The topic which I have chosen is one which eminently deserves a thorough, systematic, and fully theoretical treatment; such an approach would involve, I suspect, a careful analysis of the often subtly different kinds of state which may be denoted by the word

"want, together with a comprehensive examination of the role which different sorts of wanting play in the psychological equipment of rational (and non-rational) creatures. While 1 hope to touch on matters of this sort, 1 do not feel myself to be quite in a position to attempt an analysis of this kind, which would in any case be a very lengthy undertaking. So, to give direction to my discussion, and to keep it within tolerable limits, I shall relate it to some questions arising out of Aristotle's handling of this topic in the Nicomachean Ethics; such a procedure on my part may have the additional advantage of emphasizing the idea, in which I believe, that the proper habitat for such great works of the past as the Nicomachean Ethics is not the museums but the marketplaces of philosophy.

My initial Aristotelian question concerns two conditions which Aristotle supposes to have to be satisfied by whatever is to be recognized as being the good for man. At the beginning of Nicomachean Ethics 1. 4, Aristotle notes that there is general agreement that the good for man is to be identified with eudaemonia (which may or may not be well rendered as 'happiness'), and that this in turn is to be identified with living well and with doing well; but remarks that there is large-scale disagreement with respect to any further and more informative specification of eudaemonia. In I. 7 he seeksto confirm the identification of the good for man with eudaemonia by specifying two features, maximal finality (unqualified finality) and self-sufficiency, which, supposedly, both are required of anything which is to qualify as the good for man, and are also satisfied by eudaemonia. 'Maximal finality' is defined as follows: "Now we call that which is in itself worthy of pursuit more final than that which is worthy of pursuit for the sake of something else, and that which is never desirable for the sake of something else more final than the things which are desirable both in themselves and for the sake of that other thing, and therefore we call final without qualification that which is always desirable in itself and never for the sake of something else." Eudaemonia seems (intuitively) to satisfy this condition; such things as honour, pleasure, reason, and virtue (the most popular candidates for identification with the good for man and with eudaemonia) are chosen indeed for themselves (they would be worthy of choice even if nothing resulted from them); but they are also chosen for the sake of eudaemonia, since

"we judge that by means of them we shall be happy". Eudaemonia, however, is never chosen for the sake of anything other than itself.

After some preliminaries, the relevant sense of "self-sufficiency" is defined thus: "The self-sufficient we now define as that which when isolated makes life desirable and lacking in nothing." Eudaemonia, again, appears to satisfy this condition too; and Aristotle adds the possibly important comment that eudaemonia is thought to be "the most desirable of all things, without being counted as one good thing among others". This remark might be taken to suggest that, in Aristotle's view, it is not merely true that the possession of eudaemonia cannot be improved upon by the addition of any other good, but it is true because eudaemonia is a special kind of good, one which it would be inappropriate to rank alongside other goods.

This passage in Nicomachean Ethics raises in my mind several queries:

(1) It is, I suspect, normally assumed by commentators that Aristotle thinks of eudaemonia as being the only item which satisfies the condition of maximal finality. This uniqueness claim is not, however, explicitly made in the passage (nor, so far as I can recollect, elsewhere); nor is it clear to me that if it were made itwould be correct. Might it not be that, for example, lazing in the sun is desired, and is desirable, for its own sake, and yet is not something which is also desirable for the sake of something else, not even for the sake of happiness? If it should turn out that there is a distinction, within the class of things desirable for their own sake (1-desirables), between those which are also desirable for the sake of endaemonia (H-desirables) and those which are not, then the further question arises whether there is any common feature which distinguishes items which are (directly) H-desirable, and, if so, what it is. This question will reappear later.

(2) Aristotle claims that honour, reason, pleasure, and virtue are all both I-desirable and H-desirable, But, at this stage in the Nicomachean Ethics, these are uneliminated candidates for identification with eudaemonia; and, indeed, Aristotle himself later identifies, at least in a sort of way, a special version of one of them (metaphysical contemplation) with eudaemonia. Suppose that it were to be established that one of these candidates (say, honour) is successful. Would not Aristotle then be committed to holding that honour is both desirable for its own sake, and also desirable for the sake of something other than honour, namely, eudaemonia, that is, honour? It is not clear, moreover, that this prima facie inconsistency can be eliminated by an appeal to the non-extensionality of the context " is desirable". For while the argument-pattern

*o is desirable for the sake of B. Bis identical with y; so, or is desirable for the sake of y' may be invalid, it is by no means clear that the argument-pattern 'o is desirable for the sake of B, necessarily B is identical with Y; so, a is desirable for the sake of y' is invalid.

And, if it were true that eudaemonia is to be identified with honour, this would presumably be a non-contingent truth.

(3) Suppose the following: (a) playing golf and playing tennis are each I-desirables, (b) each is conducive to physical fitness, which is itself I-desirable, (e) that a daily round of golf and a daily couple of hours of tennis are each sufficient for peak physical fitness, and (if you like, for simplicity), (d) that there is no third route to physical fitness. Now, X and Y accept all these suppositions; X plays golf daily, and Y plays both golf and tennis daily. It seems difficult to deny, first, that it is quite conceivable that all of the sporting activities of these gentlemen are undertaken both for their own sake and also for the sake of physical fitness, and, second, that (protanto) the life of Y is more desirable than the life of X, since Y has the value of playing tennis while X does not. The fact that in Y's life physical fitness is overdetermined does not seem to be a ground for denying that he pursues both golf and tennis for the sake of physical fitness; if we wished to deny this, it looks as if we could, in certain circumstances, be faced with the unanswerable question, "If he doesn't pursue each for the sake of physical fitness, then which one does he pursue for physical fitness?"

