Normative Spontaneity

John Skorupski

I am very grateful to Kurt Sylvan for his insightful and generous discussion of my book, The Domain of Reasons (henceforth ‘DR’).[i] The main part of the book is, as he says, a contribution to what he calls ‘the reasons first project.’ I defend the thesis that the concept of a reason relation is the primitive normative concept to which all normative concepts can be reduced.[ii] Sylvan makes some interesting points about the way I defend it. However his focus is on what I say, in the final part of DR, about the epistemology and metaphysics of reason relations, where he has both criticisms and suggestions for improvement. Since a number of philosophers have had reactions to this metanormative account that are similar to his, I am happy to have the opportunity here to respond specifically to those criticisms and suggestions – and to explain why I think I should hold my ground.

Sylvan considers that the “most distinctive” aspect of my metanormative account, which “sets its starkly apart from the work of other reasons-firsters, is its Kantian flavour.” (10). But he is unconvinced by its basic idea. According to this, propositions about reason relations are truth-apt – in the simple and single sense of the word ‘true,’ the sense that obtains across all domains. Further, some purely normative propositions are true. Yet there are no substantive facts about reason relations: reason relations are irreal. I call this view irrealist cognitivism. As we shall see, it rejects both the correspondence conception of truth and the Carnap/Quine thesis that reference to an entity entails its existence. I shall suggest that some who think it confused or mysterious do so because they have not fully taken these aspects of the position into account: they criticise it from within realist metaphysical assumptions, even though it explicitly rejects those assumptions. And I shall suggest that this applies to Sylvan.

Sylvan thinks that my view of reason relations would be strengthened if I accepted some version of ‘Kantian constructivism;’ and that that would, moreover, bring me closer to Kant. I disagree on both counts. If I understand what contemporary ‘Kantian constructivism’ is, I do not agree that adopting it would strengthen my account of normativity. Nor do I agree that adopting it would make me ‘more Kantian’ (p. 10). On the contrary, I believe my cognitivist but irrealist account is closer to Kant than is anything that could be called ‘constructivism’ about reason.

Questions about Kant, I should add, are not central to Sylvan’s discussion. He is interested in whether I should be a Kantian constructivist in the contemporary sense of that label. But, for me, this is one of many cases in which awareness of the history of philosophy illuminates the viability of options that are being illegitimately ignored in some particular contemporary discussion. In the present case, thinking about Kant’s actual view helps one to get away from the false dichotomy ‘realist or non-cognitivist.’ So I will come back to Kant. But let me first pursue my diagnosis of why Sylvan finds my position ‘mysterious’ and why he thinks that going constructivist would improve it.

I Epistemology.

In The Domain of Reasons I argue that warrant for, and knowledge of, a priori propositions about reason relations has two pillars: first-person spontaneity and critical discussion with other spontaneously reason-sensitive judges. My notion of spontaneity is as close to Kant’s as it can be if one takes the notion out of its Kantian transcendental framework. Like Kant, I think that knowledge of nature requires receptivity whereas knowledge of reason (i.e. of a priori truths about reason relations) does not.

Sylvan’s critique of this epistemology of pure spontaneity takes the form of a dilemma, which he nicely states in the following passage:

While Skorupski helpfully illuminates the elusive notion of spontaneity on

pp. 406–10, I am left with questions about its epistemic role. I have two related

worries, which lead to a dilemma. Firstly, I do not understand how spontaneity could be a reliable guide to truths about reason-relations unless one accepts a cognition-dependent, constructivist account of them. And if spontaneity isn’t a reliable guide to normative truth but merely a source of internalist justification for normative propositions, I cannot see how it helps us to understand knowledge of normative truths. Although Skorupski’s account of epistemic warrant isn’t couched in terms of reliability, he doesn’t deny that reliability is necessary for normative knowledge; indeed, it is explicitly part of his view about a priori knowledge in general on p.161, and concerning the special case of normative knowledge on p.162 he just says that the way in which reliability is secured isn’t via a receptive faculty (619).

