While it is widely accepted in philosophy that some forms of moral discourse are “normative,” there is less of an orthodox view about the normativity of discourse about prudence and welfare.Footnote 1 ‘Normative’ is, of course, a term of art and different theorists define it differently, so any investigation of the normativity of discourse about prudence and welfare would do well to begin by defining ‘normative’. On one useful gloss, the normativity of an area of discourse consists in the fact that judgments made in that idiom directly function to settle the thing to do, think or feel, on pain of irrationality or are conceptually linked in the right way to such judgments The normativity of moral discourse, for example, might then consist in the fact that first-person judgments of the form “I am morally required to perform this action” settle the agent on performing that action if they can, on pain of irrationality. Other moral judgments, like judgments about what outcomes are morally good or bad or what someone has moral reason to do, are normative in virtue of either themselves settling the thing to do (or think or feel) or standing in privileged inferential relations to judgments which do. For example, the fact that there is a moral reason to do something plausibly entails that if there is no reason not to do it then one morally ought to do it, where the latter settles the thing to do; similar inferential relations can easily be found for concepts like ‘morally good’ and ‘morally bad’.

In this paper, I argue that while some forms of prudential discourse are indeed normative in this sense, discourse about welfare itself is not normative. On the view I defend here, the judgment expressed by “prudentially, I ought to eat less fatty food” is normative, but the judgment expressed by “eating less fatty food would maximize my long-term welfare” (or simply “eating less fatty food would be good for me”) is not normative. On the other hand, as I argue in Sect. 3, judgments about welfare do presuppose a certain sort of competence with normative concepts, and so in that more minimal sense they are normative. Distinguishing prudential concepts from the concept of welfare in this way is unorthodox; most theorists tend to treat the concept of welfare and prudential concepts as a “package deal” when it comes to their normativity. Indeed, claims about what would maximize my long-term welfare and claims about what I prudentially ought to do might seem to stand in conceptual entailment relations which make the distinction proposed here unstable. I argue that although there are interesting conceptual entailment relations here, they do not support the normativity of the concept of welfare, and the resulting view is not subject to any such instability.

I begin by laying out what I take to be the strongest existing case for the normativity of both prudential discourse and welfare discourse taken as a “package deal.” I then explain why the arguments for this thesis are more convincing in the case of prudential discourse than in the case of discourse about welfare (Sect. 1). I then develop two objections to the thesis that judgments about welfare are normative (Sect. 2). These objections naturally lead to the question of just what the content of welfare judgments could be if they are not normative. A key challenge here is to develop a view of the content of welfare judgments that can accommodate the depth of seemingly normative disagreement about what welfare is. Hedonists, desire-fulfillment theorists and Aristotelians (for example) all seem to disagree deeply about the nature of welfare. One of the best arguments for the normativity of welfare judgment is that the sort of disagreement in play here is relevantly similar to the kinds of disagreement used to motivate the normativity of moral discourse. I develop a positive view of the content of welfare judgments which can explain why competence with the concept of welfare requires competence with normative concepts, and can explain the relevant patterns of disagreement, but does so in a way that explains why the concept of welfare is not a normative concept (Sect. 3). They key move is to understand the concept of welfare as being such that competence with it entails accepting certain conditionals linking propositions about welfare to prudentially normative propositions but only on the further condition that there are any true prudentially normative propositions.

1 The case for the “package deal”

My focus here will be on Guy Fletcher’s defense of what I am here calling the “package deal” in Dear Prudence (Fletcher, 2021). The “package deal” holds that both discourse about what one prudentially ought to do and discourse about welfare are normative, and indeed that both of those discourses are about the same “flavor” of normativity. I focus on Fletcher’s work because he offers the most sustained, systematic and convincing existing defense of the idea that what he characterizes as “prudential discourse” is normative. I should, however, note at the outset that Fletcher is far from alone in taking the view that welfare is a normative concept. Stephen Darwall, for example, not only endorses and argues for the normativity of welfare, he asserts that its normativity is “widely accepted”:

“It seems to be widely accepted that welfare is a normative notion in the sense that an ‘ought’ or normative reason claim follows from the proposition that something is for someone’s good.” (Darwall, 2002: 4)

Darwall may exaggerate just how widely accepted the normativity of the concept of welfare is. As Fletcher nicely documents, the answer to the question “is welfare normative?” is often taken as obvious, but with different theorists having a different view of what the obvious answer is.

Nonetheless, Darwall is at least right that it is somewhat widely accepted that the concept of welfare is normative. Darwall’s own take on the normativity of the concept of welfare is somewhat idiosyncratic in ways that mean a deep dive into his views would be distracting here. On the other hand, Darwall’s view is interesting enough that I shall say a few things about it before putting it to one side. Darwall holds that the normativity of welfare entails reasons not for the agent whose welfare is at stake, but for those who care about that person. The more orthodox take on the normativity of welfare is that it (at least) entails prudential reasons for the agent whose welfare is at stake. On this more orthodox conception, these reasons are agent-relative, in the terms of the trade—very roughly, they provide reasons indexed to the agent whose welfare is at stake. Whereas the normativity of welfare on Darwall’s conception is agent-neutral—my welfare entails reasons not for me in particular, but for anyone who cares about me. An obvious worry about Darwall’s approach is that “caring about someone” sounds awfully similar to “caring about someone’s welfare” or “wanting someone to have a good life,” which threatens to make his analysis circular. Darwall is, of course, well aware of this worry and discusses it at length in his book, and I will not here rehearse the debate on this point.Footnote 2 I put Darwall’s approach to one side to focus more on specifically prudential conceptions of the normativity of the concept of welfare; Darwall’s account is better construed in terms of an altruistic or impersonal conception of its normativity. However, if the objections I make to the more orthodox conception are sound, then they should equally undermine Darwall’s approach. I return to Darwall’s views briefly in Sect. 3.

