In the Preface to this enormous work, Derek Parfit makes a reference to Sidgwick's ‘great, drab book’ The Methods of Ethics (xxxiii).Footnote ¹ The juxtaposition of adjectives is amusing and oddly endearing; for Parfit, Sidgwick's drabness is not in any serious competition with his greatness. Parfit's two-volume book is not drab – though it can be exhausting – and, although it is probably too early to tell for sure, it also shows certain signs of greatness. Time and time again, one is impressed, even moved, by Parfit's enthralling and indefatigable resourcefulness and level of immersion in these issues, and by his dogged and transparently sincere commitment to discovering the truth about them. Despite its length, On What Matters demands careful study by everyone with a serious interest in philosophical ethics.

Given the size and complexity of Parfit's book, not everything can be adequately discussed, even within the compass of a critical review. Accordingly, I propose to focus on certain aspects of Parfit's treatment of practical reason, metaethics and normativity, which occupies Parts One and Six. The reconciling project in normative ethics which Parfit advances in other parts of the book will not be discussed here. The additional material on Nietzsche, cosmology, modal ontology and sundry other matters which also finds a home in these pages will also go unexamined.

As a matter of fact, I cannot even provide a comprehensive treatment of Parts One and Six. To permit a decent level of critical detail to develop, I will examine just two topics which run through the thread of Parfit's argument in these parts of On What Matters: Parfit's dismissal of Subjectivism about reasons, and his principal reasons for thinking that Non-Naturalist Cognitivism is the only satisfactory metaethical theory.Footnote ² I will discuss these topics in that order.

I

For Parfit, a reason for action is a fact or consideration that counts in favour of action. That is not much of an elucidation, but we should not expect successful reductive analyses for so primitive a concept. The notion of a reason is utterly central to Parfit's treatment of metaethics and normativity. He favours an Objectivist, or a value-based, object-given theory of reasons for action, which holds that ‘reasons for acting all derive their force from the facts that give us reasons to have certain desires and aims’ (1, 47). Objectivist views contrast with Subjectivist views, which are desire-based and subject-given, and which hold that ‘our reasons for acting are all provided by, or depend upon, certain facts about what would fulfil or achieve our present desires or aims’ (1, 45).

This way of stating the contrast, with its embedded distinction between object-given and subject-given reasons, leads to a strained characterization of Subjectivism: focusing for present purposes on only non-instrumental desires, Parfit imputes to Subjectivists a dependence on the ‘state-given’ reason that our wanting an event would be in itself good, whereas Objectivists are allowed to help themselves to the ‘object-given’ reason that the event that we want would be in itself good, or worth achieving (1, 50). But this seems uncharitable. Subjectivists need not say that our wanting an event to happen is itself good, as though agents are only interested, when their desires are fulfilled, in the mere fact that their desires have been fulfilled, as opposed to the worldly facts which make it the case that those desires have been fulfilled. Subjectivists can surely acknowledge that what we principally take an interest in are the worldly properties of the event which is desired, but that what ultimately grounds the value of those properties is our wanting them. So when Parfit complains that ‘If Subjectivism is true, we must make our choices in the dark’ (1, 46), Subjectivists might retort that it is Parfit who has turned out the lights, not they.

The simplest form of Subjectivism is given by the Desire-Based Theory, which holds that we have a reason to do whatever would fulfil any of our present desires (1, 58). Of course, Subjectivists are aware of the legion of counterexamples to forms of desire-based theory which fail to eliminate error, since we can be cruelly misled by ignorance and false beliefs. (This territory is notoriously thick with petrol-drinkers, grass-counters, psychotically embittered squash players, and so on.) But even tidied-up versions of desire-based theories are apparently vulnerable to the Agony Argument, which Parfit outlines as follows:

Subjectivism is vulnerable to this counterexample because it tenders out our reasons to whatever agents desire to do, just as long as they have satisfied the error-expunging procedural constraints imposed on them by the more refined versions of Subjectivism (1, 62). Given Subjectivism's failure to exercise quality control on the choice of ends, it is conceivable that a fully informed agent would elect to suffer agony. Furthermore, since it is only this sort of unusual case which can properly distinguish between Subjectivism and Objectivism, given most actual people's interest in their well-being, the case is highly significant; it cannot be dismissed as a far-fetched thought experiment with no application to real human beings. Subjectivism is therefore committed to holding that this particular agent has no reason to spare himself from agony. For Parfit, that constitutes a reductio ad absurdum of Subjectivism in all its versions.