The dilemma, then, is this: either spontaneity gives one no account of normative knowledge (even if it yields an account of normative warrant) – in which case it is not an adequate account of the epistemology of the normative, since we rightly think that we know some pure normative truths – or if it yields an account of normative knowledge it has to be backed up by the constructivist metaphysics which I reject. Since I think that we know some normative truths I am forced to constructivism by my own epistemology, according to which warrant for and knowledge of the truth of normative propositions rests solely on the interplay of spontaneous propensities to judge, and not at all on any form of receptivity.

Now as Sylvan notes I hold that reliability is a necessary feature of knowledge but reject the view that knowledge must always be secured by a reliable receptive faculty. So I am not a ‘causal theorist of knowledge;’ I do not identify reliability with the special case of reliable receptivity. In some cases reliable judgement requires reliable receptive faculties, but not in the purely normative case. If I am a somewhat deaf person, then if I rely on the evidence of my ears alone I am an unreliable judge of what was said in a pub conversation. That is because I lack a sufficiently reliable receptive faculty giving me first-person access to the domain of sound. But if I am an unreliable judge of what would count as blameworthy, that is not because I lack a receptive faculty to some domain of normative facts about blameworthiness; my unreliability in this case is not a matter of unreliable receptivity.

Warrant in the purely normative case rests solely on spontaneity informed by intelligent interaction with others. This must also be the final basis of second-order judgements about how reliable another’s judgements are in a particular normative sphere. It will involve judgements about whether their first-person purely normative judgements are based on genuinely spontaneous impressions, or purely on testimony, or can be dismissed as in some way factitious (influenced, in Kant’s phrase, by ‘alien causes’ such as wishful thinking and many other things). From such discussions emerge judgements about who are more and less reliable judges in a given normative sphere. There is no test outside the circle of spontaneous judgements and critical discussion.

Much can be said about this epistemology of spontaneity and discussion of course ; I try to say some of it in chapter 16 of DR. The point I want to make here is that my account of it is not very different to other ‘reflective equilibrium’ accounts of normative knowledge. I have no particular disagreement with many others who have written about the epistemology of the normative in this spirit. Nor are these ideas novel: they go back a long way. They go further back than Kant and Sidgwick, but Sidgwick’s well-known account of ‘intuition’ exemplifies them. As he makes clear with his four tests of ‘intuition,’ for him too an ‘intuition’ is a spontaneous disposition to judge, educated through comparison with the dispositions of others, but still fallible.[iii] He is not an intuitionist in Kant’s sense of the word intuition (Anschauung) – he does not hold that normative  knowledge involves a special receptive faculty.

So the idea that there is normative knowledge, but that it does not involve any kind of receptivity to mind-independent existents is not new. In so far as DR contributes to it, it is by (i) working out a version of the receptivity/spontaneity contrast in contemporary terms, and (ii) working out, perhaps more fully than others, its metaphysical (or anti-metaphysical) implications.

Sylvan questions whether this epistemology can account for normative knowledge, as against normative warrant. There is of course a difference between warrant and knowledge: if you are warranted in the belief that p it does not follow that p, whereas if you know that p it does follow. DR puts forward what I call the ‘WK principle,’ according to which if we are warranted in believing that p we are warranted in believing that we know that p. In particular, if we are warranted in believing any proposition about reason relations we are warranted in believing that we know it. It follows that if we are never warranted in accepting that we know a normative judgement, then we are not warranted in accepting any normative judgement.

But the Critical outlook I defend in DR does not take this skeptical stance against common sense. It accepts that we are perfectly warranted in holding that we know some normative truths, hence warranted in holding that the relevant judgements pass whatever reliability conditions normative judgements must pass to count as normative knowledge. If those conditions required some kind of sui generis receptivity to a domain of substantive facts about reason-relations, as non-naturalist realism about the normative requires, then normative knowledge would not be possible. Since the Critical approach takes for granted the common-sense view that we have normative knowledge, and simply asks how that is possible, it must reject this account.

II Metaphysics.

But, Sylvan protests,

I do not understand how spontaneity could be a reliable guide to truths about reason-relations unless one accepts a cognition-dependent, constructivist account of them.