Before laying out Fletcher’s case for “the package deal,” I need to do a little exegetical work. One potentially confusing terminological issue is that Fletcher uses the broad umbrella label “prudential discourse” to include not only claims about what one prudentially ought to do, and what there is most prudential reason to do, but also claims about what would best promote someone’s welfare (Fletcher’s preferred terminology is ‘wellbeing’; see fn. 1 above). In my view, only the former is properly classified as “prudential.” Pre-theoretically, prudence is, roughly, a matter of looking after one’s long-term welfare, so talk of prudence is implicitly talk about welfare. However, the opposite entailment does not hold; talk about welfare is not implicitly talk about prudence. Someone who rejects norms of prudence altogether, thinking one ought to “live in the moment and not worry about tomorrow” might still trade in judgments of welfare. Indeed, they might characterize their view in terms of the rationality of giving more weight to present welfare than the same amount (or even more) welfare in their distant future. Admittedly, Fletcher can use ‘prudential discourse’ as a technical term, but since the distinction between welfare and prudence is at the core of the view I develop here, it is worth underscoring how in ordinary language ‘welfare’, while obviously relevant to prudence, does not itself express a prudential concept—in the sense of committing a competent user who makes positive judgments about welfare to making the corresponding judgments about what people prudentially ought to do.

On Fletcher’s view, prudential claims can usefully be divided into (a) evaluative prudential claims and (b) directive prudential claims. The former are claims about what would be good for some welfare subject. Fletcher endorses an end-relational contextualist semantics for “good for” along the lines of the one developed by Stephen Finlay (Finlay, 2014). On this semantics, for something to be good for X, is (roughly) for that thing to promote some end E associated I some way with X, where the context determines the relevant end. Fletcher argues that for welfare subjects—subjects who are the right kind to have welfare at all—the default reading of ‘good for X’ is ‘promotes X’s welfare’. By contrast, for artefacts like automobile engines, the relevant end is the proper functioning of the artefact—as in “regular oil changes are good for automobile engines.”Footnote 3 These evaluative claims about what is good or bad for someone are, on Fletcher’s view, normative. In particular, they are normative in that in contexts in which there are no relevant moral reasons in play by your lights, judging that “some option would be (sufficiently) bad for you is treated as settling whether you ought to bring about that option (and vice versa for what would be best for you).” (Fletcher, 2021: 39) Judgments about what are good for you, then, conceptually entail agent-relative prudential reasons for action. On Fletcher’s view, they also entail agent-relative reasons for certain attitudes (preference or desire), but I will here overlook this aspect of his view since the commitment to agent-relative reasons for action is sufficient for the objections I press later. Moreover, I take the connection between prudence and reasons for action as more immediately plausible than the corresponding connection to reasons for attitudes. In addition to these evaluative claims, prudential discourse in Fletcher’s taxonomy includes deontic prudential claims—claims about what one prudentially ought to do, what there is prudential reasons to do, etc. Fletcher argues that this discourse as a whole, both in its evaluative and deontic forms, is normative, thus defending the “package deal.”

How does Fletcher argue for the normativity of prudential discourse in this broad sense? Rather than lay out a specific conception of the normative, which would inevitably be somewhat tendentious, he argues by parity with the moral case. The idea is that it is relatively uncontroversial that moral discourse is normative if any discourse is, so relying on the moral case as a paradigm should be relatively uncontroversial. Fletcher then argues that there are five “marks of the normative” which are commonly cited in the moral case to justify the view that moral discourse is normative. While no one of these conditions is sufficient for normativity, the idea is that the conjunction of all five makes a strong case for normativity. Each of these five “marks of the normative” can, he argues, be found equally in the prudential case, thus supporting the conclusion that prudential discourse is normative. The five “marks of the normative” on Fletcher’s account are:

  1. (1)

    Prescriptive language: The discourse is couched in terms like ‘ought’, ‘must’, ‘requirement’, ‘reason’, ‘good’, ‘bad’, etc.

  2. (2)

    Authority: Judgments couched in the discourse are taken as directly relevant to determining what one ought to do, all things considered—to settling the thing to do.

  3. (3)

    Affect Connection: Judgments couched in the discourse are “closely connected” with certain affective responses.

  4. (4)

    Disagreement: The discourse admits of widespread and fundamental disagreement without conceptual incompetence.

  5. (5)

    Testimonial Non-Deference: It seems weird and problematic to defer entirely to someone else when making the relevant judgments.

I agree that these are plausible “marks of the normative.” Moreover, I agree that these marks are present in the case of judgments about what one prudentially ought to do, what there is prudential reason to do, what one prudentially must do. Clearly prudential judgments are made with prescriptive language. It is also very plausible that prudential judgments carry rational authority. Judging that one is prudentially required to do something and at the same time judging there is no reason (moral or otherwise) not to perform that action, while remaining indifferent to that option, would be irrational. This is extremely plausible, and it corroborates the normativity of prudential discourse.

The link to affect is also plausible. Fletcher actually adverts to preference, which is not really affective in the usual sense of that word—preference is not associated with a distinctive affect or feeling. However, he also makes the connection to reactive attitudes, which is more plausible—we tend to blame ourselves and others for their prudential mistakes. Pity, including self-pity also seems tightly connected with prudential failings, as indeed are guilt, shame, and other forms of self-reproach. We also find widespread fundamental disagreement about the requirements of prudence. Finally, it does seem at least somewhat weird and problematic to let someone else dictate what counts as making your life goes well, at least at the most fundamental level. Obviously, as Fletcher notes, I might defer to my doctor about what would best promote my health. It would, though, be weird for me to defer to my doctor (or anyone else) when it comes to whether I should put my health ahead of athletic achievement. At least, it would be weird to defer entirely to someone else in such fundamental matters.

On the strength of Fletcher’s arguments, I agree that judgments about what one prudentially ought to do, what is prudentially required, and what there is prudential reason to do are all normative. However, I do not think these arguments are equally compelling in the case of judgments about welfare. The first point obviously holds up—we do use prescriptive language in Fletcher’s sense—in particular ‘good for’ and ‘bad for’—to express judgments about welfare. Indeed, welfare itself is plausibly an evaluative term (as is well-being), since it is derived from ‘well’ which is clearly evaluative, being the adverbial form of ‘good’. However, as Fletcher rightly notes, this mark alone is hardly sufficient for normativity. We also use evaluative language when we call something a “good toaster” or when we say some conditions would be “good for E. coli bacteria,” but these are not plausibly classified as normative judgments. Unless some of the other “marks of the normative” hold up for ‘welfare’ the case for its normativity will be unconvincing.Footnote 4

