Derek Parfit has recently circulated an argument against what he calls Non-Analytical Naturalism, which he understands as the thesis that normative truths are reducible to natural truths. He begins by stipulating that he will use ‘normative’ as an abbreviation for ‘irreducibly normative’:
When some normative word cannot be analyzed or defined in non-normative terms, we can call this word, and the concept it expresses, irreducibly normative. That is what I shall mean by ‘normative’…
His central argument then appears to be that Non-Analytic Naturalism is (by definition) inconsistent with normativity, and hence false.
Now, I have great sympathy for the idea that it should be a test of adequacy on any reductive theory, that it can account for the normativity of wrong, good, and so on. And I think it is a quite interesting question whether any reductive theory can account for this, and moreover what constraints, if any, would be placed on a reductive theory that would be able to do so. I simply don’t see, however, how Parfit’s argument contributes anything at all to this investigation, or in fact amounts to anything other than a stipulation that he is going to use words in a way that presupposes his conclusion.
It’s easy to stipulate that by ‘normative’ we will mean, inter alia, ‘something that a reductive theory cannot account for’. That makes the step to the conclusion that reductive theories cannot account for normativity trivial. The interesting question is whether the senses of normative words that Parfit claims are irreducible are in fact so.
If anyone sees anything that look more like an argument lying behind Parfit’s remarks, I’m very much hoping to discover what it is supposed to be.
Mark:
I spent just a few minutes skimming through Parfit’s text, so I can’t really say that I have a good grasp of the text. Howeve, if my version of Word is displaying the diagrams correctly, he defines Non-Analytical Naturalism as a view that accepts that normative concepts cannot be analyzed in non-reductive terms (hence the “non-analytical”), but claims that normative truths are not irreducibly normative. So you might think that this is not the best way to capture a view like Boyd’s or that no one would hold a view characterized this way, but given that NAN by definition accepts the substance of the passage you quoted (even if need not accept that the claim in question is true by definition), Parfit can’t be accused of begging the question against NAN. Having said that I must confess that I don’t know exactly what the dialogue between the Natural Utilitarian and Sidgwick is supposed to establish. As far as I can tell Parfit’s argument there goes as follows (this is obviously a rough sketch):
(A) Sidgwick and Ross believe that there are two distinct properties: the property of maximizing utility and the property of being right.
(B) However, given that, by hypothesis NU is right, the property of being right and the property of maximizing utility is the same property.
(C) So S & R’s belief described in (A) is a false belief.
(D) Thus either there is no such thing as what S & R believes to be the property of maximizing utility or there is no such thing as what S & R believes to be the property of being right.
(E) S & R have true beliefs of the form “X maximizes utility”
(F) From (D) and (E): There is no such thing as what S & R believes to be the property of being right.
But (D) clearly does not follow from the previous steps. After all S & R’s belief has three conjuncts: there is a property of being right, there is a property of maximizing utility, and these properties are non-identical. And the obvious thing to say is that the mistaken conjunct is the last one, but (D) overlooks this possibility. With some minor violence to Sidgwick and Ross (and to grammar), one can try to run the argument by replacing “property of being right”, “NU”, and “property of maximizing utility” with “Hesperus”, “Phosphorus”, “astronomy”, and “individual” to see that one is running into trouble when one gets to (D).
But again, I just skimmed through the text, so I must be missing something here.
Hi, Sergio.
I agree with your diagnosis of the Sidgwick/Ross discussion – I discern more than one philosophical flaw in the appendix, and was trying to restrict myself to commenting on the most glaring. Your explication goes into more detail than the version in the text, but it resonates with an explanation I remember him giving in a workshop at Rutgers several years ago.
On the characterization of Non-Analytic Naturalism: I agree that as defined it fails to include Cornell Realism. It interests me, however, because I myself believe that normative properties and relations are reducible to non-normative ones, but that the correct reductive theses are non-analytic.
So if I get to speak for myself, I don’t accept Parfit’s stipulation that ‘normative’ implies irreducibility. That is the thesis for which he is supposed to be providing an argument. If he wants to stipulate that ‘normative’ implies irreducibility, then the argumentative burden merely gets shifted, to having to argue for the view that there are ‘normative’ truths (in this sense). But ignoring the clever use of persuasive definition, denying that there are ‘normative’ truths in this sense does not mean denying that there are normative truths, in any ordinary sense. I don’t see where Parfit argues either of these things, or even tries to.