Well, as just noted, I quite agree that this epistemology cannot be linked to a realist metaphysics. But does that imply that we have to replace it with a ‘cognition-dependent,’ constructivist’ metaphysics? That is the key question.

At this point I must sketch some important underlying ideas, albeit without the explanatory discussion that they ideally require. The first thing to note is what I shall call the Thesis: any knowledge of cognition-independent existents must involve receptivity. It is central to this discussion; it is also central to Kant’s Copernican revolution: he reacts to it by accepting that space and time are existents and inferring from the possibility of knowing them that they are in some elusive way cognition-dependent –‘subjective,’ ‘empirically real but transcendentally ideal.’ The last chapters of DR try to show in detail why this is not the right way to treat reason relations. I argue, in those chapters, that the right way is to reject not their cognition-independence but their existence. At the same time – this is the cognitivism part – I take it that we refer to them, quantify over them and make true and false assertions about them. Evidently then I reject Quine’s criterion of ontological commitment. The only condition that can legitimately be placed on reference (I say) is that we should know, and be able to communicate to each other, what we are talking about – what our topic is.  

Let me add, without going further into the issue here, that I distinguish in DR three kinds of non-existent entities that we can and do refer to: fictional entities (Sherlock Holmes), putatively existent but actually non-existent entities (phlogiston) and reason relations. I do not assimilate them: I am neither a fictionalist nor an error theorist about reason relations. Fictions differ from putative existents as products of free imagination differ from products of hypothesis. Reason relations are a wholly different category from either of these:  they are the ineliminable objects of thought as thought.

These are controversial views for anyone committed to a correspondence conception of truth and a Quinean conception of existence. They are however the views I defend. Their starting point is the (to my mind) entirely natural, pre-philosophical assumption that factuality and normativity are distinct: (purely) normative knowledge is not factual knowledge.

I think this starting point should be uncontroversial, a glimpse of the obvious. However it faces the stubborn dogma that all cognitive content is factual content. There are various routes to this, such as that of the logical positivists. But probably the key source nowadays is the correspondence conception of truth, which says that for a proposition to be true is for it to stand in a relation of correspondence, or ‘truth-making,’ to a fact.

Here again I must be brief. It is, right from the start, important not to conflate two notions of fact, nominal and substantive. ‘Fact’ in the nominal sense means the same as ‘truth.’ To say that for every true proposition there is a fact, in this sense of fact, is to say no more than that for every true proposition there is a truth. In this discussion note, for clarity, I am using ‘fact’ only in a substantial sense, not in a nominal sense. I have in mind what we ordinarily mean when we talk about the facts and sticking to the facts, or when we say, for example, that the fire in the power station was caused by the fact that some wires were insufficiently insulated. A fact consists in the possession, by some existents, of some property or properties – substantial properties, not nominal properties that have a merely semantical standing as e.g. senses of predicates. In DR I advance the further, synthetic a priori, claim that existence is causal standing and that a substantial property is a property which, if instantiated, has causal standing. Truth, in contrast, is a nominal, not a substantial property. 

The correspondence conception of truth holds that every truth ‘corresponds to’ is ‘made true by’ a fact; it is therefore committed to some substantial notion of fact (not necessarily to what I just suggested is the ordinary or default sense of the word). It entails that if there are normative truths there are normative facts. DR rejects the correspondence conception – across the board, not just in the normative case. It affirms that there are normative truths but denies that there are normative facts. It denies that the semantic values of singular terms have to be existents, and that meaningful predicates have to ‘correspond’ to substantial properties. In giving a homophonic truth-conditional semantics for normative sentences, it says, we do not commit ourselves to the existence of any normative existents or normative facts. Semantics is one thing, metaphysics is another. Assigning semantic values to constituents of sentences is an innocuous semantical task; the idea that all these semantic values are existents is a far from innocuous metaphysical thesis.

Sylvan is well aware of these claims in DR, where they are discussed at length, and I assume that he rejects them. But I am surprised that he does not highlight them, since they are pivotal to my rejection of ‘constructivism.’