What, then, about authority and the connection to affect? Here it must be conceded that most people do tend to treat judgments about their welfare as bearing on what they ought to do, and judgments about welfare do tend to trigger certain feelings as well. However, this is not sufficient to support the thesis that the concept of welfare is itself normative. For, plausibly, many people hold substantive views about their reasons for action and what they ought to do which entail that promoting their own welfare is always a reason, indeed a prudential reason, for action. Given those background normative judgments, they will of course infer that they have certain reasons for action from premises about what best promotes their welfare, and this in turn will explain why they tend to have certain affective reactions to failures to promote one’s welfare, etc. That, though, is so far compatible with the concept of welfare itself being a purely descriptive one. Compare: many people hold that keeping your promises generates a moral reason for action, and so will move seamlessly from “I promised” to “I morally ought to do it.” This, though, is perfectly consistent with the independently plausible thesis that ‘promise’ is a purely descriptive concept. It is not as if a nihilist could not competently and sincerely recognize promises and distinguish them from other speech-acts. Here we must distinguish descriptive concepts which pick out normatively relevant properties from properly normative concepts. A perhaps even clearer example is the concept of physical pain, which most people take to be normatively relevant but which itself seems to be a descriptive concept. For properly normative concepts, a judgment employing that concept must in itself be relevant, on pain of irrationality, to what one ought to do—independently of any further substantive background normative theory. In the case of ‘welfare’, this stronger condition is plausibly not met. However, it will be easier to explain why this condition is not met after I have laid out my objections to the idea that ‘welfare’ is normative in the next section, so I return to this point there.

What about disagreement? Here I am more concessive. The pattern of disagreement we find about the nature of welfare is very similar to the one we find in the moral case, and this stands in need of explanation. However, I shall argue (in Sect. 3), this can most plausibly be explained in a way that does not suggest that welfare is itself a normative concept in the intended sense.

Finally, I take a similar approach to Fletcher’s point about testimony—that we tend not to defer entirely to others about fundamental questions of our own welfare, but here too my proposed positive account of the concept of welfare can account for this. Before developing this positive account, though, I first need to lay out my objections to the idea that the concept of welfare is a normative one.

2 Why the concept of welfare is not normative

I offer two objections to the thesis that the concept of welfare is normative. The first objection draws on the work of Derek Parfit. This objection is more ambitious than the second one, in that it amounts to a sort of error theory about prudential normativity in general. This argument relies on the truth of Parfit’s views, and so is to that extent more controversial. The second objection is more modest. The second objection is, in effect, a version of G.E. Moore’s “Open Question Argument.” This argument relies only on the coherence of Parfit’s views. In fact, the second objection is even more modest than this, as it can go through even if the coherence of Parfit’s views is rejected. It will be sufficient for this second objection that certain general views about reasons for action which are widely taken to be coherent are, indeed, coherent. The role of Parfit’s views here will perhaps come as no surprise to Parfit aficionados; apparently the original title for Reasons and Persons was Against Prudence!Footnote 5

Before laying out these objections, I should clarify what is at stake in the debate over whether the concept of welfare is normative, and whether it is a prudentially normative one in particular. The issue, then, as Darwall put it, is whether an ‘ought’ or normative reason claim follows from a claim about what would be good for someone. Or, as Fletcher puts it, whether a judgment about my future welfare in itself settles the thing to do, at least modulo there being no other competing reasons, such as moral reasons, in play. Although I initially focus on the link between welfare and prudential normativity, I shall eventually widen the scope of my critique to the idea that welfare is any kind of normative concept.

Parfit’s work on personal identity is highly relevant here because prudence is an agent-relative concept, which requires that I give special weight to my future welfare just because it is mine. The idea of some future person stage being a stage of the same person that I am now is therefore doing some “heavy lifting” for those who take prudence seriously. Moreover, as classically conceived, prudence plausibly requires that (apart from discounting for uncertainty about the future), one be temporally impartial about one’s welfare, so that the same amount of pleasure counts equally in favour of an action no matter whether that pleasure occurs now or in ten years, again, discounting for uncertainty about outcomes to one side. Even if strict impartiality is rejected as not part of prudence, the idea that your future welfare matters, at least to some extent to what you should do now because it is your welfare in the future seems fairly central to our concept of prudential reasons. Parfit’s conception of personal identity puts a lot of pressure on both these elements of prudence.

I cannot here hope to summarize the intricacies of Parfit’s view of personal identity or his many ingenious arguments for that view. I shall instead simply offer a very brief summary of the view and then explore the implications of the view to the present topic. Parfit argues that personal identity is not some “further fact” over and above the facts about overlapping chains of psychological connectedness between different “person stages.” The intended contrast is with a dualist in the philosophy of mind, who maintains that there is some such further fact. More specifically, Parfit argues that personal identity is a matter of (a) the holding of “relation R,” where relation R is psychological connectedness and/or psychological continuity, with “the right cause,” where (b) that relation does not take a “branching” form. (Parfit, 1984: 216) For present purposes we may ignore the complications arising from the “with the right cause” condition. What matters is that personal identity is ultimately a matter of psychological connectedness (very roughly, the holding of memories, beliefs, plans, desires, values, etc. over time), and that for personal identity this connectedness cannot take a branching form. Parfit argues convincingly that relation R cannot take a branching form if personal identity is to be preserved by considering “fission” cases. In these cases, my brain is split and put into two new bodies. By hypothesis, this transfer preserves enough psychological connectedness in both “Lefty” and Righty” for relation R to hold between me before the operation and Lefty as well as between me after the operation and Righty. However, I cannot be identical to both Lefty and Righty, since Lefty and Righty are two different people and identity is transitive. If it is not obvious that Lefty and Righty are different people, just imagine that each of them goes on to live a very different life from the other. Each develops his own personality, values, relationships and has his own new memories. Moreover, if Lefty and Righty were the same person then we would be disposed to punish both of them for the crimes of either, but this would of course be a gross injustice. Righty should not go to jail for Lefty’s crimes. Since Lefty and Righty are distinct persons, I cannot be identical to both of them. At the same time, it would be totally arbitrary to choose either Lefty or Righty as being “me” as opposed to the other. Therefore, I do not survive fission, and this explains why our concept of personal identity must require that relation R takes a non-branching form.

Parfit argues, convincingly in my view, that although we do not literally survive in fission cases, we should not regard fission as being as bad as death. After all, if only the left hemisphere of my brain survived and were transferred to a new body, we would happily consider that a form of survival, assuming enough psychological connectedness held. However, it would be weird to think that the value of Lefty’s survival is diminished by the survival of Righty. As Parfit memorably puts the point, why should we consider a double success a failure? For this reason, Parfit argues that what really “matters” in personal survival is simply the holding of relation R, no matter whether it takes a branching form or not. Indeed, Parfit argues that we might well regard fission as better than ordinary survival. The number of years lived would be doubled, just for a start. Further, suppose I am passionate about two careers, and torn between them, and with only one life I could not reasonably expect to carry out both successfully. If I undergo fission, then each of the resulting people could pursue one of these careers. Indeed, each of them could take pride in the other, too, just as we can take pride in our children’s achievements so too could they take pride in each other’s achievements given their special relationship (Parfit, 1984: 264).