Let us now consider how close an analogy to this example we can construct if we search for one which replaces references to physical fitness by references to eudaemonia. We might suppose that X and Y have it in common that they have distinguished academic lives, satisfying family situations, and are healthy and prosperous; that they value, and rightly value, these aspects of their existences for their own sakes and also regard them as contributing to their eudaemonia. Each regards himself as a thoroughly happy man. But Y, unlike X, also composes poetry, an activity which he cares about and which he also thinks of as something which contributes to his endaemoria; the time which y devotes to poetic endeavour is spent by X pottering about the house doing nothing in particular. We now raise the question whether or not Y's life is more desirable than X's, on the grounds that it contains an I-desirable element, poetic composition, which X's life does not contain, and that there is no counterbalancing element present in X's life but absent in Y's.

One conceivable answer would be that Y's life is indeed more desirable than X's, since it contains an additional value, but that this fact is consistent with their being equal in respect of eudae-monia, in line with the supposition that each regards himself as thoroughly happy. If we give this answer we, in effect, reject the Aristotelian idea that eudaemonia is, in the appropriate sense, self-sufficient. There seems to me, however, to be good reason not to give this answer. Commentators have disagreed about the precise interpretation of the word "eudaemonia", but none, so far as I know, has suggested what I think of as much the most plausible conjecture; namely, that "eudaemonia" is to be understood as the name for that state or condition which one's good daemon would (if he could) ensure for one; and my good daemon is a being motivated, with respect to me, solely by concern for my well-being or happiness.To change the idiom, "eudaemonia" is the general characterization of what a full-time and unhampered fairy godmother would secure for you. The identifications regarded by Aristotle as unexcitingly correct, of eudaemonia with doing well and with living well, now begin to look like necessary truths. If this interpretation of

"eudaemonia" is correct (as I shall brazenly assume) then it would be quite impossible for Y's life to be more desirable than X's, though X and Y are equal in respect of eudaemonia; for this would amount to Y's being better off than X, though both are equally well-off!

Various other possible answers remain. It might be held that not only is Y's life more desirable than X's, but Y is more eudaemon (better off) than X. This idea preserves the proposed conceptual connection between eudaemonia and being well-off, and relies on the not wholly implausible principle that the addition of a value to a life enhances the value of that life (whatever, perhaps, the liver may think). One might think of such a principle, when more fully stated, as laying down or implying that any increase in the combined value of the H-desirable elements realized in a particular life is reflected, in a constant proportion, in an increase in the degree of happiness or well-being exemplified by that life; or, more cautiously, that the increase in happiness is not determined by a constant proportion, but rather in some manner analogous to the phenomenon of diminishing marginal utility. I am inclined to see the argument of this chapter as leading towards a discreet erosion of the idea that the degree of a particular person's happiness is the value of a function the arguments of which are measures of the particular H-desirables realized in that person's life, no matter what function is suggested; but at the present moment it will be sufficient to cast doubt on the acceptability of any of the crudest versions of this idea. To revert to the case of X and Y: it seems to me that when we speak of the desirability of X's life or of Y's life, the desirability of which we are speaking is the desirability of that life from the point of view of the person whose life it is; and that it is therefore counterintuitive to suppose that, for example, X who thinks of himself as "perfectly happy" and so not to be made either better off or more happy (though perhaps more accomplished) by an injection of poetry composition, should be making a misassess-ment of what his state of well-being would be if the composition of poetry were added to his occupation. Furthermore, if the pursuitof happiness is to be the proper end, or even a proper end, of living, to suppose that the added realization of a further H-desirable to a life automatically increases the happiness or wellbeing of the possessor of that life will involve a commitment to an ethical position which I, for one, find somewhat unattractive: one would be committed to advocating too unbridled an eudae-monic expansionism.

A more attractive position would be to suppose that we should invoke, with respect to the example under consideration, an analogue not of diminishing marginal utility, but of what might be called vanishing marginal utility; to suppose, that is, that X and Y are, or at least may be, equally well-off and equally happy even though Y's life contains an H-desirable element which is lacking in X's life; that at a certain point, so to speak, the bucket of happiness is filled, and no further inpouring of realized H-desirables has any effect on its contents. This position would be analogous to the view I adopted earlier with respect to the possible over-determination of physical fitness. Even should this position be correct, it must be recognized that the really interesting work still remains to be done; that would consist in the characterization of the conditions which determine whether the realization of a particular set of H-desirables is sufficient to fill the bucket.

The main result, then, of the discussion has been to raise two matters for exploration; first, the possibility of a distinction between items which are merely I-desirable and items which are not only I-desirable but also H-desirable; and, second, the possibility that the degree of happiness exemplified by a life may be overdetermined by the set of H-desirables realized in that life, together with the need to characterize the conditions which govern such overdetermination.

