in a relation such that they cannot both be true. But we seem unable to reproduce this kind of clash with respect to Els. The intentions of Jones and Smith clearly might diverge as represented by the following two statements:
(15) Shall_Jones (Roberts will do A)
(16) Shall_Smith (Roberts will not do A)
The difficulty here, however, is that while Smith and Jones are clearly involved in some sort of conflict, they do not seem to be contradicting each other. Sellars has suggested that the most we can say here is that a disagreement in attitude is involved (cf. SM, 189).
This point is tied to the second difficulty with Els -- the impossibility of two persons having the same intention (cf. SM, 188). When we introduced the notion of an El in IV, we noted that Els always express the intentions of the speaker -- that they are in this sense abidingly egocentric. The intentions that I express are always my intentions, although, of course, I can report your intentions, describe them, raise questions about them, and so forth. But if intentions are in this strong sense egocentric, it would seem impossible for two persons ever to have the same intention. We must be careful here because the surface grammatical form of expressions of intentions suggests that we frequently do have the same intention. Thus, both Smith and Jones might intend that Roberts do A; indeed it might be true to say that all the members of some group intend that some action be done or some state of affairs be brought about, e.g., all of the citizens of a city might intend that the sewer system be repaired. But, although it appears in these cases that different persons have the same intention, it turns out that this cannot be true in the same strong sense in which two people, for example, can have the same belief.
This becomes clear if one recalls the way in which Els sustain their connection with action. Though two persons might intend that some state of affairs be brought about, it is really the case that each of them intends that he do what he can to bring about that state of affairs. If Jones intends that the sewer system be repaired, the cash value of this intention is that he intends that he, Jones, take certain steps to bring it about that the sewer system be repaired, and likewise for Smith.
The same ought-judgments clearly can be expressed by different persons, however, and this is a partial explanation of why ought-judgments have an external negation. If both Jones and Smith believe that Roberts morally ought to do A, they are, in a strong sense, believing the same thing; and if Jones asserts that Roberts ought to do A while Smith asserts that it is not the case that Roberts ought to do A, they are, in a straightforward sense, contradicting each other.
These difficulties are important ones for Sellars because they would seem to suggest that any attempt to treat ought-judgments as special forms of Els is bound to founder on the lack of intersubjectivity of the latter. His response is not, however, to abandon the claim that moral judgments are Els but rather to suggest that they are Els of a special kind. His problem is to construct an El that will accommodate the intersubjectivity of moral judgments while retaining the features of "simple" Els that attracted him to them in the first place.
The special kind of El that Sellars introduces is what we might call the We-EI.²⁴ He suggests that there are EIs that express the intentions of a group but are asserted (or expressed) by members of a group. If there are such intentions, then we could express Smith's and Jones' agreement about what Roberts is to do by statements of the following form:
(17) Jones: Shall_we (Roberts will do A)
(18) Smith: Shall_we (Roberts will do A)
This formulation indicates how the intentions in question are the same and we can also see how, in virtue of this sameness, real contradiction between we-intentions can arise.
It is remarkable, I think, that Sellars spends so little time defending or explaining his notion of we-intentions, since they play a crucial role in his overall account of moral judgments. He does tell us some of the things that We-EIs are not, and he also makes a number of provocative remarks about how they are related to a special form of consciousness. It is important to consider his remarks here and develop some of the insights they contain as well as some of the difficulties with them. I am going to suppose that Sellars is right in claiming that we-intentions as he introduces them would handle the intersubjectivity characteristic of moral judgments; the questions I want to raise are rather questions about what we-intentions really are.
One way of approaching this topic is to consider one simple account of what we-intentions are. This account, as we shall see, will not do justice to Sellars' notion of we-intentions, but its failure is illustrative of some important points. This account would suggest that we-intentions are really compounds of a sort that involve as components both I-intentions and certain biographical statements about other members of one's group. Consider, for example, a group made up of three men -- Jones, Smith, and Roberts. When Jones we-intends that P do A, we would characterize this intention in the following way:
Shall_Jones (P will do A) and Smith and Roberts also intend that P do A.