Why does any of this matter to whether concept of welfare is a normative concept? The hypothesis under consideration is that welfare is normative in that it entails prudential reasons—that an action would promote your welfare is a reason to perform that action. On Parfit’s view, though, it isn’t really that it is your welfare that matters—it is, rather, that it is the welfare of a person who is R-related to you. In normal, real-world cases, these do not tend to come apart, but in fission cases they do. Consider the following thought experiment. I have a condition that will kill me in six months. I can either live out my remaining days with a rather minimal quality of life—just enough, by my lights, to make life worth living but no more. Alternatively, I could have my brain split immediately, keeping my consciousness largely intact in each hemisphere (due to sufficient redundancy) and have each half implanted in a younger, healthier body. I judge that each of these two (Lefty and Righty) would then have long lives with enormous amounts of welfare. However, this is a unique opportunity. I can only have the surgery right now; I cannot put it off until closer to my death. I correctly infer, on Parfitian grounds, that neither of these two people will be me; the fission will effectively be my death. Which action will maximize my welfare? Plausibly, not having the surgery and living out my six months naturally in my own body. That way, I get six more months of life with a positive balance (barely) of additional welfare throughout that period.Footnote 6 Whereas if I have the surgery, I have no more welfare. I therefore correctly judge that not having the surgery would maximize my welfare. Now suppose welfare is a normative concept, in that judging that an option would maximize your welfare entails a prudential requirement to choose that option, which in turn would entail you ought to do it if there are no other reasons in play to outweigh it anyway. From the fact that not having the surgery would maximize my welfare, it follows that I am under a prudential requirement not to have the surgery, where prudence is genuinely normative. This simply does not follow, though. I might quite reasonably agree with Parfit that what matters in survival is relation R and thus choose to have undergo the surgery. This need not involve any conceptual incompetence. A natural corollary of Parfit’s view is I should be especially concerned about the welfare of those who stand in relation R to me now. On that view, there are not really any prudential reasons at all, since prudence wrongly ascribes significance to whether some future person stage will be me, rather than whether that person stage is R-related to me-now. Granted, prudence is an excellent heuristic in the real world, since fission is entirely science fiction and in the real world whether it is me and whether it is R-related to me now travel in lockstep. Nonetheless, an excellent heuristic is not a fundamental conceptual truth.

Moreover, Parfit’s views supply a second reason to deny that the concept of welfare entails prudential reasons, one that Parfit himself discusses at length. Plausibly, prudence entails that we should care equally about our welfare no matter when it occurs; it is temporally impartial. The mere fact that some welfare will occur later in my life does not in itself diminish its normative force. However, if what really matters is relation R, then, Parfit argues we should discount the welfare of future stages of ourselves insofar as we are less psychologically connected to them. Temporal differences as such still don’t matter on this view, but degrees of psychological connectedness do, and these can make it rational to care far less about the welfare of my future self than prudence would dictate. Here is Parfit:

My concern for my future may correspond to the degree of connectedness between me now and myself in the future…it can be rational to care less, when one of the grounds for caring will hold to a lesser degree. Since connectedness is nearly always weaker over longer periods, I can rationally care less about my further future. (Parfit, 1984: 313)

As Parfit puts it, we have a discount rate, but not with respect to time itself, but with respect to the weakening of the relation which really matters. J.L. Mackie, having (like Parfit) defended a broadly Lockean view of personal identity draws a similar conclusion:

Even the rationality of prudence—in the sense of equal concern for the interests and welfare at all future times of this same person, oneself—is not quite as self-evident as is commonly supposed. Personal identity is not absolute...our concept of personal identity through time itself functions as a sort of institution, aided by a contingent present desire for our own future welfare.” (Mackie, 1976: 191)


Mackie later adds:

...that there is no factual basis for the employment of our present absolute concept…has an important bearing on moral philosophy, especially in bringing it about that there is no exclusive rationality in having an equal concern for all of one’s future selves. (Mackie, 1976: 199)

How might one object to these Parfit-style arguments? In the case of the argument from fission cases, one might object that there will typically be contingent reasons not to go in for fission. It would, for example, often wreak havoc on intimate relationships, just for a start. But these will be contingent, and if the details of the story are spelled out in the right way, there may be no such reasons. Perhaps I am a loner with no real relationships. One might instead object that the mere fact that I am under a prudential obligation to not have the surgery does not mean that I do not have other non-prudential reasons to opt for fission. After all, there are still altruistic reasons, and those reasons might favor fission, which will bring two new happy people into the world. Indeed, we could even hold that because these people would be R-related to me now, this gives me special reason to bring them into the world, in addition to reasons of impartial altruism. This, though, is also not the right way to think about the case if Parfit is right. This makes the case seem like a dilemma, akin to deciding whether to have children who you know will be very happy but when you also know that to have them you will die in childbirth rather than having a few more months of life instead. Given Parfit’s view, the case should not seem like a dilemma. Instead, it seems like whatever truth there ever was in prudential theories of reason for action is exhaustively captured by the reasons provided by the welfare of those who stand in relation R to me-now.

This suggests another objection. The preceding argument puts pressure only on a very specific conception of the normativity of the concept of welfare—a prudential one. That this prudential conception is not contained in the concept of welfare does not prove that welfare is not a normative concept in some other sense. Perhaps welfare is normative, but not in virtue of conceptually entailing prudential reasons. Perhaps it is instead normative in virtue of conceptually entailing reasons for me-now to promote the welfare of those who stand in relation R to me-now. This, though, seems extremely implausible. Our folk concept of welfare evolved long before Parfit wrote Reasons and Persons, a book which (sadly!) most people have still not read. Relation R is a fairly complex, technical concept that has no obvious purchase on our ordinary concepts. Indeed, the influence of theological conceptions on the folk suggest that if anything the folk conception of personal identity presupposes that it is a “further fact,” perhaps a fact about the persistence of a soul or Cartesian mind. Given those widespread background beliefs, and given how subtle Parfit’s relation-R is, the idea that the concept of welfare encodes reasons of this sort seems very unlikely.