The Agony Argument is designed to cut through the checks and balances installed by the more refined, error-free versions of Subjectivism. But the news for Subjectivists is about to get even worse, because Parfit takes his conclusions to show that Subjectivism is, in any case, unable to take advantage of these refinements. Why is this? In the Incoherence Argument, Parfit claims that Subjectivism is in no position to enter a plea for these additional informational resources, because agents will not be taking reason-giving instruction from the object-given properties which will be disclosed by these information-revealing exercises (1, 93–5). If an agent already has an actual present desire for something – a gin and tonic, for example – then there may be a principled case for making available to her error-free beliefs about the contents of the glass in front of her. Since she wants a gin and tonic, not a petrol and tonic, the accuracy of her beliefs will matter to the fulfilment of that very desire. But there is no principled case that Subjectivists can make for agents’ error-free exposure to environments where their present desires have not yet been determinately formed.

The Incoherence Argument is unfair. The advantage of having an accurate grasp of what these object-given properties are is independent of the fact that, for Objectivists, some of these properties will be directly reason-giving. The Incoherence Argument reflects Parfit's earlier misimpression that Subjectivists are only interested in state-given reasons, rather than in the worldly properties which engage our desires. Because we are desiring creatures who are interested in the world, not just ourselves, the world to which we are exposed is one about which Subjectivists can properly insist we have true beliefs, rather than false beliefs, or no beliefs.

The Agony Argument also seems to me to be in tension with another argument Parfit advances in section 6, where he considers ‘hedonic reasons’. Parfit thinks, perhaps surprisingly, that pain is not intrinsically bad purely in virtue of the way it feels; only the combination of pain with hedonic disliking makes the mental state of being in a state of pain bad (1, 54). Parfit then distinguishes between these hedonic likings and dislikings and the meta-hedonic desires which are desires for the continuation, or cessation, of these mental states: ‘What we dislike is some sensation. What we want is not to be having a sensation that we dislike’ (1, 54).

Parfit's distinction between hedonic likings and dislikings on the one hand, and meta-hedonic desires on the other hand, allows him to dismiss the authority of desires in even this area of agents’ lives, where it might have seemed that desires enjoy a few normative perks. But Parfit's distinction is artificial, because the likings and disliking are plausibly, in part, disguised desires. After all, hedonic likings and dislikings are states which disclose an agent's attitude to the mental states she is in. Since a hedonic disliking is not the very same state as the state of being in pain, and since it therefore cannot consist in a simple registration of the phenomenology or ‘raw feels’ of the pain state, there is nothing else for the disliking to be except attitudinal. A hedonic disliking thus cannot be fully understood in any way which excludes the agent's aversion to being in that state, but then it will be implausible to think that the agent's aversion to being in that state excludes her desire not to be in that state. On Parfit's way of carving up this territory, the meta-hedonic desire will be largely surplus to requirements, because the liking or disliking will already be doing all the reason-giving heavy lifting. It is true that there is no pressing need to recruit distinct meta-hedonic desires into the analysis, but that is because the hedonic disliking is not innocent of association with desire.

Why is the Agony Argument in tension with Parfit's treatment of hedonic reasons? If future agony is accompanied by a future hedonic disliking, then future agony is already covered by a reason-giving thinly disguised desire; and, if future agony is not accompanied by a future hedonic disliking, then, by Parfit's own admission, it is not a bad mental state to be in, and so the Agony Argument cannot refute Subjectivism. Now Parfit argues that it is only meta-hedonic desires, rather than hedonic dislikings, which can be applied to future experiences, since without the actual presence of the agony, there will be no actually existing hedonic disliking to be taken into account (1, 54). But that consideration does not show that it is improper for an informed, error-free Subjectivist account to insist that the hedonic disliking which will be formed at the time that the agent suffers the agony should be taken into account in her present deliberations. (And, if the agent simply does not know whether the agony, when it is inflicted upon her, will be accompanied by a hedonic disliking, then she is not in a position to form a relevantly informed desire, which means that she will lack a reason to acquiesce in the imposition of that agony.)

Parfit mentions, in passing, temporally neutral versions of Subjectivism, and accounts which combine a desire-based theory of well-being with a value-based theory of reasons (1, 74–5). But he is not inclined to dwell on them at any length. Since these represent the sorts of theoretical pigeonholes which most serious Subjectivists will actually be occupying, Parfit's relative neglect of these options seems odd.