(4) Let us move in a different direction. I have already remarked that, with respect to the desirability-status of happiness and of the means thereto, Aristotle subscribed to two theses, with which I have no quarrel (or, at least, shall voice no quarrel).

(A) That some things are both 1-desirable and H-desirable (are both ends in themselves and also means to happiness).

(B)

That happiness, while desirable in itself, is not desirable for the sake of any further end.I have suggested the possibility that a further thesis might be true (though I have not claimed that it is true), namely:

(C) That some things are I-desirable without being H-desirable (and, one might add, perhaps without being desirable for the sake of any further end, in which case happiness will not be the only item which is not desirable for the sake of any further end).

But there are two further as yet unmentioned theses which 1 am inclined to regard as being not only true, but also important: first,

(D) Any item which is directly H-desirable must be I-desirable.

And second,

(E) Happiness is attainable only via the realization of items which are 1-desirable (and also of course H-desirable).

Thesis (D) would allow that an item could be indirectly H-desirable without being I-desirable; engaging in morning press-ups could be such an item, but only if it were desirable for the sake of (let us say) playing cricket well, which would plainly be itself an item which was both 1-desirable and H-desirable. A thesis related to (D), namely, (D'). (An item can be directly conducive to the happiness of an individual x only if it is regarded by x as being I-desirable) seems to me very likely to be true; the question whether not only (D') but (D) are true would depend on whether a man who misconceives (if that be possible) certain items as being I-desirable could properly be said to achieve happiness through the realization of those items. To take an extreme case, could a wicked man who pervertedly regards cheating others in an ingenious way as being I-desirable, and who delights in so doing, properly be said to be (pro tanto) achieving happiness? I think Aristotle would answer negatively, and I am rather inclined to side with him; but I recognize that there is much to debate. A consequence of thesis (D), if true, would be that there cannot be a happiness-pill (a pill the taking of which leads directly to happiness); there could be (and maybe there is) a pill which leads directly to "feeling good" or to euphoria; but these states would have to be distinguishable from happiness.

Thesis (E) would imply that happiness is essentially a dependent state; happiness cannot just happen; its realization is conditionalupon the realization of one or more items which give rise to it.

Happiness should be thought of adverbially; to be happy is, for some x, to x happily or with happiness. And reflection on the interchangeability or near-interchangeability of the ideas of happiness and of well-being would suggest that the adverbial in question is an evaluation adverbial.

The importance, for present purposes, of the two latest theses is to my mind that questions are now engendered about the idea that items which are chosen (or desirable) for the sake of happiness can be thought of as items which are chosen (or desirable) as means to happiness, at least if the means-end relation is conceived as it seems very frequently to be conceived in contemporary philosophy; if, that is, x is a means to y just in case the doing or producing of x designedly causes (generates, has as an effect) the occurrence of y. For, if items the realization of which give rise to happiness were items which could be, in the above sense, means to happiness, (a) it should be conceptually possible for happiness to arise otherwise than as a consequence of the occurrence of any such items, and (b) it seems too difficult to suppose that so non-scientific a condition as the possession of intrinsic desirability should be a necessary condition of an item's giving rise to happiness. In other words, theses (D) and (E) seem to preclude the idea that what directly gives rise to happiness can be, in the currently favoured sense, a means to happiness.

The issue which I have just raised is closely related to a scholarly issue which has recently divided Aristotelian commentators; battles have raged over the question whether Aristotle conceived of eudaemonia as a 'dominant or as an 'inclusive' end. The terminology derives, I believe, from W. F. R. Hardie; but I cite a definition of the question which is given by Ackrill in a recent paper:

"By 'an inclusive end' might be meant any end combining or including two or more values or activities or goods... By 'a dominant end' might be meant a monolishic end, an end consisting of just one valued activity or good." One's initial reaction to this formulation may fall short of overwhelming enlightenment, among other things, perhaps, because the verb 'include' appears within

J. L. Ackrill, Aristotle on Eudaemonia (Dawes Hicks Lecture, Oxford: Oxford University Press, 1974), 5.the characterization of an inclusive end. I suspect, however, that this deficiency could be properly remedied only by a logico-metaphysical enquiry into the nature of the 'inclusion relation' (or, rather, the family of inclusion relations), which would go far beyond the limits of my present undertaking. But, to be less ambitious, let us, initially and provisionally, think of an inclusive end as being a set of ends. If happiness is in this sense an inclusive end, then we can account for some of the features displayed in the previous section. Happiness will be dependent on the realization of subordinate ends, provided that the set of ends constituting happiness may not be the empty set (a reasonable, if optimistic, assumption).

Since the "happiness set" has as its elements I-desirables, what is desirable directly for the sake of happiness must be I-desirable. And if it should turn out to be the case, contrary perhaps to the direction of my argument in the last section, that the happiness set includes all I-desirables, then we should have difficulty in finding any end for the sake of which happiness would be desirable.

So far so good, perhaps; but so far may not really be very far at all. Some reservation about the treatment of endaemonia as an inclusive end is hinted at by Ackrill:

It is not necessary to claim that Aristotle has made quite clear how there may be 'components' in the best life or how they may be interrelated The very idea of constructing a compound end out of two or more independent ends may arouse suspicion. Is the compound to be thought of as a mere aggregate or as an organized system? If the former, the move to cudaemonia seems trivial-nor is it obvious that goods can be just added together. If the latter, if there is supposed to be a unifying plan. what is it?"