Of course, this does not exhaust the possibilities. Perhaps the concept of welfare encodes purely agent-neutral reasons, so that judging that an action would increase anyone’s welfare entails altruistic reasons to perform that action. This is much more plausibly part of our folk wisdom, and so less of a “stretch” to see as contained in the pre-theoretical concept of welfare than anything as esoteric and reductive as agent-relative reasons grounded in Parfit’s relation R. However, this also does not seem plausibly to be a conceptual truth. Someone who finds rational egoism plausible does not, on the face of it, seem to be making a conceptual mistake, though numerous philosophers have boldly tried to argue that they are. What is plausible in all of this is the substantive view that welfare is a normatively relevant property, in much the way that having made a promise or causing pain is normatively relevant. In both cases, the concept itself is descriptive but it picks out a property which can ground normative reasons for action. Indeed, this point leads directly into my second objection to the normativity of the concept of welfare.

My second objection to the normativity of welfare discourse is effectively a version of G.E. Moore’s famous (some would say “notorious”) Open Question Argument. Moore argued that goodness was not reducible to any concatenation of natural properties by arguing that the concept of goodness is not reducible to any concatenation of naturalistic concepts. A major problem with this form of argument was its leap from a conclusion about concepts to a further conclusion about metaphysics. The concept of water is not conceptually reducible to the concept of H20, yet water is H20, and this point has not been lost on modern metaethics which has moved on in ways it would be distracting to summarize here. In any event, my intended conclusion is entirely conceptual, so the worry about moving from a purely conceptual thesis to a metaphysical one will not arise here.

Moore argued that for any concatenation of naturalistic concepts N, it will always be a conceptually “open question” whether something that is N is good. Being a conceptually open question is simply being such that wondering whether the entailment from ‘is N’ to ‘is good’ holds without betraying incompetence with the concept of goodness. Simply asserting that the relevant questions are conceptually open might seem question-begging. Sympathetic commentators have suggested a more charitable reading of Moore’s argument is as an argument to the best explanation, rather than a simple deductive inference. On this reading, the argument begins with the premise that competent speakers tend to find the relevant question conceptually open. This contrasts with cases in which we have a clear conceptual truth. Competent speakers do not find e.g. “I know it is a closed 2-dimensional figure with 3 sides, but is it a triangle?” open. The argument then asserts that the best explanation of speaker’s consistently finding the relevant questions open is simply that they are conceptually open, from which it follows that the concept of goodness is not reducible to the relevant naturalistic properties. Of course, the argument cannot canvass all the logically possible naturalistic analyses, but by showing how compelling the argument is for the wide range of analyses people have offered one is given reason to think that no such analysis will evade the argument.

Moore was trying to show that an evaluative concept (the concept of goodness) was not reducible to any concatenation of naturalistic concepts. My aim, by contrast, is to show that a concept is not normative. Rather than showing a seemingly evaluative concept like the concept of goodness is not reducible to anything naturalistic, my aim is almost the opposite one of showing a seemingly naturalistic or descriptive concept (the concept of welfare) is not, in fact, actually a normative concept in disguise. A complication here is that I do think the best analysis of ‘welfare’ will include some normative concepts, so in that very broad sense I allow that it is a normative concept. However, as should become apparent (in Sect. 3), on this analysis ‘welfare’ is not a normative concept in the standard sense of functioning to settle the thing to do, think or feel—not even modulo there being no competing normative reasons (e.g. moral reasons) in play. On my analysis, judgments about someone’s welfare do not entail any substantive conclusions about normative reasons of any kind. Indeed, on my analysis judgments about welfare are compatible with global nihilism, the view that there are no normative reasons whatsoever, which is surely a good litmus test for a concept’s normativity.

My primary target is the view that ‘welfare’ is a normative concept in the sense that judging that a possible action of a given agent would increase that agent’s welfare conceptually entails that there is prudential reason for that agent to perform that action—and that it would maximize her welfare conceptually entails that she prudentially ought to perform it. My suggestion is that competent speakers can, without betraying conceptual confusion, find questions like “I know it would promote my welfare, but do I have prudential reason to do it?” to be open, and that the best explanation of this is that the relevant questions are, in fact, open. The discussion of Parfit provides one natural segue into seeing how these questions can seem open. Someone competent with the relevant concepts might have the following sorts of monologues, after all:

“I know investing in a mutual fund would maximize my long-term welfare, but does that give me any prudential reason to invest in a mutual fund? The person who would cash in on that investment will likely be so different from me as I am now—will have so little psychological connectedness to me as I am now—that I don’t see why I should particularly care how that person does, as opposed to a total stranger or distant relative.”

“I know that living another six mediocre months instead of having the fission procedure tomorrow would maximize my welfare, but why should I care at all about that? I agree with Parfit—all that really matters in survival is relation R anyway, whether branching or non-branching, and that relation would be sustained through fission. Parfit is right—prudence itself is bunk—there are no prudential reasons, strictly speaking.”


Insofar as these kinds of thoughts do not betray conceptual confusion, the thesis that welfare judgments conceptually entail normative prudential judgments is in trouble. Another way of putting the point is simply that the normative views Parfit defends do not seem conceptually confused, yet he trades in talk of welfare and rejects prudential normativity on principled grounds—grounds many readers have, after all, found not only right but profoundly right. It seems heavy-handed to dismiss what some consider the most powerful lessons of Reasons and Persons as conceptually confused.

We do not have to rely entirely on the coherence of Parfit’s views to make the point, though. A commitment to a purely agent-neutral altruism also seems compatible with competence with the concept of welfare, but this also entails the rejection of purely prudential reasons. The following inner monologue also seems entirely coherent:

Granted, investing my extra disposable income in this mutual fund would best promote my welfare, but why am I especially important? I could instead donate that money to charity and better promote lots of other people’s welfare. I was moved by Jesus’s injunction to treat thy neighbor as thyself, and I also thought Thomas Nagel’s The Possibility of Altruism made a strong case that all reasons are impersonal. I don’t see any reason to think my welfare gives me any more reason than the welfare of a total stranger. This includes not only reasons for action, but reasons for attitudes. Because my welfare is no more important than anyone else’s, I don’t have any reason to prefer outcomes in which I do better than others, not even as a tie-breaking consideration (when aggregate welfare is equal between the two outcomes).