The point of this, of course, is that while Jones cannot do the intending for Smith and Roberts, he can report their intentions. Given that this is so, we might think that we-intending is just ordinary I-intending that is accompanied by the belief that everyone else in one's group intends the same thing. To avoid certain obvious objections, this characterization could be fixed up by giving the biographical statements a suitably counterfactual character. So Jones, rather than asserting that Smith and Roberts in fact agree with him, might assert that they would agree with him if they had at their disposal the facts that he has at his disposal, were free of emotional disturbance, and so forth. We might further modify it so that Jones does not assert that all the members of the group would share his intention but that most would or that the normal ones would, and so forth. In this way, this view of we-intending would take on some of the formal features of so-called ideal-observer theories in ethics.
But however we alter this basic model for we-intending, it will fail to capture what Sellars has in mind. It fails for two reasons. First, Sellars simply says that I-intending and we-intending are different forms of intending (cf. IIO, 203). He could not accept a view that would make we-intending simply a matter of I-intending plus something extra. The intendings in the two cases must be different. This is also made clear in Sellars' discussion of moral weakness, where he construes weakness of will as a matter of the conflict between I-intending and we-intending (cf. KBDW, 17-19). But if we-intending fits the model we have suggested above, there simply could not be conflict between the I-intending of some person and his we-intending. This conflict would be impossible because we-intending would just be I-intending plus certain beliefs.
But a second reason why this view will not do is that Sellars claims that we-intending involves a radical move away from ordinary I-intending. In some places he says that we-intending involves a special "form of consciousness"; in others, that it involves a special "form of life." He also suggests that we-intending results from "internalizing the concept of a group." (Cf. IIO, 203-5.) These ways of characterizing we-intending should make it clear, if it were not already, that Sellars sees a radical difference between I-intending and we-intending.
But how are we to construe this talk about a new form of consciousness? One way of approaching the topic, is to reflect on what is involved, in coming to see oneself as a member of a group. We might consider first an amateur theatrical group that is trying to put on its spring production. We can readily understand how the members of this group might come to have the same intentions with respect to the success of their production. But to put the point this way is to suggest that group participation is a matter of the coincidence of a number of I-intentions, and again this does not seem sufficient to capture what Sellars is after.
Consider instead a local seamstress who is approached by members of the group and requested to make certain costumes that they need for their production. And further suppose that in the beginning of her association with this group, the seamstress is not aware that she is making costumes for a dramatic production. She approaches her work as just one more job, and she is guided in it by the norms of money-making, craftsmanship, or whatever, which usually guide her. Now let us imagine that the seamstress becomes aware of the role that her work is playing in the larger goals of the dramatic group and even that she becomes, as we might say, interested in the success of the production. Her enthusiasm might be such that she is inducted into the group as a member and continues to do the same work without compensation. It seems obvious that if events transpired in this way, we could plausibly say that the seamstress comes to see her relationship to the members of the dramatic group in a new way. Indeed they would cease being "them" and become "us.," More important, the seamstress could come to see the standards applicable to her particular work in a new way. Where before she was just "doing a job" where her work was subject to her self-criticism in the same way that any other "job" would be, she now sees her work as playing a role in the larger enterprise of the dramatic group. In particular, the success of the particular costumes she is making cannot be determined apart from some role they may play in the overall success of the production.
Is this change that the seamstress undergoes what Sellars is getting at when he talks about a new form of consciousness? I am inclined to think so. In coming to see her particular project -- in this case, the making of costumes -- as being constrained in certain ways by the overall aims of the group, I think she is at least coming close to having what Sellars calls we-intentions.
Though this example may help us to capture the flavor of Sellars' we-intentions, some difficulties remain. One of these is the problem of construing the relation between I-intentions and we-intentions. In terms of our example, we might wonder how the we-intentions that the seamstress acquires in becoming one of "us" are related to her I-intentions. Sellars says that "there is a particularly close logical connection" between the two, but he denies that the relation is one of entailment. In particular, he claims that "we intend that x do A" need, not entail that "each of us intends that x do A." (IIO, 203.) And it is for this reason that he wants to say that we-intending is not merely the logical sum of the I-intendings of those who make up the group. Sellars also claims (being true to his nominalistic predilections) that we-intendings are not the intendings of any sort of group mind. But what, we are inclined to ask, are the existence conditions of a we-intention? Under what conditions is it true of the members of a group that they we-intend some action or state of affairs? The answer we are tempted to give here is that a group we-intends some object just in case the group has a shared intention for this object. But this answer will not do as Sellars makes clear in the following passage:
The group has shared intentions by virtue of the fact that its members intend in the mode "shall_w". But that the members intend in this mode does not guarantee that in "point of fact there are shared intentions. Intending in the mode "shall_w" is a "form of life", a conceptual framework within which moral discourse exists and without which it is impossible. Yet the actual existence of shared universal intentions is no more an antecedent condition of participation in moral discourse than actual agreement on matters of fact is an antecedent condition of participating in factual discourse. In each case the forms of discourse set this agreement as a task. To abandon the idea that disagreement on moral matters is even in principle capable of resolution is not to retreat to a moral solipsism; it is to abandon the moral framework itself, and to retreat to the language or "form of life" of purely personal intention. (IIO, 205.)