II

In Part Six, some of these debates concerning reasons for action are revisited, but this time in a more metaphysical key, and with different aims. Parfit kicks off with an examination of Bernard Williams's Internalist account of reasons.Footnote ³ In these sections Parfit is less concerned with the Subjectivist content of Internalism than with Williams's concept of a reason, which is supposedly revealed in Williams's occasional remarks to the effect that Externalism is, not just false, but unintelligible. Such remarks by Williams tend to strike me either as a bit of mischievous hectoring, or as simply another way of saying that Externalists, in the absence of convincing arguments for their position, are reduced to table-thumping. But Parfit soberly takes these remarks at face value, and subsequently engages with Williams as a representative of Analytical Subjectivism, where it is the analytical part, rather than the subjectivist part, which matters to his critical interrogation. Parfit writes (2, 269):

On Parfit's view, when we say (A), we cannot intend it to mean any claim such as (B) or (C). This is because any analysis of (A) in terms of (B) or (C) will risk a total loss of normativity. Why? The Analytical Subjectivist faces the following dilemma. Either (A) manages to say something informative or substantive about our notion of a reason, or it doesn't. If it does, then purely naturalistic facts such as those reported in (B) and (C) will not do, since any substantive claim about the nature of a reason must involve relating two distinct properties to each other; it must retain the normative property and relate it to some other property. This constitutes the first horn of the dilemma.

Here is the second horn of the dilemma: if (A) does not manage to say something informative or substantive about the concept of a reason, then (A) is merely a concealed tautology, with a content which amounts to one of the following claims:

So Williams is either guilty of changing the subject, or he is saying something completely trivial. Either normativity is irreducible, and requires the existence of irreducibly normative facts and properties, or normativity is just an illusion. This argument is an important basis for Parfit's commitment to Non-Naturalist Cognitivism.

There is a further conclusion to be drawn: Williams and other Analytical Subjectivists actually lack the concept of a normative reason. Parfit expresses the point gently, but clearly (2, 272):

Can this really be true of Williams? Parfit says, near the outset of his argument:

Fair enough. But his criticisms of Analytical Subjectivism clearly commit him to something stronger, along the lines of the following:

This packs a lot into our bare concept of a reason. Why exactly must we accept it?

III

Inspired by minor differences in the metaethical flora and fauna which come to his attention, Parfit makes the same basic argument over and over again, though its essentials are present in the attack on Analytical Subjectivism. But it will do us no harm to consider the greater detail and slightly different emphases which emerge in the later discussion.

There are two objections which Parfit comes heavily to rely upon. The first is the Normativity Objection (2, 327). In the Burning Hotel case, you will die unless you jump from your hotel window into the canal. Naturalists will hold that facts (B) and (C) are the same (2, 326):

But Parfit cannot see how the fact stated in (B) could be the very same fact as that which is stated in (C):

The second objection is the Triviality Objection; but to explain this, we need a little more background. The most plausible forms of naturalism might concede that normative concepts are irreducibly normative, but that the properties to which normative claims refer are natural. This is Non-Analytical Naturalism, which we associate with Allan Gibbard and others.Footnote ⁴ Non-Analytical Naturalism sometimes draws inspiration from various kinds of a posteriori identities established by science: between, for example, ‘water and ‘H₂O’, or between ‘heat’ and ‘molecular kinetic energy’. The meanings of ‘water’ and ‘heat’ will not, by themselves, yield those truths, yet their discovery by science will not tempt us to insist that there is anything to water over and above H₂O, or that there is anything to heat over and above molecular kinetic energy. Water and heat were always, as we might put it, things which were awaiting the empirical discovery of these identities. Couldn't it be the same with, say, rightness, wrongness, and other normative concepts?

Parfit argues not. First, and just to illustrate the difference between the two camps, consider how Non-Analytical Naturalists and Non-Naturalists will characterize Act-Utilitarianism (2, 336). Naturalists will claim that the property of being an act that would maximize happiness is the very same as the property of being what we ought to do. I will call this the Non-Analytical Naturalist Claim. Non-Naturalists, such as Sidgwick, will claim instead that the property of being an act that would maximize happiness is the same as the property that makes an act have the different property of being what we ought to do. I will refer to this as the Non-Naturalist Claim.

The ‘different property’ alluded to in the Non-Naturalist Claim is the distinct, second-order, and irreducibly normative property which the happiness-maximizing property has of being what we ought to do. The difference between the Non-Analytical Naturalist Claim and the Non-Naturalist Claim may strike you as minor, but for Parfit it carries extraordinary significance: the Non-Naturalist Claim allows Act-Utilitarianism to be genuinely normative, whereas the Non-Analytical Naturalist Claim condemns Act-Utilitarianism to triviality. Why is that? Consider the Non-Analytical Naturalist Claim: it purports to be substantive, in the sense that it purports to be significantly informative, and can be disagreed with. Now consider the claim ‘The property of being an act that would maximize happiness is the same as the property of being an act that would maximize happiness’. This claim, by contrast, is true, but trivial: everything is identical with itself. To be non-trivial, the Non-Analytical Naturalist Claim must describe the relation between two distinct properties (2, 352). But the problem with the Non-Analytical Naturalist Claim is that it is a negative claim. It denies that there are two distinct properties in play, and says instead that the property of being an act that would maximize happiness is the very same property of being what we ought to do. So the Non-Analytical Naturalist Claim advances only a concealed tautology.