From these very pertinent questions, Ackrill detaches himself, on the grounds that his primary concern is with the exposition and not with the justification of Aristotle's thought. But we cannot avail ourselves of this rain check, and so the difficulties which Ackrill touches on must receive further exposure.

Let us suppose a next-to-impossible world W, in which there are just three I-desirables, which are also H-desirables, A, B, and

C. If you like, you may think of these as being identical, respect-ively, with honour, wealth, and virtue. If, in general, happiness is

• Ibid. 10-n1.to be an inclusive end, happiness-in-W will have as its components A, B, and C, and no others. Now one might be tempted to suppose that, since it is difficult or impossible to deny that to achieve happiness-in-W it is necessary and also sufficient to realize A, to realize B, and to realize C, anyone who wanted to realize A, wanted to realize B, and wanted to realize C would ipso facto be someone who wanted to achieve happiness-in-W. But there seems to me to be a good case for regarding such an inference as invalid. To want to achieve happiness-in-W might be equivalent to wanting to realize A and to realize B and to realize C, or indeed to wanting A and B and C; but there are relatively familiar reasons for allowing that, with respect to a considerable range of psychological verbs (rep-resented by v), one cannot derive from a statement of the form

'x y's (that) A and x y's (that) B' a statement of the form 'x y's (that) A and B. For instance, it seems to me a plausible thesis that there are circumstances in which we should want to say of someone that he believed that p and that he believed that q, without being willing to allow that he believed that both p and q. The most obvious cases for the application of the distinction would perhaps be cases in which p and q are inconsistent; we can perhaps imagine someone of whom we should wish to say that he believed that he was a grotesquely incompetent creature, and that he also believed that he was a world-beater, without wishing to say of him that he believed that he was both grotesquely incompetent and a world-beater. Inconsistent beliefs are not, or are not neces-sarily, beliefs in inconsistencies. Whatever reasons there may be for allowing that a man may believe that p and beheve that g without believing that p and q would, I suspect, be mirrored in reasons for allowing that a man may want A and want B without wanting both A and B; if I want a holiday in Rome, and also want some headache pills, it does not seem to me that ipso facto I want a holiday in Rome and some headache pills.

Moreover, even if we were to sanction the disputed inference, it would not, I think, be correct to make the further supposition that a man who wants A and B (simply as a consequence of wanting A and wanting B) would, or even could, want A (or want B) for the sake of or with a view to, realizing A and B. So even if, in world W, a man could be said to want A and B and C, on the strength of wanting each one of them, some further condition wouldhave to be fulfilled before we could say of him that he wanted each of them for the sake of realizing A and B and C, that is, for the sake of achieving happiness-in-W.

In an attempt to do justice to the idea that happiness should be treated as being an 'inclusive end, let me put forward a modest proposal; not, perhaps, the only possible proposal, but one which may seem reasonably intuitive. Let us categorize, for present pur-poses, the l-desirables in world Was 'universals. I propose that to want, severally, each of these I-desirables should be regarded as equivalent to wanting the set whose members are just those I-desirables, with the understanding that a set of universals is not itself a universal. So to want A, want B, and want C is equivalent to wanting the set whose members are A, B, and C (the happiness-in-W set). To want happiness-in-W requires satisfaction of the stronger condition of wanting A and B and C, which in turn is equivalent to wanting something which is a universal, namely, a compound universal in which are included just those universals which are elements of the happiness-in-W set. 1 shall not attempt to present a necessary and sufficient condition for the fulfilment of the stronger rather than merely of the weaker condition; but I shall suggest an important sufficient condition for this state of affairs. The condition is the following: for x to want the conjunction of the members of a set, rather than merely for him to want, severally, each member of the set, it is sufficient that his wanting. severally, each member of the set should be explained by (have as one of its explanations) the fact that there is an 'open' feature F which is believed by x to be exemplified by the set, and the realization of which is desired by x. By an open feature I mean a feature the specification of which does not require the complete enumeration of the items which exemplify it. To illustrate, a certain Oxford don at one time desired to secure for himself the teach-ing, in his subject, at the colleges of Somerville, St Hugh's, St Hilda's, Lady Margaret Hall, and St Anne's. (He failed, by two colleges.)

This compound desire was based on the fact that the named colleges constituted the totality of women's colleges in Oxford, and he desired the realization of the open feature consisting in his teach-ing, in his subject, at all the women's colleges in Oxford. This sufficient condition is important in that it is, I think, fulfilled with respect to all compound desires which are rational, as distinct fromarbitrary or crazy. There can be, of course, genuinely compound desires which are non-rational, and 1 shall not attempt to specify the condition which distinguishes them; but perhaps I do not need to, since I think we may take it as a postulate that, if a desire for happiness is a compound desire, it is a rational compound desire.