Again, this monologue does not seem to betray conceptual confusion. A purely impersonal altruism, which in effect rejects the idea of specifically prudential reasons in this way does not seem to betray conceptual confusion. Of course, the defender of the normativity of the concept of welfare might now switch gears, insisting now that welfare judgments conceptually entail altruistic reasons. However, not only would this abandon the link between welfare and prudential normativity, it is also vulnerable to an objection that is the flip-side of this one. Plausibly, the conceptual coherence of rational egoism, which categorically rejects altruistic reasons, undermines the conceptual entailment from welfare judgments to judgments about agent-neutral altruistic reasons too. It is unclear how you could plausibly argue that welfare encodes one, but not the other of these conceptions of normativity, but equally it is unclear how you could argue that it encodes both since purely rational egoism and pure altruism also seem coherent. The idea that ‘welfare’ is normative entails that it is normative in one of these more specific ways, but privileging any one of them (pure altruism, pure egoism, or a hybrid) also seems arbitrary.

There is yet another riff on the Moorean argument. The apparent conceptual coherence of nihilism about practical reasons also undermines the thesis that welfare is normative. Again, the following inner monologue seems conceptually in good order:

Granted investing in this mutual fund would maximize my welfare, but why does that matter? Why does anything matter? Mackie is right, normativity relies on a kind of non-natural Platonic property that simply does not and could not exist. There are no reasons for action, and so obviously no prudential reasons for action either.

Once again, this seems conceptually OK. Note, by contrast, that the following inner monologue does not seem coherent, thus corroborating my rejection of the “package deal”:

Granted investing in this mutual fund is what I prudentially ought to do, but why does that matter? Why does anything matter? Mackie is right, normativity relies on a kind of non-natural Platonic property that simply does not and could not exist. There are no reasons for action, and so obviously no prudential reasons for action either.

The commitment to a prudential ‘ought’ already “buys into” normativity, and so going on to ask whether there are any prudential reasons would indeed betray conceptual confusion, whereas simply making judgments about what would promote one’s welfare does not “buy into” normativity in this way. Returning to a loose end from Sect. 1 in relation to Fletcher’s five “marks of the normative,” this is why welfare judgments do not carry “authority” in the needed sense. One can coherently make judgments about what would promote one’s welfare while endorsing nihilism about practical reasons (and, indeed, reasons for attitudes, though I have said less about that here).

Obviously more could be said about both these arguments. The first argument relies on Parfit’s view of personal identity, which of course is not uncontroversial, but I cannot argue at length for that here—and anyway that view has already been extensively discussed in the literature so I would be reinventing the wheel if I did. If the first argument simply establishes that defending the normativity of the concept of welfare requires rejecting Parfit’s view of personal identity, that would itself be interesting, though. The existing literature on whether welfare is normative does not typically engage with Parfit’s views, but if I am right then they pose a direct challenge to the thesis that welfare is a normative concept.

The second argument does not depend on the truth of Parfit’s views, but only on their conceptual coherence (and the coherence of various other normative views), so in that sense it is less tendentious. However, the second argument is a form of Moore’s Open Question Argument, which is itself controversial. The fact that I am using the argument only to establish a conceptual conclusion, as opposed to a metaphysical one, at least avoids some of the more powerful objections to Moore’s own use of this form of argument. However, there are of course many other important objections to the Open Question Argument. A discussion of these would also be out of place here, though; that is another debate altogether, and a well-worn one at this stage. If I have made it plausible here that if the form of argument is not in principle hopeless then it has substantial purchase here, that too will be progress.

However, none of this is likely to be convincing without at least the beginnings of a positive account of the concept of welfare. Moreover, I have still not addressed what I consider the best argument that the concept of welfare is a normative one—namely the argument from disagreement I mentioned but put to one side in Sect. 1. In the following section, I try to remedy both these defects by developing a positive account of the concept of welfare that can explain why it is amenable to widespread fundamental disagreement while at the same time denying that the concept of welfare is a normative one in the intended sense.

3 Welfare: A conditionally normative concept

One plausible way of understanding the content of a concept is in terms of its inferential role. This approach can be developed in a broadly inferentialist theory of concepts, or in the spirit of the so-called “Canberra plan,” where a list of platitudes associated with a given concept is drawn up and these platitudes are used to fix the extension of the concept via so-called “Ramsey-Lewis” sentences (Cf. Jackson, 1998). The account I develop here could be combined with either of these approaches, or indeed any approach which characterizes concepts in terms of inferential roles or platitudes in the first instance.

In terms of the Canberra plan, what are the platitudes associated with the concept of welfare? Some of them are surely substantive theses about what constitutes welfare, at least for creatures relevantly like us. For example, the following are plausible candidate platitudes:

Ceteris paribus, more pleasure increases a subject’s welfare.

Ceteris paribus, more pain decreases a subject’s welfare.

Other platitudes might address the scope of the concept of welfare—that is the kinds of entities capable of having welfare in the first place. One such candidate would be:

S: Only sentient beings can have welfare—they are the only welfare subjects.

So far, these platitudes need not have any normative content. However, we could add the following to our list of platitudes:

PR: If a possible action of A’s would increase A’s welfare, then A has prudential reason to perform that action.

PO: If a possible action of A’s would maximize A’s welfare, then A prudentially ought to perform that action.

An important advantage of including these in our list of platitudes is that doing so helps explain the ubiquity of deep disagreement about welfare. At least, modulo the assumption that some account of why normative concepts in general engender ubiquitous and deep disagreement is in the offing. However insofar as these are platitudes, the concept of welfare is, after all, a straightforwardly normative concept, a thesis I am arguing against. I therefore need to reject the thesis that PR and PO are genuine platitudes.

Before I propose my alternative, it will be useful to review another debate, which to my mind is very suggestive of the approach I favor for the concept of welfare. In their work on what they call the “moral fixed points,” Terence Cuneo and Russ Shafer-Landau argue that we should understand the content of our moral concepts in terms of certain “fixed points” (Cuneo and Shafer-Landau 2014). These fixed points are meant to be conceptual truths. Among the fixed points are theses like the following:

RS: It is wrong to engage in the recreational slaughter of a fellow person.