Various things can be meant by this term. One kind of constructivism about reason relations would react to the Thesis by treating them in analogy with Kant’s treatment of space and time: like space and time reason relations exist, but cognition-dependently. Against this, cognitivist irrealism says that reason relations are irreal but in no sense cognition-dependent.

Nor does it endorse another, probably more common, form of constructivism, according to which what makes a claim about reason-relations true is something like: what the dispositions of reason-sensitive judges would be under some (non-trivialising) ideal conditions. This form of constructivism, unlike the former form, is a reductive view which ‘analyses out’ reference to reason relations. But both derive impetus from an underlying commitment to the correspondence conception of truth: they take it for granted that if a normative proposition is true, it must have a truth maker, and they then accommodate the Thesis by holding that reason relations are cognition-dependent.

The contrast, then, is that DR responds to the Thesis by affirming that reason relations are cognition independent but denying that they exist. (As one can replace the nominal use of the word ‘fact’ by ‘truth,’ so one can, if desired, avoid the quantificational use of the verb ‘to exist,’ by using instead the verb ‘to be’ – as in ‘There are many characters in War and Peace, some of whom, like Pierre, did not exist, while others, like Napoleon, did.’ This is a matter of clear terminology – it does not introduce a special metaphysical category of ‘being.’)

III Constructivism, irrealism, and cognitive internalism.

So much for a sketch of the broad metaphysical issues. But Sylvan also pushes my view towards constructivism by another route: he argues that a specific thesis which I call ‘cognitive internalism,’ and discuss sympathetically in DR, would, if I accepted it, commit me to constructivism. He makes interesting points about this thesis itself, which deserve a more extended discussion than I can give here. However I do want to explain why I do not think that the putative implication to constructivism holds.

Consider a claim of the form were truths p_i  to obtain there would be reason to y. I take it to be an abbreviated  universal statement:

(1) (x) (were truths p_i [iv] to obtain that would be a reason for x to y).

Reasons are universalisable in this sense. The question however is what is the scope of the universal quantifier – the range of the variable x?

This is not an easy question. In DR I consider various options without coming to a definite conclusion. But as Sylvan says I am sympathetic to the basic idea, which I associate both with Kant and with Williams, that truths of a given kind are reasons for you only if you can potentially tell in a first-person way (not just on testimony) that truths of that kind are indeed reasons – only if you can potentially recognize for yourself their reason-giving force. People vary greatly in their reason sensitivity across widely different areas, so let’s say, only if you are relevantly reason-sensitive. This is cognitive internalism. Note that it is a thesis about reasons as such, not only about what I call warranted reasons (ie. the reasons a person would be warranted in believing he has, given his epistemic state). We then have

(2) For all relevantly reason-sensitive x, were truths p_i to obtain that would be reason for x to y.

Cognitive internalism accepts (2). If a person is not relevantly reason-sensitive with respect to some domain of responses, y, (2) does not hold that p_i are a reason for that person to y.

In DR I imagine a person I call gratitude-blind Tom. Tom is highly reason-sensitive in all sorts of areas, say logic, or music criticism; but has no grasp, actual or potential, of reasons to feel grateful. This is because gratitude is not an emotion he ever feels, or has the potential to feel. It is not a part of his emotional make-up. In that case, according to cognitive internalism, it’s not just that he has no first-person reason to believe that being done a good turn is reason for him to say thank you: It is not, of itself, a reason for him to say thank you. Nonetheless (3) remains true:

(3) For all relevantly reason-sensitive x, were some person to do x a genuine good turn there would be reason for x to thank that person.

Tom does not fall within the range of x in (3), because he is not relevantly reason-sensitive to reasons of gratitude. And note that (3) is a truth that Tom cannot know in a spontaneous, first-person, way. So there are universal normative truths that are inaccessible to him on the basis of his own first-person spontaneity; it’s just that these truths do not include him in their scope.

According to cognitive internalism, then, truths p_i are a reason for x only if x can tell, recognize, know, that they are. Since these words are factives, we can make this a biconditional:

(4) Truths p_i are a reason for x if and only if x can tell, recognize, know, that they are.