In this passage Sellars is concerned to distinguish the following three items: (1) the presence of shared intentions within a group, (2) the presence of intending in the mode "shall_w" within a group, and (3) the presence of moral discourse within a group. With respect to these items, he further suggests that (2) and (3) always go together -- indeed that they are two aspects of one and the same phenomenon. To engage in moral discourse is to intend in the mode "shall_w." But (1) is not a necessary condition of either (2) or (3). Although Sellars says that "the group has shared intentions by virtue of the fact that its members intend in the mode shall_w," he goes on to add that "that the members intend in this mode does not guarantee that in point of fact there are shared intentions." The achievement of shared intentions by a group is rather a "task'' that is set by intending in the mode shall_w; it is not an "antecedent condition" of such intentions.
The upshot of this passage then is that we cannot look to the presence of shared intentions as the existence conditions of we-intending. Indeed, we-intending seems just to consist in the presence of certain forms of discourse, in particular the presence of moral discourse. But this account of we-intending seems to cause difficulties for Sellars in that he introduced we-intentions to explain the intersubjectivity that is characteristic of moral judgments; but when we press for an account of we-intending, the response is that we-intending is what goes on when we engage in moral discourse or reflection. This same circle is involved when, in the passage above, Sellars characterizes we-intending as a form of life or a conceptual framework "within which moral discourse exists and without which it is impossible." If we ask what the form of life or conceptual framework here referred to is, the only possible answer would seem to be one which would characterize it in terms of moral thought and talk.
The point I am making, of course, is that if the notion of we-intending is to be of use in elucidating the logical force of moral judgments, this mode of intending has to be characterizable independently of any reference to our linguistic behavior in using moral judgments. If I understand him, however, Sellars has not shown how such an independent characterization is possible.
One more related question needs to be raised. You will recall that Sellars invoked intentions as the fundamental conceptual items on which moral judgments are to be modeled primarily in order to be able to account for the practical force of moral judgments. It is because Els are logically tied to action, through the mediation of volitions, that they are suited to this role. But though I-intentions have been constructed in order to have this practical force, it is not at all clear that we-intentions will have this force or at least that they will have it in the same way. I-intentions, as we have seen, in the fullness of time become volitions, and volitions issue in action. But what about we-intentions? Do we-intentions stand in the same relation to volitions as do I-intentions? I know of no place in Sellars' writings where he explicitly takes up this question, but it is clearly crucial to his overall theory. Whatever may be the practical force which accompanies I-intentions, it is irrelevant to the practical force that moral judgments bear; moral judgments are, after all, not I-intentions but rather we-intentions. These questions would not arise were Sellars to characterize we-intentions in such a way that they entail I-intentions; but as we have seen above, he rejects this view. Indeed, as I have pointed out, it is at the heart of his treatment of the problem of moral weakness, that I-intentions and we-intentions must be able to conflict.
One wishes that Sellars had introduced the notion of we-intentions more precisely by locating it within his general account of practical reasoning. In doing this he could have said more about (1) how they arise, (2) how they are related to I-intentions, and (3) how they are related to volitions. Until these matters are settled, it will be difficult to make any final assessment of Sellars' account of the intersubjectivity of moral judgments; indeed, it will be difficult even to know how to understand this account.
VII
In our discussion of Sellars' account of moral judgments so far we have seen how he has attempted to flesh out Prichard's insights by showing that moral discourse is action-guiding, intersubjective, and unique. By construing moral judgments as intersubjective Els, he thinks he has fulfilled the first two goals. And the uniqueness of moral discourse is a function of the uniqueness of the we-intentions on which Sellars models them. There is one rather large and troublesome question, however, left over for his account. This last difficulty is set by his agreement with Kant that moral judgments are implicitly "reasonable" in a way strongly analogous to the reasonableness of true beliefs.