An implication of this account is that error theorists who think nothing is morally wrong are thereby “conceptually deficient,” since they are committed to rejecting “fixed points” like RS. If nothing is wrong, then ipso facto recreational slaughter is not wrong. Error theorists have, very plausibly in my view, replied that it is heavy-handed to dismiss them as conceptually deficient on these grounds. In particular, they have argued that RS is too strong to be one of the platitudes. Instead, whatever is plausible in treating RS as a platitude is better captured with the following conditional platitude:

RS*: If anything is wrong, the recreational slaughter of a fellow person is wrong. (Evers & Streumer, 2016: 4)

Treating this as a platitude would not entail that error theorists are conceptually deficient, for they may be as adept as anyone in recognizing what any recognizable morality would have to include among its prescriptions. It would also encode the idea that there are conceptual limitations on the kinds of reasons that can count as moral. It would do so without making the truth of moral realism a sort of trifling conceptual truth like ‘all bachelors are married’. Evers and Streumer usefully compare their debate over the moral platitudes with an analogous debate one might have with a Christian who holds that it is a “fixed point” that God rewards benevolence. This would entail that the atheist, who denies that God exists, is conceptually deficient. Again, this seems heavy handed, and we should instead include at most a conditional platitude such as “If anything is rewarded by God, benevolence is rewarded by God.”

Now return to the concept of welfare. In my view, including PR and PO among the platitudes associated with the concept of welfare is akin to including RS among the platitudes associated with the concept of moral wrongness. Here too, we can and should instead go for conditional analogues of those platitudes:

PR*: If there are any prudential reasons, then if a possible action of A’s would increase A’s welfare, then A has prudential reason to perform that action.

PO*: If there are any prudential ought’s, then if a possible action of A’s would maximize A’s welfare then A prudentially ought to perform that action.


Including these weaker platitudes instead of PR and PO would still preserve a kind of link between welfare and prudential normativity. It would also allow a nihilist to be competent with the concept of welfare and correctly judge that certain things are conducive to a person’s welfare. They would not thereby be committed to the truth of any prudentially normative ‘ought’ claims because they could coherently deny the antecedents of PR* and PO*—that is, they could deny that there are any prudential reasons or prudential ought’s. Similarly, someone convinced by Parfit that relation R is what matters and that this discredits prudence, which puts too much weight on personal identity over time, could deny the antecedents of PR* and PO*. And someone who thinks all reasons are agent-neutral could for that reason deny the antecedents of PR* and PO*. Such people could all remain competent with the concept of welfare.

In more general terms, PR* and PO* can accommodate the possibility of deep disagreement about the nature of welfare. Such disagreement can reflect the deep normative disagreement over what it would take for there to be prudential reasons or true prudential ‘ought’ claims. Insofar as some explanation of the possibility of such disagreement is in the offing from whatever the best meta-normative theory tells us, that explanation can be applied here in this way. Similarly, whatever the best explanation of why we do not accept testimony about the normative can explain why we wouldn’t accept testimony about the nature of welfare. Even someone who rejects the category of prudential normativity could have views about what it would take for there to be such normativity, and would not revise their views simply on the strength of the testimony of someone else.

It might be objected that someone who admitted that an action would maximize his long-term welfare but denied that he prudentially ought to do it would thereby betray a kind of conceptual incompetence—incompetence with the concept of prudential ought. This is true in a sense, but one which does not undermine the theory on offer here. Prudence is a “thick” normative concept, and as with many such concepts it is possible to use the same form of words to advert to the associated descriptive content without endorsing the relevant norms. Hume, for example, famously referred to modesty as a “monkish” virtue and heaped scorn upon it, arguing that it was no virtue at all. Similarly, someone who rejects the normativity of prudence might competently use ‘prudentially ought’ to advert to whatever maximizes welfare without endorsing the normativity of prudence. It is only in this non-normative sense of ‘prudentially ought’ that someone who denies that an action which maximizes an agent’s welfare is one the agent prudentially ought to perform thereby betrays conceptual incompetence. Since this is not the sense of ‘prudentially ought’ that is relevant to whether the concept of welfare is a normative one, this does not undermine the theory on offer. Parfit, in fact, makes the same point by comparing prudence with chastity:

Consider the claim that someone is unchaste. Many now believe that there is nothing morally wrong in unchastity. And, for these people, the charge ‘unchaste’ ceases to be a criticism. A similar claim applies to ‘imprudent’…If we believe that an imprudent act is not irrational, the charge ‘imprudent’ might, for some of us, cease to be a criticism. It might become, like ‘unchaste’, merely a description. (Parfit, 1984: 318)


Understanding the concept of welfare as only conditionally normative in this way has a lot to recommend it. For a start, this approach can explain why there is pervasive and deep disagreement about the nature of welfare just as well as accounts on which the concept of welfare is a normative one. On the one hand, for those who treat prudence as normative, and so endorse the antecedents of PR* and PO*, debates about the nature of welfare will de facto just be normative debates. Moreover, the account can explain why those who reject the normativity of prudence can usefully take part in these debates. For even an error theorist can have views on what conception of welfare would figure in the best theory of prudence. This would be analogous to the atheist having views about the most plausible reading of the morality of the New Testament, e.g. This, in turn, also means that any otherwise plausible explanation of why it would be weird/problematic to defer entirely to others about normative matters like morality and prudence can carry over as an explanation of why it would be weird to defer in this way in determining the nature of welfare. For even those who reject the normativity of prudence will need to deploy their concept of prudence (if only implicitly) to see what welfare would have to be like for it to be such that if there were any prudential normativity then welfare would provide its most apt “target.” So insofar as we can explain the weirdness of deferring entirely to others on normative matters in general, the explanation can be carried over to deference about welfare.

Indeed, we can go further in developing a view of this sort. There is no reason to restrict the relevant platitudes to ones about prudence. It is equally plausible that an analogous conditional with respect to altruistic reasons could be incorporated into the meaning of ‘welfare’:

AR*: If there are any altruistic reasons, then if a possible action of A’s would increase someone’s welfare, then A has altruistic reason to perform that action.

It is not, however, clear whether we should incorporate an altruistic analogue of PO into the meaning of ‘welfare’. The obvious conditional would be a broadly consequentialist one, according to which if there are any altruistic reasons then if an action would maximize aggregate welfare, then one altruistically ought to perform it. However, it is not clear that altruism must take a maximizing form, as a matter of the bare concept of altruism. Perhaps altruism would give special weight to the needs of the worst off, for example. I here remain neutral on what the best version of the conditional normativity view is, and whether it should include some such altruistic conditional normativity platitude and if so what it should be. To put my cards on the table, I think some such altruistic conditional normativity platitude should be included in the best theory in this vicinity, but I am less sure precisely what form that platitude should take. However, my aim in this paper is only to defend the general framework. I defer a more detailed exposition of the best version of such a theory for another occasion.