But this is not yet constructivism. For a constructivist holds that what makes the left hand side of (4) true is some fact about x’s disposition to accept p_i as a reason, where x satisfies some appropriate factual constraints. This is also Sylvan’s understanding :

Kantian constructivism …  holds that to be a reason just is [my emphasis] to be a consideration that any rational agent could recognize as such in virtue of her constitution as a rational agent (624).

The constructivist would have to spell out ‘rational agent’ and ‘recognise’ non-question-beggingly, in a way that avoids primitive reference to reason relations as reason relations, and this seems to me to be an insurmountable problem. Be that as it may, my present point is that cognitive internalism does not force one to the constructivist view. Cognitive internalism is perfectly consistent with holding (as cognitive irrealism holds) that nothing ‘makes’ (in the truth maker sense) the left hand side of (4) true: if it is true it is simply true. So (3) for example is simply true – there is reason to feel grateful for a good turn.

Actually, nothing makes any proposition true in any sense intended by the correspondence theorist. To be sure, it is trivially true that a factual assertion is true only if the fact it asserts to obtain does obtain. This must be accepted by any theorist of truth who countenances talk of facts at all. Unlike the correspondence theory, however, this tautology is consistent with the DR thesis that purely normative truths are not factual truths. In this particular sense, then, factual assertions are not ‘simply’ true: their truth depends on the facts in a sense of ‘depend’ that any theorist of truth must accept. In contrast, purely normative assertions do not say that any fact obtains; the information they convey is normative and depends on no facts: such assertions are, if true, simply true.

We can add to the two kinds of constructivism considered so far another. It takes the form of a collectivist-voluntarist form of expressivism, according to which we construct and endorse our norms through negotiated agreements. I suggest in DR that it arises from the idea of autonomy, or self-legislation, though I also argue that it is not to be found in Kant. I will come back to this in the next section. The point I want to make here, about all three forms of constructivism – the non-cognitivist as well as the cognitivist – is that they all accept the realist/non-cognitivist dichotomy, all accept that if there are normative truths they must be factual truths, and do so because they all accept realism in a wider, global (or metaphysical) sense encapsulated in the correspondence conception of truth. If this metaphysical realism is the background assumption, it seems obvious that unless a truth-maker for claims of form (1) can be found, we have to go non-cognitivist about them. In this respect constructivism about the normative is on a par with realism about the normative. Both make the same metaphysically realist assumption – whereas cognitivist irrealism, while it is most directly irrealism about reason relations, is also committed to the rejection of metaphysical realism overall.

Irrealism about reason relations does agree with (non-naturalistic) normative realism about them in taking truths about reason relations to be irreducible. ‘There is reason to pursue pleasure’ is true if and only there is reason to pursue pleasure. But it sees a truth-condition as just that: a semantic condition on the truth of a sentence, not a specification of a fact to which the proposition expressed by the sentence has to ‘correspond.’ Likewise with semantic values: a semantic value specifies what a term refers to; it does not assume (or deny) the metaphysical thesis that what can be referred to exists. To appreciate the cognitivist irrealist position it is essential to distinguish clearly between metaphysical and merely semantical notions, and not to assume mistakenly that metaphysical doctrines somehow ‘fall out’ of truth-conditional semantics (or any other semantics).

I suspect that Sylvan is, at least implicitly, thinking about cognitivist irrealism from within a framework of global realism. It is suggestive that he thinks that unless I go constructivist my position is ‘dangerously close’ to that of my ‘non-naturalist realist opponents,’ and also interesting that he quotes Parfit as one of these, while at the same time citing a passage from Parfit that doesn’t sound realist at all:

There are some claims that are irreducibly normative in the reason-involving sense, and are in the strongest sense true. But these truths have no ontological implications. For such claims to be true, these reason-involving properties need not exist either as natural properties in the spatio-temporal world, or in some non-spatiotemporal part of reality’ (620-21)[v]

As Sylvan says, this sounds a lot like cognitivist irrealism; unsurprisingly, I don’t disagree with it. I’m not sure whether he finds Parfit’s view “mysterious” because he thinks it is realist or because he thinks it is irrealist. I agree that it would be mysterious if it were the former. Perhaps it is its non-naturalism that he finds mysterious. My view is irrealist, and for just that reason consistent with naturalism, at any rate in that basic sense of naturalism according to which the natural facts are all the facts. It is not naturalism but realism that I reject.