Sellars raises this question in a number of different ways in different places. Thus, he talks in one place about whether normative ethics is a rational discipline, by which he means "a field of inquiry in which good reasons can be offered for answers to questions belonging to that field." (SE, 194.) The context of this discussion makes it clear that by "good reasons" he means reasons as good as those that are given in the sciences and mathematics. In another place, Sellars asks how we are to understand the difference between a merely inherited set of categories for evaluating actions and a fully reasoned and reflective morality. The suggestion here is that any account of morality that could not accommodate its reasoned character would be inadequate (cf. IIO, 205-6).
The question Sellars here raises is the classical question of the justification of moral judgments. One might suppose that he has already answered this question by tying moral discourse to intersubjective we-intentions; but this would be a mistake. If Sellars is successful in his identification of moral judgments with we-intentions, he will certainly have accounted for the intersubjectivity of moral discourse; but if his account were to end here, he would make the truth (or assertibility) of moral judgments relative to the we-intentions of a group. In raising the question of the reasonableness of moral judgments, however, he wants to raise questions of the adequacy of particular we-intentions. In the discussion below, it will simplify things if we set out Sellars' views on justification in ethics without for the most part bringing in his interpretation of moral judgments as intersubjective Els. It should be clear from what has gone before how this discussion could be translated into the we-intention mode.
Sellars' general approach to the justification of moral judgments is an attempt to reconcile his basically Kantian or deontological view of moral motivation with a utilitarian or teleological view of moral justification. He maintains, that is, that particular moral judgments are ultimately justified by showing how they are involved in the pursuit of some overarching end or goal. He realizes of course that he must square this strategy with his insistence that moral action is action on principle where the action performed is performed for its own sake. While many moral philosophers have regarded such a compromise as impossible to achieve, Sellars proposes to bring these two features of moral judgments together by distinguishing two different levels at which intentions can operate in action.²⁵
Suppose, for example, that someone has a basic intention to bring about some state of affairs S (e.g., a satisfying life for himself or a maximization of the general welfare). And further suppose that he discovers that a necessary condition of realizing S is that everyone act in certain specifiable ways in certain specifiable situations. He may discover, further, that the best way to insure that everyone acts in these ways is for everyone to acquire dispositions to act in these ways in these circumstances without regard to the consequences of particular actions. Thus, it would be reasonable for him to set about promoting the development of such dispositions in himself and others. What we would have in such a case, Sellars claims, is an entry of intentions into action at two different levels. At the first level is the intention to bring about S; but this intention can, in the way we have seen, "lead to" intentions of the "Shall (X will do A, if X is in C)" sort, which we have seen in section V are characteristic of action on principle (cf. IIO, 207).
This general framework for justification would seem to allow Sellars to reconcile the Kantian strain in his thought with a teleological justificatory base. Particular moral actions would be actions on principle in that they would spring from dispositions to act in certain ways in certain situations; the practical reasoning associated with such actions would be of the sort we explored in section V. But the decision to acquire and nurture in oneself and others these dispositions to act, and hence the actions themselves, would be ultimately justified with reference to one's goal in acquiring such dispositions. The motivational structure of moral action will then be true to Kant's strictures, while the ultimate justification for such action will have a teleological form.
In order to fill in this framework, Sellars must identify some goal or end that will support the superstructure of moral judgment and moral action as we know it. Such a goal will have to satisfy two conditions. First, it must be such that moral judgments with roughly the same content as those which most persons accept can be derived from it; it must, that is, give rise to particular judgments that are recognizably moral. But, secondly, and more importantly, it must be a goal that is intrinsically reasonable to pursue. Since the point of the enterprise of justification is to show that moral judgments are reasonable, it will not do simply to show that they are derivable in the manner suggested from some goal. This would show at most that they have a certain relative reasonableness; but Sellars is determined to show that moral judgments are reasonable in the stronger sense of Kant's categorical imperative. And in order to do this he must show that they are derivable from some goal that is intrinsically reasonable to pursue.