An attraction of this conditional normativity approach is that it explains how the concept of welfare gets its life from its role in norms which have been ubiquitous in all human society—norms of altruism and prudence. This can explain why the idea that the concept of welfare is a normative one can seem very plausible. In particular, it can explain the ubiquity of deep disagreement, and indeed normative disagreement, as to the nature of welfare. However, by making the conceptual connection conditional in this way, the theory can avoid a commitment to the thesis that welfare is an unconditionally normative concept—that is, one which directly functions to settle the thing to do. The concept does not always play this role at all; nihilists can use the concept of welfare as articulated here. When the concept does play this role, it does so only indirectly—the role is mediated by substantive judgments about what one prudentially or morally ought to do—judgments not mandated by mere competence with the concept of welfare itself.

This functional distinction notwithstanding, one might object that in one clear sense of ‘normative’ it is sufficient for a concept to be normative that its analysis includes normative concepts. Since my proposed analysis of ‘welfare’ does include normative concepts, in what sense is it not normative? On the proposed account, the concept of welfare is normative in one sense, but not in another. However, the sense in which it is normative is weaker than the sense in which e.g. the concept of the all things considered ought or the concept of moral duty is normative. Here it is useful to distinguish strong and weak senses in which a concept can be normative:

  1. (1)

    A concept C is strongly normative if the direct application of that concept to an action (or feeling or belief) settles the thing to do (or feel or think) or defeasibly settles the thing to do (or feel or think), modulo the absence of competing reasons/values.

  2. (2)

    A concept C is weakly normative if the analysis of the concept ineliminably includes strongly normative concepts.


Strong normativity is unconditional; strongly normative concepts directly function to settle the thing to do (or feel or think). By contrast, weak normativity can be unconditional; weakly normative concepts can be such that their deployment does not directly function to settle the thing to do (or think or feel). If conceptual analysis is understood broadly enough to include just listing "the platitudes" then the conditional normativity of the concept of welfare will mean it is weakly normative but not strongly normative. It isn't strongly normative since 'Doing this would enhance your welfare' [or even 'massively boost your welfare'] does not settle that you should do it, even modulo the absence of any competing reasons/values. After all, accepting this is compatible with global nihilism (an error theory like Mackie’s famous one), on my interpretation of the concept of welfare.

It is, in a way, this—the compatibility of the first-order truths about welfare with global nihilism that makes it not normative in an interesting sense that distinguishes it in a theoretically interesting way from the concept of the all-things-considered 'ought’, since the latter trivially is incompatible with global nihilism. This distinction will also (plausibly in my view) make first-order moral judgments come out as strongly normative, again preserving the needed contrast between moral claims and claims about welfare. That is because the truth of 'Stealing is morally wrong' entails that one has reason not to steal—and not just that one has such reason if some further conditions are met. If an error theory about the normative is true then 'stealing is morally wrong' is not true, on my view, anyway. The point, then, is that the concept of prudentially ought is strongly normative, but the concept of welfare is only weakly normative.

Reflection on the error theory is useful here in another way in this context. The theory developed here sheds light on a novel “companions in guilt” argument Fletcher has recently and independently developed against the moral error theory. Fletcher argues as follows:

(1) Prudential Parity premise. If the arguments for the moral error theory are sufficient to establish that the moral error theory is true, then those arguments (or appropriate analogues of them) are sufficient to establish that the prudential error theory is true.

(2) Prudential existence premise. The prudential error theory is false.

So, (3) the arguments for the moral error theory are not sufficient to establish that the moral error theory is true. (Fletcher, 2018: 478)

The force of the argument derives from the overwhelming plausibility of the idea that some outcomes are bad for welfare subjects like us. In particular, the idea that a life of intense, overwhelming pain is a life with a lower level of welfare than a life filled with ecstatic pleasure seems undeniable. Yet, Fletcher argues, the prudential error theory entails this. Insofar as the arguments (e.g. arguments from disagreement) for the moral theory transpose smoothly to the prudential case, as Fletcher argues, this is bad news for the moral error theory.

It may be obvious by now how the present account, if sound, disarms this objection. The force of Fletcher’s argument relies on reading “prudential error theory” as including within its scope claims about welfare. If, however, welfare is only a conditionally normative concept, then the arguments for the moral error theory do not carry over as arguments for the prudential error theory so understood, simply because welfare is not a normative concept tout court—it is only conditionally normative. Thus, in the sense of ‘prudential error theory’ which makes (2) overwhelmingly plausible, premise (1) is false. We can, of course, salvage premise (1) by reading ‘prudential error theory’ as my preferred terminological choices would suggest, so that it does not include within its scope simple propositions about levels of welfare, but only claims about what agents prudentially ought to do, prudentially must do, have prudential reason to do. On that reading, though, premise (2) is far less compelling, also for reasons discussed here. Anyone convinced by Parfit, e.g., will reject prudence as a bankrupt normative framework, and the perspective offered by his conception of personal identity provides powerful warrant for that rejection. Similarly, anyone convinced by Nagel (e.g.) that all reasons are agent-neutral will also reject specifically prudential reasons but need not deny that certain outcomes are good or bad for people in terms of their welfare. In any event, the rejection of prudential ought propositions (etc.) is far less obviously hopeless than the thesis that pain is not bad for people, etc., in which case the thesis that welfare is only conditionally normative at least seriously reduces the force of Fletcher’s argument.

4 Conclusion

I have in this essay argued that while the concept of prudential reasons for action and the concept of a prudential ought are robustly normative, the concept of welfare is not normative in this way. The view that the concept of welfare is robustly normative is incompatible with certain very plausible ideas about personal identity developed by Parfit and runs into several powerful versions of the “Open Question Argument.” However, we can preserve what is plausible about the idea that the concept of welfare is normative by understanding its content, in part, in terms of certain conditionals—conditionals such as, e.g., if there are any prudential reasons at all then that an action would promote the agent’s welfare provide the agent with a prudential reason to perform the action. The concept of welfare then, is not normative if being normative means functioning directly to settle the thing to do, as the concepts of prudential ought and the concepts of morally ought and altruistically ought to do. The concept of welfare does, though, still get its life from its role in these systems of norms, and it would be impossible to be fully competent with the concept of welfare without being competent with those normative concepts. The concept of welfare is, then, not normative full-stop, but is in this sense conditionally normative. This helps disarm an otherwise convincing challenge to the moral error theory. It is also an important finding it is own right, enriching our understanding of the concept of welfare.