Note lastly that the case for rejecting the correspondence conception of truth and the Quinean criterion of ontological commitment is by no means based solely on the difficulties these doctrines pose for a sensible metanormative view. On the contrary, as others have spelt out in various contexts and ways, these all-too-influential dogmas generate puzzles in a wide variety of fields. Irrealist cognitivism about the normative fits into a much broader, independent case for rejecting them.[vi]

IV Was Kant a constructivist about reason?

I turn finally to the question, was Kant a constructivist? I am not a Kantian überhaupt. I note in DR that there are two fundamental ways in which my view of freedom, reason and morality differ from his. One very significant difference centres on whether there are reasons for feelings as well for beliefs and actions. In my view there are, in Kant’s view there are not – not really reasons for feeling, as against practical reasons for trying to cultivate feelings. This difference takes my account of morality away from Kant’s in ways which are quite basic, but which are not relevant here.  The other very significant difference is more relevant, in that it bears more closely on the metanormative issue: unlike Kant I don’t believe that transcendental idealism is required to make freedom and reason intelligible.

Still, there is a parallelism with Kant – he thinks that transcendental realism makes freedom impossible[vii] and must be rejected, I think that metaphysical realism, the realism of the correspondence conception, makes it impossible and must be rejected. It must be rejected, however, not to resolve the supposed conflict between freedom and determinism that Kant sets out in the third antinomy, but, rather, to allow a metanormative account of reason relations that can accommodate Kant’s compelling conception of freedom as rational self-determination.

Kant was indeed a constructivist about nature, empirical reality. Constructivism is implicit in the crucial ‘Copernican’ claim that transcendental idealism is required to defend empirical realism, as Kant develops it through his fundamental distinction between spontaneity and receptivity, his claim that both are required in every judgement about the empirical world, and his deductions of the categories. There is firm grounding here for a constructivist interpretation of his empirical-realist view of nature.

But none of this applies to Kant’s view of reason. The distinctive feature of Kant’s account of our pure knowledge of reason – or, as I would say, of reason relations – is that such knowledge is a product of spontaneity alone. It involves no receptivity to any input from a mind-independent domain of reality, natural or other. Since there is no such input there is nothing to construct from it, as there is in the case of empirical knowledge, where the material is provided by intuition (in Kant’s sense of the word.)  Knowledge of reason relations is presupposed in the construction of empirical reality, but there is nothing from which reason relations are themselves, in turn, ‘constructed.’

Nor is there any basis for reading Kant as a constructivist about reason in the senses discussed in sections II and III.  His overall[viii] epistemology of reason is that of spontaneity and discussion – but I have already argued that this does not entail constructivism unless one is a metaphysical realist. He is also, I believe, a cognitive internalist, but here again I have argued that cognitive internalism does not entail constructivism unless one adds a metaphysical-realist premise.

Some, however, have argued that Kant is a constructivist about reason on the basis of his doctrine of autonomy – that to act freely is to ‘give oneself the law.’ This doctrine can sound like a voluntarist form of constructivism. But what did Kant mean by it? Despite his not infrequent Rousseauesque rhetoric his meaning is actually quite limited and specific. This becomes clear when we note his distinction between the giver and the author of the law. Thus, for example,

One who commands (imperans) through a law is the lawgiver (legislator). He is the author (autor) of the obligation in accordance with the law, but not always the author of the law. In the latter case the law would be a positive (contingent) and chosen law. A law that binds us a priori and unconditionally by our own reason can also be expressed as proceeding from the will of a supreme lawgiver . . . but this signifies only the idea of a moral being whose will is a law for everyone, without his being thought as the author of the law.[ix]

A law that that “binds us a priori and unconditionally by our own reason” is not a positive law and has no author. A fortiori, we are not its author – no one is, not even God. It stands fast as an a priori and unconditional normative truth. What we can ‘author’ is the ‘obligation’ to act in accordance with the law: that is, we can bind ourselves to do so. When we recognise the law of reason the effect on our emotions and will is to create respect, strike down self-conceit, reduce self-interest and produce moral resolve: a command to oneself to act from the law. This is self-legislation; we thus require no command from another. However a ‘holy will’ would not experience requirements of reason as obligations at all. It would simply act from them without the finite being’s need for resolve. In this sense it does not ‘give itself the law.’