The argument Parfit uses to displace the Non-Analytical Naturalist Claim will seem familiar. It is, at bottom, the same argument as that which was applied to Analytical Subjectivism. For Parfit, the Triviality Objection is a Non-Naturalist skeleton key; it can unpick every Naturalist lock. But it should be kept in mind that an equally deep objection – perhaps an even deeper one – comes from the Normativity Objection. This is because, for Parfit, the Triviality Objection is concerned with a false, and necessarily false, antecedent. We can put the point this way, using Act-Utilitarianism, once more, as the illustrative example. Act-Utilitarian Naturalists make the two following claims (2, 343):

The Triviality Objection says, in effect, that if (C) were true, then (A) would be trivial. Since (A) is not trivial, (C) is false. But Parfit thinks, in any event, that (C) could not be true. That impossibility must be due to the Normativity Objection, not the Triviality Objection.

IV

Parfit pursues these arguments with prodigious energy, and in microscopic detail. The elements of repetitiveness or déjà vu across chapters 24 to 27 seem to be explained by his determination to leave no stone unturned in his demolition of the various forms of Naturalism. The stakes for him are very high, after all, and the intellectual angst is sometimes palpable: Parfit voices the worry on more than one occasion that if Non-Naturalism is not true, his life's work will have been in vain (2, 304, 367). This is evidently because everything he has had to say about normative ethics and political philosophy will have been reduced, by his lights, to triviality. (As if philosophical bounty-hunters in pursuit of ‘Theory X’ didn't already have enough to worry about.)

But are Parfit's arguments fair? We start with the Triviality Objection. Naturalists are likely to take umbrage at Parfit's inflexibility about the conditions which need to be satisfied if a claim is to avoid triviality. For Parfit, we succeed in making substantive claims only when those claims are understood to exhibit a relationship between (at least) two distinct properties. Unless this ‘distinctness of properties’ condition is in place, Naturalists cannot make a substantive claim, and will be merely peddling tautologies. Parfit seems incapable of interpreting identity claims between normative and natural properties without imputing philosophical larceny to the Naturalist. It is not obvious what grounds this conviction. The claim that rightness is the very same property as the property of maximizing happiness is informative, and can certainly be disagreed with. Its aim is to secure a metaphysical location for rightness, rather than silently consenting to rightness's outright disappearance, or abandonment. The claim, by contrast, that the property of maximizing happiness is the very same property as the property of maximizing happiness is simply an unexciting reminder that every property is self-identical. How can a surprising and contentious identity claim, such as that which is advanced by the Non-Analytical Naturalist, fail to be more informative than a claim about self-identity? How can the former claim be simply a disguised form of the latter claim? Parfit considers such Naturalistic protests in a couple of places, responding to work by Gibbard (2, 346–9) and then Mark Schroeder (2, 358–61). But he does little more, on each occasion, than simply rehash his earlier argument that a substantive claim requires the positive ascription of a distinct property. This is evidently Parfit's bottom line, and one doubts that anything is going to bounce him out of it. But I don't see anything more here than a philosophical stand-off; Naturalists simply don't strike me as being open to the charge of obtuseness which Parfit is levying on them.

Further progress may be possible. In his response to Schroeder, Parfit allows himself to be distracted by Schroeder's Subjectivism, whose implications he is unwilling to accept (2, 361). That should not have been to the point in this particular part of the discussion, since Parfit's business with Subjectivism has been officially concluded. But it inadvertently provides us with some evidence of what might ultimately lie behind Parfit's aversion to Naturalism. To explore these issues further, we need to examine Parfit's reasons for holding the Normativity Objection.

V

Parfit writes:

We do not find Naturalists campaigning for the identity of rivers and sonnets, or for the identity of justice and the number 4. If a similar chasm separates the normative and the natural, it is surely less obvious. What is the gulf supposed to consist in?