Before coming to Sellars' own account of the ultimate end of morality, it will be instructive to consider two alternative views that, though they ultimately prove defective, point the way to Sellars' view, the first account of the ultimate object of morality is one associated with what Sellars calls the "art of living'' conception of morality (cf. SM 200-204). Sellars connects this view to a central strand of ethical thought in classical Greek philosophy. On this view, moral judgments tell us what we must do if we are to achieve a satisfying life for ourselves; actions are morally required in the sense that they are required for a satisfying life. Hence the moral ought-judgment
(1) Jones morally ought to A
is taken as (roughly) equivalent to
(2) If Jones wants to lead a satisfying life, then he ought to A.

There are obvious objections, however, to this view. In the first place, it seems clear that the actions that such a view would enjoin would not be the same actions that would be required by morality. Although, as Sellars points out, Plato labored mightily to defend the coincidence of the demands of the art of living and of morality, almost everyone would agree that he failed. But even if he had succeeded in showing that the art of living was materially equivalent to morality, he would not have overcome our feeling that there is a conceptual distinction here (cf. SM, 204). The binding force we attribute to moral judgments seems different in kind from the binding force attributable to action guides derived from such an art of living. If nothing else, the tension we frequently feel between the demands of morality and the demands of prudence would support our feeling that we have two distinct things here.
One might suppose that the difficulties that the art-of-living conception of morality encounter could be avoided if we substitute for this Platonic notion the more all-encompassing model of benevolence. Thus, a moral judgment would tell me not what is required for me to lead a satisfying life but rather what is required for me to promote the general welfare. The moral ought-judgment
(3) Jones morally ought to A
is, on this view, taken as roughly equivalent to
(4) If Jones wants to promote the general welfare, he ought to do A.

This view certainly appears more plausible as an account of the moral 'ought' than does the art-of-living conception. In particular, the actions that would be enjoined by a commitment to benevolence seem to be materially equivalent to those required by morality. But again there are difficulties. For one thing, Sellars thinks that we should balk at a conception of morality that rests the binding force of moral judgments on what people want (cf. IIO, 206). Like Kant, he wants to keep separate the realms of the moral and the "merely pathological." A more serious objection, moreover, is that this view allows a certain amount of slippage between one's commitment to the general welfare and one's decision to perform particular actions that will promote the general welfare. To make this clear, we must look a bit more closely at the point of view of benevolence.
One who takes the point of view of benevolence constructs principles of action by determining which principles, if observed by everyone, would promote the general welfare. He would recognize certain empirical generalizations of the form

If the general welfare is to be promoted, everyone must do actions of type A in situations of type S.
He would then strive to develop dispositions in himself and in others to act in accord with this principle. The difficulty here is that it would be silly to strive to promote absolute unanimity in the commitment of persons to these principles. We know that it is impossible, and probably unnecessary, to bring everybody's actions in line with such principles. What is important is that most people act in accord with these principles. But if we need not strive for universal conformity, does not each person have reason to slip through the net of morality in virtue of the fact that only most people need to conform to the relevant principles?²⁶
This query is as old (at least) as Glaucon and Adeimantus, and Sellars thinks it is unanswerable. If benevolence is taken as the moral point of view, then no particular agent has a compelling reason for bringing his actions in line with moral principles. As Sellars says, "The bridge between benevolence and the life of principle requires a sense of 'benevolence' which logically precludes the above compromise." (IIO, 209-10.)
But if benevolence fails as a model for the moral point of view, Sellars thinks nevertheless that the general welfare is the end in terms of which moral judgments are ultimately justified. In order to make good the claim of the general welfare to play this role, however, we must first see how the two difficulties with the point of view of benevolence can be avoided. These two difficulties were: (1) benevolence is based on merely wanting the general welfare to be promoted, while the ultimate end of morality must be something that is intrinsically reasonable to pursue; and (2) one could avoid the requirement to act in accord with principles of action derived from this end, since all that is really required is that most people observe these principles.
Sellars thinks that these difficulties can be avoided if we bring in, one more time, the notion of we-consciousness. The sense of "benevolence" that is required is not the impartial love of everybody,
which is, as Kant saw, the espousal of a principle of conduct (roughly that one ought to help other people realize their ends), but the simple recognition of people generally as we. (IIO, 210.)
Here again, as in his account of we-intentions, Sellars spends remarkably little time elucidating a notion that he introduces to play a crucial role in his overall argument. He does suggest, however, how this notion of we-consciousness will overcome the difficulties in treating the general welfare as the ultimate end of morality.