The proposal which I have made does, I think, conform to acceptable general principles for metaphysical construction. For it provides for the addition to an initially given category of items (universals') of a special sub-category (compound universals") which are counterparts of certain items which are not universals but rather sets of universals. It involves, so to speak, the conversion of certain non-universals into 'new' universals, and it seems reasonable to suppose that the purpose of this conversion is to bring these non-universals, in a simple and relatively elegant way, within the scope of laws which apply to universals. It must be understood that by 'laws' I am referring to theoretical generalities which belong to any of a variety of kinds of theory, including psychological, prac-tical, and moral theories; so among such laws will be laws of various kinds relating to desires for ends and for means to ends.

If happiness is an inclusive end, and it, for it to be an inclusive end the desire for which is rational, there must be an open feature which is exemplified by the set of components of happiness, our next task is plainly to attempt to identify this feature. To further this venture I shall now examine, within the varieties of means-end relation, what is to my mind a particularly suggestive kind of case.

At the start of this section I shall offer a brief sketch of the vari-eties, or of some of the varieties, of means-end relation; this is a matter which is interesting in itself, which is largely neglected in contemporary philosophy, and which I am inclined to regard as an important bit of background in the present enquiry. I shall then consider a particular class of cases in our ordinary thinking about means and ends, which might be called cases of'end-fixing, and which might provide an important modification to our consideration of the idea that happiness is an inclusive end.

123 of1 shall introduce the term 'is contributive to' as a general expression for what I have been calling 'means-end' relation, and I shall use the phrase is contributive in way w to' to refer, in a general way, to this or that particular specific form of the con-tributiveness relation. I shall, for convenience, assume that anyone who thinks of some state of affairs or action as being contributive to the realization of a certain universal would have in mind that specific form of contributiveness which would be appropriate to the particular case. We may now say, quite unstartlingly, that x wants to do A for the sake of B just in case x wants to do A because (1) x regards his doing A as something which would be contributive in way w to the realization of B, and (2) x wants B. That leaves us the only interesting task, namely, that of giving the range of specific relations one element in which will be picked out by the phrase contributive in way w, once A and B are specified.

The most obvious mode of contributiveness, indeed one which has too often been attended to to the exclusion of all others, is that of causal antecedence; x's contributing to y here consists in x's being the (or a) causal origin of y. But even within this mode there may be more complexity than meets the eye. The causal origin may be an initiating cause, which triggers the effect in the way in which flipping a switch sets off illumination in a light bulb; or it may be a sustaining cause, the continuation of which is required in order to maintain the effect in being. In either case, the effect may be either positive or negative; I may initiate a period of non-talking in Jones by knocking him cold, or sustain one by keeping my hand over his mouth. A further dimension, in respect of which examples of each variety of causal contributiveness may vary, is that of conditionality. Doing A may be desired as something which will, given the circumstances which obtain, unconditionally originate the realization of B, or as something which will do so provided that a certain possibility is fulfilled. A specially important subclass of cases of conditional causal contributiveness is the class of cases in which the relevant possibility consists in the desire or will of some agent, either the means- taker or someone else, that B should be realized; these are cases in which x wants to do A in order to enable, or to make it possible for, himself (or someone else) to achieve the realization of B; as when, for example, x puts a corkscrew in his pocket to enable him later, should be wish to do so, to open a bottle of wine.But, for present purposes, the more interesting modes of contributiveness may well be those other than that of causal con-tributiveness. These include the following types.

(1) Specificatory contributiveness. To do A would, in the prevailing circumstances, be a specification of, or a way of, realizing

B; it being understood that, for this mode of contributiveness, B is not to be a causal property, a property consisting in being such as to cause the realization of C, where C is some further property.

A host's seating someone at his right-hand side at dinner may be a specification of treating him with respect; waving a Union Jack might be a way of showing loyalty to the Crown. In these cases, the particular action which exemplifies A is the same as the item which exemplifies B.

Two further modes involve relations of inclusion, of one or

another of the types to which such relations may belong.

1. To do A may contribute to the realization of B by including an item which realizes B. I may want to take a certain advertised cruise because it includes a visit to Naples.
2. To do A may contribute to the realization of B by being included in an item which realizes B. Here we may distinguish more than one kind of case. A and B may be identical; 1 may, for example, be hospitable to someone today because I want to be hospitable to him throughout his visit to my town. In such a case the exemplification of B (hospitality) by the whole (my behaviour to him during the week) will depend on a certain distribution of exemplifications of B among the parts, such as my behaviour on particular days. We might call this kind of dependence "component-dependence". In other cases A and B are distinct, and in some of these (perhaps all) B cannot, if it is exemplified by the whole, also be exemplified by any part. These further cases subdivide in ways which are interesting but not germane to the present enquiry.

We are now in a position to handle, not quite as Aristotle did, a 'paradox' about happiness raised by Aristotle, which involves Solon's dictum "Call no man happy till he is dead", I give a simplified, but I hope not distorted, version of the 'paradoxical' line of argument. If we start by suggesting that happiness is the end for man, we shall have to modify this suggestion, replacing

"happiness" by "happiness in a complete life". (Aristotle himselfapplies the qualification "in a complete life" not to happiness, but to what he gives as constituted of happiness, namely, activity of soul in accordance with excellence). For, plainly, a life which as a whole exemplifies happiness is preferable to one which does not.

But since lifelong happiness can only be exemplified by a whole life, non-predictive knowledge that the end for man is realized with respect to a particular person is attainable only at the end of the person's life, and so not (except possibly at the time of his dying gasp) by the person himself. But this is paradoxical, since the end for man should be such that non-predictive knowledge of its realization is available to those who achieve its realization.