Go back to the Burning Hotel case, which was first mentioned in section III. We already know that Parfit refuses to accept that (B), which states a fact about a reason, could be identical with (C), which states a psychological fact about motivation. But is that because of the Naturalism, or because of the Subjectivism? Parfit tends to suppose that the Naturalist must analyse facts about reasons in terms of natural facts – whether actual or dispositional – about motivation. But does this represent the only possibility for Naturalism? On another view, normative facts might be analysed, not in terms of facts about motivation, but in terms of other natural facts about actions, agents, states of affairs, institutions, and other items in the natural or social world that have a claim on us in deliberation. In Burning Hotel, the fact that I have a reason might be identical with the natural fact that jumping into the canal is the only way of saving my life.

A virtue of this suggestion, I am about to argue, is that it helps to make better sense of the fact that some natural facts are normatively significant. To see why, consider the following trio of claims:

1. (1) There are normative facts.

2. (2) Normative facts are non-natural facts.

3. (3) Some natural facts are normatively significant.

The first two claims, (1) and (2), are obviously accepted by Parfit. Parfit should also concur with (3). The fact that jumping out of the window will save my life is surely normatively significant. It gives me my reason for jumping out of the window. The ordinary natural fact that jumping out of the window will save my life must surely matter to me.

The problem is that (2) will be in tension with (3) if Parfit also wishes to hold (4):

1. (4) Normatively significant facts are normative facts.

This is because (4), in combination with (2), entails (5):

1. (5) Normatively significant facts are non-natural facts.

And (5) is straightforwardly inconsistent with (3). But does Parfit hold (5)? He might perhaps say that he holds, not (5), but (6):

1. (6) A natural fact may be normatively significant, but it is only the fact that a natural fact is normatively significant that is the normative fact.

Claim (6) makes it clear that a natural, normatively significant fact, such as the fact that jumping out of the window is the only way to save my life, is not itself a normative fact.

But what independent grounds could there be for holding (6)? The explanation could not be, merely, because it is concerned with a second-order fact, or with the fact that some other fact obtains. For consider a natural, empirically observable fact about motivation: (C), for example, which says that jumping into the canal is what, having deliberated on the fact that doing so is the only way of saving my life, I would choose to do. If facts about other facts are, by their very nature, non-natural, then we would have no choice but to endorse (7), which is the product of the application of this suggestion to (C):

1. (7) The fact that jumping into the canal is what, after rationally deliberating on the fact that jumping is the only way of saving my life, I am most strongly motivated to do, is a non-natural fact.

If we were prepared to endorse (7), then both facts about reasons and facts about motivation would all be non-natural facts. That surely establishes too much. There must be special considerations, then, applying to reasons but not motivation, why Parfit thinks that facts about reasons are all non-natural facts. What are they?

There is another problem with (6): the fact that jumping out of the window is the only way of saving my life is not, in itself, genuinely normatively significant. It is not the fact that the natural fact is normatively significant which is the normative fact; rather, it is the fact which makes the natural fact normatively significant that is the normative fact. And what makes the natural fact normatively significant, in turn, is that it is related in some way – presumably via the relation of supervenience or something similar – to the fully normative fact that I have a reason for jumping. Call this relation, whatever it is, the S-relation. I suggest, then, that a more perspicuous replacement for (6) is given by (8):

1. (8) A natural fact is not normatively significant in itself, or in virtue of being picked out by the best normative moral theory, but only in virtue of being S-related to the normative fact that makes the natural fact normatively significant.

Do we really need ties of normative dependence of this baroque level of complexity? I don't think we do.

VI

Why did anyone ever bother to try analysing normativity in ways that refuse to depart from the constraints imposed by Metaphysical Naturalism? Plainly, to avoid ‘queerness’ worries of the sort Mackie and others drew attention to.Footnote ⁵ But Parfit takes his preferred form of Cognitivist Non-Naturalism – its full title is eventually rendered as Non-Metaphysical Non-Naturalist Normative Cognitivism (2, 486) – to exclude any burdensome ontological implications.

Parfit's refusal to shoulder any such ontological burdens is driven by what might be called the safety in numbers move: since numbers and other abstract entities do not have spatio-temporal locations, and since they cannot be reasonably excluded from a suitably expansive or tolerant conception of reality, then neither should the normative facts and properties demanded by Non-Naturalist Cognitivism be excluded from this conception of reality. This kind of philosophical ointment is often applied by moral realists, but it will not impress metaethical theorists who insist that, if values can be successfully analysed in other ways, then they should be analysed in other ways, on pain of exposure to a charge of special ontological pleading. After all, there is arguably nothing for numbers to be except abstract entities; for values, by contrast, there are other possibilities.

Parfit spends a large part of On What Matters trying to convince us that these ontologically weightless facts and properties make all the difference to our practical lives. They allow us to have lives that matter. But, for me, their existence came to seem less and less important. The things that matter don't need the things which, on Parfit's view, make them matter.