Sellars' explanation of how the commitment to promote the general welfare can be intrinsically reasonable is summed up in the following passage:
It is a conceptual fact that people constitute a community, a we, by virtue of thinking of each other as one of us, and by willing the common good not under the species of benevolence -- but by willing it as one of us, or from a moral point of view. (SM, 222.)
His thought here would seem to be that in the version of benevolence we have explained above the interests of others are something we can take an interest in only by wanting to promote these interests. And whether or not one has the appropriate want will depend on the contingent features of one's affective life. But when one comes to see others as "one of us," there is no longer a gap to be bridged by a want. Indeed, one's commitment to promote the interests of one's fellows would seem to be a necessary feature of seeing them as one's fellows. Insofar as one sees other persons as "one of us," therefore, one necessarily aims at promoting their welfare. And to say this, Sellars seems to think, is just to say that the intention to promote the general welfare is intrinsically reasonable.
Sellars responds to the second objection in this passage:
If we replace "most people but not I" by "we but not I" in the formulation of the objection, we move from consistency to incoherence. There is no logical place for a compromise between benevolence and self-love, where "benevolence" is understood as the consciousness of oneself and one's fellow men as we. (IIO, 210.)
The objection alleged that any person could escape the requirements of the relevant principle of action, since at most it would be required that most people act in accord with these principles in order to promote the general welfare. Sellars suggests in this passage that when one sees the general welfare as the welfare of us, the logical gap that allowed one to promote conformity with moral principles on the part of most people while making an exception of myself would be closed. One would no longer have the logical space for discriminating between our welfare and my welfare, since one's conception of one's own welfare would be inextricably bound up with one's conception of the welfare of the group.
Sellars' invocation of the notion of we-consciousness in order to respond to these objections to making the general welfare the ultimate end of morality requires more examination than I can give it here. My own feeling is that his account is essentially incomplete in that he has not shown with sufficient clarity how his conception of we-consciousness is to be distinguished from "ordinary" benevolence. Nor does his identification of we-consciousness with "what Royce called Loyalty" and with "what Christians call love of Neighbor (caritas)" help much (cf. IIO, 212). In his treatment of the justification of moral judgments, as in his account of their intersubjectivity, Sellars shows a profound grasp of the difficulties to be met; in both cases, however, his final solutions call for further clarification on his part.
While I have pointed out a certain incompleteness in Sellars' account of the intersubjectivity and the reasonableness of moral judgments, I have done nothing to show that Sellars is not on the ''right track" in both of these areas. One has the impression indeed that Sellars invokes the notions of we-intending and we-consciousness in order merely to gesture in the direction where he thinks the ultimate solutions are to be found. In pointing to the essential "groupiness" of moral discourse, he is, of course, picking up on a classical theme in moral philosophy, a theme that has been carried forward recently in an intriguing fashion by Thomas Nagel.²⁷ One can only hope that Sellars will further develop these notions in his future work.

NOTES
1 In order to work within this model, of course, one must believe that one has some grasp of the distinction between the "merely factual" and the moral. I, at least, do not have any such grasp, and it is a measure of the distance we have moved away from the problematic of CM in recent years that many other moral philosophers would share my skepticism.
2 It would be difficult, however, to point to any "naturalist" who actually held this position. One of the recurrent themes of CM has been the search for and positive identification of living, breathing naturalists. Virtually every classical moral philosophers was at one time or another accused of being a naturalist, but the evidence was always inconclusive. Among recent philosophers, Ralph Barton Perry was the favorite candidate throughout the 1930s and 1940s, while Phillipa Foot has assumed this role in the 1950s and 1960s. Again there seems to me good reason for rejecting both of these characterizations, especially in the case of Mrs. Foot. In any event, whether there were ever any moral philosophers who espoused naturalism as defined within CM is irrelevant, since naturalism as so defined did play a significant role within CM in the minds of those who opposed it. Whether one thinks that what was being opposed was an "ideal type" or a straw man will depend on ones more general attitude toward the presuppositions of CM.
3 The clearest (and crudest) example of this argument is found in Charles Stevenson, "The Emotive Meaning of Ethical Terms" Mind (1937), where he argues that a naturalistic account of the meaning of ethical terms cannot do justice to their "magnetism."
4 See Prichard's "Does Moral Philosophy Rest on a Mistake?" and "Moral Obligations,' both found in Moral Obligation (Oxford: Oxford University Press, 1949).