I suggest that we need to distinguish non-propositional, attributive ends, such as happiness, and propositional ends or objectives, such as that my life, as a whole, should be happy. Now it is not in fact clear that people do, or even should, desire lifelong happiness; it may be quite in order not to think about this as an objective.

And, even if one should desire lifelong happiness, it is not clear that one should aim at it, that one should desire, and do, things for the sake of it. But let us waive these objections. The attainment of lifelong happiness, an objective, consists in the realization, in a whole life, of the attributive end happiness. This realization is component-dependent; it depends on a certain distribution of realizations of that same end in episodes or phases of that life.

But these realizations are certainly non-predictively knowable by the person whose life it is. So, if we insist that to specify the end for man is to specify an attributive end and not an objective, then the "paradox' disappears.

The special class of cases to which one might be tempted to apply the term 'end-fixing' may be approached in the following way. For any given mode of contributiveness, say causal contributiveness, the same final position, that x wants (intends, does) A as contributive to the realization of B, may be reached through more than one process of thought. In line with the canonical Aristotelian model, x may desire to realize B, then enquire what would lead to B, decide that doing A would lead to B, and so come to want, and to do, A.

Alternatively, the possibility of doing A may come to his mind, he then enquires what doing A would lead to, sees that it would lead to B, which he wants, and so he comes to want, and perhaps do, A. I now ask whether there are cases in which the followingconditions are met: (1) doing A is fixed or decided, not merely entertained as a possibility, in advance of the recognition of it as desirable with a view to B, and (2) that B is selected as an end, or as an end to be pursued on this occasion, at least partly because it is something which doing A will help to realize.

A variety of candidates, not necessarily good ones, come to mind.

(1) A man who is wrecked on a desert island decides to use his stay there to pursue what is a new end for him, namely, the study of the local flora and fauna. Here doing A (spending time on the island) is fixed but not chosen; and the specific performances, which some might think were more properly regarded as means to the pursuit of this study, are not fixed in advance of the adoption of the end. (2) A man wants (without having a reason for so want-ing) to move to a certain town; he is uncomfortable with irrational desires (or at least with this irrational desire), and so comes to want to make this move because the town has a specially salubrious cli-mate. Here, it seems, the movement of thought cannot be fully conscious; we might say that the reason why he wants to move to a specially good climate is that such a desire would justify the desire or intention, which he already has, to move to the town in question; but one would baulk at describing this as being his reason for wanting to move to a good climate.

The example which interests me is the following. A tyrant has become severely displeased with one of his ministers, and to humiliate him assigns him to the task of organizing the disposal of the palace garbage, making clear that only a high degree of efficiency will save him from a more savage fate. The minister at first strives for efficiency merely in order to escape disaster; but later, seeing that thereby he can preserve his self-respect and frustrate the tyrant's plan to humiliate him, he begins to take pride in the efficient discharge of his duties, and so to be concerned about it for its own sake. Even so, when the tyrant is overthrown and the minister is relieved of his menial duties, he leaves them without regret in spite of having been intrinsically concerned about their discharge.

One might say of the minister that he efficiently discharged his office for its own sake in order to frustrate the tyrant; and this is clearly inadequately represented as his being interested in the efficient discharge of his office both for its own sake and for thesake of frustrating the tyrant, since he hoped to achieve the latter goal by an intrinsic concern with his office. It seems clear that higher-order desires are involved; the minister wants, for its own sake, to discharge his office efficiently, and he wants to want this because he wants, by so wanting, to frustrate the tyrant. Indeed, wanting to do A for the sake of B can plausibly be represented as having two interpretations. The first interpretation is invoked if we say that a man who does A for the sake of B (1) does A because he wants to do A and (2) wants to do A for the sake of B. Here wanting A for the sake of B involves thinking that A will lead to B. But we can conceive of wanting A for the sake of B (analogously with doing A for the sake of B) as something which is accounted for by wanting to want A for the sake of B; if so, we have the second inter-pretation, one which implies not thinking that A will help to realize B, but rather thinking that wanting A will help to realize B.

The impact of this discussion, on the question of the kind of end which happiness should be taken to be, will be that, if happiness is to be regarded as an inclusive end, the components may be not the realizations of certain ends, but rather the desires for those realizations. Wanting A for the sake of happiness should be given the second mode of interpretation specified above, one which involves thinking that wanting A is one of a set of items which collectively exhibit the open feature associated with happiness.

III

My enquiry has, I hope, so far given some grounds for the favourable consideration of three theses:

1. happiness is an end for the sake of which certain I-desirables are desirable, but is to be regarded as an inclusive rather than a dominant end;
2. for happiness to be a rational inclusive end, the set of its components must exemplify some particular open feature, yet to be determined; and
3. the components of happiness may well be not universals or states of affairs the realization of which is desired for its own sake,but rather the desires for such universals or states of affairs, in which case a desire for happiness will be a higher-order desire, a desire to have, and satisfy, a set of desires which exemplifies the relevant open feature.