About reason as such then, Kant is an objectivist, even though our mode of awareness of reason is the subjective phenomenology of finite sensuous beings. At the same time, however, this objectivism does not make Kant a realist. A realist view would commit him to the idea that there are some facts to which a priori and unconditional principles of reason correspond. What facts could these be? Empirical facts or noumenal facts? Either option has clearly unKantian implications (if they correspond to empirical facts they are not a priori, if to noumenal facts they cannot be known). But there is no reason for him to adopt either view. Only the false non-cognitivist/realist dichotomy can make one think otherwise.[x]

To conclude. I am happy to agree with Sylvan that my metanormative account of reason relations has a ‘Kantian flavour.’ I do not accept however that taking a constructivist line on normativity would make it more Kantian. I think my metanormative account is already fully in accord with Kant, and that adopting ‘Kantian constructivism’ about reason relations would make it less so. To interpret Kant as a constructivist is to saddle him with the false dichotomies of recent meta-ethics.

Let me end, as I began, by thanking Kurt Sylvan again for the careful attention he has devoted to my book – and especially for making me more vividly aware of what parts of it an intelligent and informed reader is likely to find controversial.

References

Irwin, Terence, 2009, The Development of Ethics, A Historical and Critical Study, Volume 3, From Kant to Rawls, Oxford: Oxford University Press.

Parfit, Derek, 2011, On What Matters, Oxford: Oxford University Press.

Priest, Graham, 2005, Towards Non-Being, Oxford: Oxford University Press.

Sidgwick, Henry, 1981 (1907), The Methods of Ethics.

Stern, Robert. 2012, Understanding Moral Obligation: Kant, Hegel, Kierkegaard, Cambridge: Cambridge University Press.

Sylvan, Kurt, 2016, ‘Skorupski on spontaneity, apriority and normative truth,’ in Philosophical Quarterly, 66, 2016: 617 – 628.

---

[i] Sylvan 2016: 617 – 628.

[ii] To be precise, I argue that they can be reduced to three primitive reason relations.

[iii] See Sidgwick 1981 (1907): 338 – 43. Similarly I have no quarrel with Sosa’s conception of intuitions as ‘attractions to assent’ (cited by Sylvan). The argument of the book is that the spontaneity of such attractions to assent is the most primitive test of their normative authority.

[iv] Let these truths be stated so as to take account of any agent-relative back reference.

[v] See Parfit 2011, Volume 2: 486 .

[vi] An excellent exposition of the wide range of difficulties Quine’s criterion produces, together with systematic semantics for quantification over non-existent objects, is provided in Priest 2005.

[vii] “Were we to yield to the illusion of transcendental realism, neither nature nor freedom would be possible” Critique of Pure Reason, A543/B571.

[viii] ‘Overall’ – in Groundwork III Kant claims (1) that reason’s substantive requirements on a free being can be deduced analytically from the very idea of that being as free, and (2) that we cannot know we are free. The two theses together would imply that we cannot know that any normative claims about reason relations apply to us, i.e. include us in their scope, though we can know what they are. I set this aside in favour of the many places where Kant accepts that we know of rational requirements that they apply to us, and where the epistemology of spontaneity and discussion comes to the fore.

[ix] Metaphysics of Morals, 6:227.

[x] A number of authors have argued effectively against the constructivist interpretation of Kant’s ethics, among them Irwin (2009) and Stern (2012). However both take it that if Kant is not a constructivist he must be a ‘realist.’ Though I am not sure how much they mean by this term, it sounds like a return to the false dichotomy.