5 See Sellars' characterization of Ross' view in IIO, 160.
6 Sellars raises objections to emotivism in a number of places, but his most sustained attack is found in SE, 405-8.
7 R. M. Hare, The Language of Morals (Oxford: Oxford University Press, 1952), p. 13.
8 Ibid., p. 20.
9 Ibid., p. 27.
10 Sellars develops these arguments in IIO, 169-73.
11 Although Sellars claims in IIO, 170, that 'tellings-to' require an audience, he does hold that I can tell myself to do something. He says in TA, 107, "While we can address imperatives to ourselves in inner speech -- tell oneself to do something -- the thought expressed by:
I shall drive this nail
should not be construed as though it would be more perspicuously expressed by:
Myself, drive this nail.
In SM, 187, he says, "We can, and do on occasion, tell ourselves to do something, but deciding what to do is no more telling ourselves what to do than deciding what is the case is telling ourselves what is the case."
12 Sellars' clearest and most concise discussion of practical reason is found in TA. Important discussions are also found, however, in IIO and SM, chap. 6.
13 As Sellars points out, to set out the details of these rules would be extremely complicated. We would have to include rules that allow us to locate ourselves temporally and spatially in a world in which temporal location at least never remains the same from moment to moment.
14 For further discussion of these issues see Gary Gutting's paper in this volume on Sellar's treatment of the free-will problem.
15 As we noted above, Sellars' account of practical reason depends heavily on his views about material rules of inference. A discussion of the central role these rules play in Sellars' account of scientific reasoning can be found in G. Gutting's paper on Sellar's philosophy of science in this volume. For our purposes, we need only note that when Sellars talks about one intention implying another he includes cases in which the action intended in the first intention is a causally necessary condition of the action intended in the second. See SRLG, 330ff.; "Particulars", 292ff.; SM, 180-81.
16 Since (R1) would appear to be derivable from (R2), one might prefer to speak here of only one rule, and in some places (see SM, 179) Sellars seems to accept this view. In TA, 110-11, Sellars speaks of their being two rules, however, and we will follow his view as expressed there. It should be clear that nothing of importance hangs on the principle of counting we use here.
17 For Sellars' more elaborate account of the correct construal of this argument, see IIO, 173ff.
18 Hare's view, as I understand it, treats universalizability as simply a logical feature of moral judgments.
19 Sellars' most extensive account of action-on-principle is found in IIO, 180ff.
20 A more obvious objection to allowing universal intentions to play the role of premises in moral reasoning is that, as Sellars says, "it would be silly to espouse such resolutives knowing full well that in a great many cases people who are in C simply have not done, nor will they do A." IIO, 192.
21 G. E. Moore, Ethics (New York: Oxford University Press, 1965), p. 33ff.
22 For example, Charles Stevenson in "Moore's Argument Against Certain Forms of Ethical Naturalism," Facts and Values (New Haven: Yale University Press, 1963), pp. 117-37.
23 Sellars' most extensive discussion of the absence of external negation in Els is found in IIO, 196ff. Also, see SM, 185ff.
24 For Sellars' discussion of we-intentions, see IIO, 200-205; SM, 215-18; SE, 41-11; PSIM, 38-40.
25 One finds what might initially seem different accounts of moral justification in SE, IIO, and SM. In SE, Sellars emphasizes the distinction between vindication and validation, which also plays a large role in his treatment of the problem of induction (see G. Gutting's paper on Sellar's philosophy of science in this volume). In SM, the account of moral justification grows out of a lengthy and subtle analysis of the distinction between hypothetical and categorical imperatives. I have chosen here to follow the account in IIO. Though not as complete, nor as subtle, as the later SM account, it seems to me to contain the essentials of his view set out in the most straightforward manner. In doing this, I neglect entirely Sellars' insightful discussion of hypothetical imperatives. My excuse for this neglect is twofold: (1) I think that Sellars' account of the hypothetical imperative, though interesting and important, is not as central to his general conception of moral justification as its prominence in SM would seem to suggest; and (2) my discussion of this topic would have made this essay, already too long, intolerably longer. Anyone interested in pursuing Sellars' views on this matter, however, will find it necessary to supplement my account here with careful attention to chap. 6 of SM.
26 For Sellars' discussion of this difficulty, see IIO, 207-9.
27 Thomas Nagel, The Possibility of Altruism (Oxford: Clarendon Press, 1970).