   At this point, we might be faced with a radical assault, which would run as follows. "Your whole line of enquiry consists in assuming that, when some item is desired, or desirable, for the sake of happiness, it is desired, or desirable, as a means to happiness, and in then raising, as the crucial question, what kind of an end happiness is, or what kind of means-end relation is involved. But the initial assumption is a mistake. To say of an item that it is desired for the sake of happiness should not be understood as implying that that item is desired as any kind of a means to anything. It should be understood rather as claiming that the item is desired (for its own sake) in a certain sort of way? 'for the sake of happiness' should be treated as a unitary adverbial, better heard, perhaps, as happiness-wise. To desire something happiness-wise is to take the desire for it seriously in a certain sort of way, in particular to take the desire seriously as a guide for living, to have incorporated it in one's overall plan or system for the conduct of life. If one looks at the matter this way, one can see at once that it is conceivable that these should be I-desirables which are not H-desirables; for the question whether something which is desirable is intrinsically desirable, or whether its desirability derives from the desirability of something else, is plainly a different question from the question whether or not the desire for it is to be taken seriously in the planning and direction of one's life, that is, whether the item is H-desirable. One can, moreover, do justice to two further considerations which you have, so far, been ignoring: first, that what goes to make up happiness is relative to the individual whose happiness it is, a truth which is easily seen when it is recognized that what x desires (or should desire) happiness-wise may be quite different from what y so desires; and, second, that intuition is sympathetic to the admittedly vague idea that the decision that certain items are constitutive of one's happiness is not so much a matter of judgement or belief as a matter of will. One's happiness consists in what one makes it consist in, an idea which will be easily accommodated if "for the sake of happiness' is understood in the way which I propose."There is much in this (spirited yet thoughtful) oration towards which I am sympathetic and which 1 am prepared to regard as important; in particular, the idea of linking H-desirability with desires or concerns which enter into a system for the direction of one's life, and the suggestion that the acceptance of a system of ends as constituting happiness, or one's own happiness, is less a matter of belief or judgement than of will. But, despite these attractive features, and despite its air of simplifying iconoclasm, the position which is propounded can hardly be regarded as tenable. When looked at more closely, it can be seen to be just another form of subjectivism: what are ostensibly beliefs that particular items are conducive to happiness are represented as being in fact psychological states or attitudes, other than beliefs, with regard to these items; and it is vulnerable to variants of stock objections to subjectivist manceuvres. That in common speech and thought we have application for, and so need a philosophical account of, not only the idea of desiring things for the sake of happiness but, also, that of being happy (or well-off), is passed over; and should it turn out that the position under consideration has no account to offer of the latter idea, that would be not only paradoxical but also, quite likely, theoretically disastrous. For it would seem to be the case that the construction or adoption of a system of ends for the direction of life is something which can be done well or badly, or better or less well; that being so, there will be a demand for the specification of the criteria governing this area of evaluation; and it will be difficult to avoid the idea that the conditions characteristic of a good system of ends will be determined by the fact that the adoption of a system conforming to those conditions will lead, or is likely to lead, or other things being equal will lead, to the realization of happiness; to something. that is, which the approach under consideration might well not be able to accommodate.

   So it begins to look as if we may be back where we were before the start of this latest discussion. But perhaps not quite; for, per-haps, something can be done with the notion of a set or system of ends which is suitable for the direction of life. The leading idea would be of a system which is maximally stable, one whose employment for the direction of life would be maximally conduciveto its continued employment for that purpose, which would be maximally self-perpetuating. To put the matter another way, a system of ends would be stable to the extent to which, though not constitutionally immune from modification, it could accommodate changes of circumstances or vicissitudes which would impose modification upon other less stable systems. We might need to supplement the idea of stability by the idea of flexibility; a system will be flexible in so far as, should modifications be demanded, they are achievable by easy adjustment and evolution; flounder-ings, crises, and revolutions will be excluded or at a minimum.

   A succession of systems of ends within a person's consciousness could then be regarded as stages in the development of a single life-scheme, rather than as the replacement of one life-scheme by another. We might find it desirable also to incorporate into the working-out of these ideas a distinction, already foreshadowed, between happiness-in-general and happiness-for-an-individual.

   We might hope that it would be possible to present happiness-in-general as a system of possible ends which would be specified in highly general terms (since the specification must be arrived at in abstraction from the idiosyncrasies of particular persons and their circumstances), a system which would be determined either by its stability relative to stock vicissitudes in the human condi-tion, or (as I suspect) in some other way; and we might further hope that happiness for an individual might lie in the posses-sion, and operation for the guidance of life, of a system of ends which (a) would be a specific and personalized derivative, determined by that individual's character, abilities, and situations in the world, of the system constitutive of happiness in general; and

   (b) the adoption of which would be stable for that individual in his circumstances.

   The idea that happiness might be fully, or at least partially, characterized in something like this kind of way wouid receive some support if we could show reason to suppose that features which could plausibly be regarded, or which indeed actually have been regarded, as characteristic of happiness, or at least of a satisfactory system for the guidance of life, are also features which are conducive to stability. I shall list some features for which, in this regard, the prospects seem good.

4. Feasibility. An adopted system of ends should be workable; the more it should turn out that actions and performances dictated by the system cannot be successfully undertaken, the stronger are the grounds for modification of the system. A particular case of the operation of this feature lies in the demand that an agent should be equipped, by nature or by training, with the competencies needed for the effective prosecution of his system of ends.
5. Autonomy. This feature is closely related to the preceding one. The less reliant one's system of ends is on aids the availability of which is not within one's control, particularly if it is within the control of others, the less dependent the system is on what Aristotle called "ektos choregia", the more stable, or the more securely stable, it will in general be. Unless one has firm guaran-tees, it is better not to have to rely on the availability of elaborate machinery or government grants.
6. Compatibility of component ends. Initially, one might suppose that there are grounds for the modification of a system of ends in so far as the fulfilment of certain ends in the system thwarts the fulfilment of certain others. But I think we have to recognize that, characteristically, an end is such as to be realizable in varying degrees, and that it would be unrealistic to demand a system in which the realization of one end was never diminished by the realization of others. What we in fact may reasonably look for is a harmony of ends; the possibility, that is, with respect to competing ends, of finding an acceptable balance in the degrees of realization to be expected for each end. How such balances are to be determined is a large and difficult question, but their unavailability would prompt modification of the system. This feature looks like an analogue of consistency, which is commonly favoured as a feature of non-practical systems, though perhaps more by some people than by others, like Wittgenstein and Norman O. Brown.
7. Comprehensiveness. (An analogue for completeness.) A system is comprehensive to the extent to which it yields decisions with respect to particular practical questions; the more undecidabilit-ies, the less the comprehensiveness. To be more accurate, the comprehensiveness of a system varies directly with its capacity to yield answers to those practical questions which should be decided in the light of general principles. In ordinary circumstances it would, for example, be inappropriate to try to invoke one's life-schemeto decide whether one should have beef or lamb for dinner tonight.

   Deficiency in comprehensiveness seems to legitimize modification.

   1. Supportiveness of component ends. A system's stability will be increased if the pursuit of some ends enhances the pursuit of others. Such enhancement may arise in more than one way; for example, a man's dedication to mathematical studies might yield increased skill as a chess-player; or his devotion to his wife might inspire him to heightened endeavour in his business of selling encyclopaedias.
   2. Simplicity. A system's effectiveness as a guide to living will depend, in part, on how easy it is to determine its deliverances on particular questions. Ifit yields answers on practical questions, but these answers are difficult to discern, the system will be at a disadvantage when compared with another, whose greater simplicity makes its deliverances more accessible.
   3. Agreeableness. One form of agreeableness will, unless coun-teracted, automatically attach to the attainment of an object of desire, such attainment being routinely a source of satisfaction. The generation of satisfaction will, then, not provide an independent ground for preferring one system of ends to another. But other modes of agrecableness, such as being a source of delight, which are not routinely associated with the fulfilment of desire, could dis criminate independently of other features relevant to such pref-erences, between one system and another. A system the operation of which is specially agreeable would be stable not only vis-d-vis rival systems, but also against the weakening effect of incontinence: a disturbing influence is more surely met by a principle in consort with a supporting attraction than by the principle alone.

   However promising the signs may so far have seemed to be, I very much doubt whether the proposed characterization of happiness can be more than a partial characterization. Eirst, there seem to be features which intuition would require that an optimal system of ends should exemplify, but which cannot be represented as promotive of stability. Many people would hold that, other things being equal, a person's system of ends should be such as to involve maximal development of his natural talents; and many would hold that, where this is possible, a system should provide scope for outstanding or distinctive personal achievement. If these views arecorrect, it seems difficult to furnish for them a justifying connection with the ideas of stability and flexibility, Second, the features associated with stability seem to be, even in combination, insufficiently selective. All the listed features, except the last, seem to be systemic in character; and difficulties seem to arise with respect to them which are reminiscent of a stock objection to a familiar form of the Coherence Theory of Truth. Proponents of the idea that membership of a coherent and comprehensive system of propositions is necessary and sufficient for being true are met with the reply that a plurality of such systems, each inconsistent with the others, is conceivable, and that, to eliminate from candidacy for truth all but one member of such a plurality, it will be necessary to appeal to an extra-systematic condition, such as incorrig-ibility or certification by observation. In somewhat similar style, we can point to systems of ends which, so far as one can tell, might be undifferentiated with respect to stability and the features associated therewith, including agreeableness, yet which intuitively would be by no means equally approvable for the guidance of liv-ing; for example, such systems as might be espoused by a hermit, by a monomaniacal stamp-collector, by an unwavering egotist, and by a well-balanced, kindly country gentleman.

   To resolve such difficulties, an extra-systematic condition seems to be required, one which will differentiate ends or systems of ends in respect of value. Here I would seek to explore a road not entirely different from that taken by Aristotle. I would like to consider the possibility that the idea of happiness-in-general might be determined by reference to the essential characteristics of a human being (rational animal); the ends involved in the idea of happiness-in-general would, perhaps, be the realization in abund-ance, in various forms specific to individual men, of those capacities with which a creature-constructor would have to endow creatures in order to make them maximally viable in human living conditions, that is, in the widest manageable range of different environments.?

   P The original lecture version of this chapter concluded with this sentence:

   "But I have now almost exactly reached the beginning of the paper which, till recently, you thought I was going to read to you tonight, on the derivability of ethical principles. It is a pity that I have used up my time."]H. P. Grice.