After a very cursory read of Parfit’s text, I find myself similarly dissatisfied. Parfit does not offer much in terms of an argument against the non-analytical naturalist. He merely states (e.g.) that the gap between the natural and the normative is “too great to be bridged,” that the property of making right is not the same as the property of being right, that “normative and natural facts differ too deeply for any form of Normative Naturalism to be true,” and that “Normative Naturalism could not be true, since normative claims could not state natural facts.”
However, I do not know if I think Parfit is guilty of the error that Mark describes. I admit that the text is a little unclear, but I think that Parfit is stipulating that a term is “intrinsically normative” when it cannot be analytically defined in non-normative terms. However, when Parfit turns to discuss non-analytical naturalists, he says that their theory is that the normative terms state the same facts as the non-normative terms, even though the normative terms might not be analytically definable via non-normative terms. So for the analytical naturalists, rightness might just be maximizing happiness, even if “rightness” and “maximizing happiness” are not definitionally equivalent.
So using Parfit’s definition of “intrinsically normative,” people like the Cornell Realists might accept that the core moral terms (“right,” “good,” etc.) are intrinsically normative! After all, Boyd holds that “morally right” cannot be analytically defined at all: its synthetic definition is allegedly some homeostatic property-cluster, or (taking certain liberties with Boyd’s text) some verbal description of such a cluster. Whatever Boyd’s theory of definitions, he clearly thinks that biconditional analyses of moral phenomena are not possible. And without a biconditional, you’re not going to get an analytic definition, which is thought by many to require a meaning-giving, necessary biconditional.
But as I say, Parfit’s explanation for his rejection of non-analytical naturalism is a little sketchy. The following passage seems to sum up his view:
Basically, the idea is that even if being such that it will maximize happiness provides a reason to perform an action, thereby making the action right, this cannot be what being right consists in. Why not? As Parfit says,
Well, okay.
I had much the same worries (and a lot more – a lot could be said about the use of the mathematics analogy). About normativity, he defines normative concepts in this way:
“Concepts are normative in what we can call the reason-involving sense when these concepts can be used to make or imply claims about reasons.Claims are in this sense normative, we can roughly say, when these claims either assert or imply that someone does, or would, or might have some reason. This, I believe, is the best sense of ‘normative’, and is the sense that I use.”
One worry is that a non-reductive naturalist can happily accept all of this. All the naturalist is going to add is that the implied, asserted, or made claims about reasons are made true by the natural facts of the world. And, he can explain the difference in meaning by for instance different implicatures of the terms. I just didn’t see how reasons were something that a naturalist could not cope with. Parfit does say something about them, mainly this:
“In claiming that these beliefs are quite different, I am assuming that the concept of a reason is irreducibly normative, as is the concept of what we must do in the decisive-reason-implying sense.”
Now the circle is really small. And, still it’s not clear why the naturalist cannot have reasons and thus normativity.
Jussi’s point about reasons is the one I made in Realism and Reduction, anticipating that the objection from normativity would amount to precisely this (it’s the same claim that Dancy makes in the piece just discussed in a recent thread).
If reasons are reducible or naturalistic, and normativity is a matter of involving reasons, then normativity can be reducible or naturalistic. So if you want to evaluate whether normativity can be reducible or naturalistic, then you need to evaluate whether reasons are reducible or naturalistic – not just assume that they aren’t.
The view I’ve defended, for example, is conjunctive. Other normative properties and relations reduce to reasons (a thesis I share with many nonreductivists) and reasons are reducible in non-normative terms (via a ‘desire-based’ theory). Hence all normative properties and relations are reducible. If I understand it correctly, Parfit’s argument against my conjunctive thesis works by simply assuming that its second conjunct is false.
In fact, I think his arguments against noncognitivism in the later part of the appendix have a very similar structure, as well – they are arguments against views which need to be understood as having two parts, which argue against the first part on the assumption that the second part is false.
I’ve actually just had correspondence with Parfit about this very issue. He will say that the Cornell Realists might themselves think that they have identified irreducible normative facts, but they haven’t. Or alternatively, that a neo-Aristotelian naturalist like Foot may think she has given a naturalistic account of reasons, but she hasn’t. The reason they haven’t is that Parfit doesn’t consider the facts that they’ve identified as genuinely normative. Why not? As Mark points out, he stipulates that ‘normative’ for him means ‘irreducibly normative’, and ‘irreducibly normative’ means ‘cannot be analysed or defined in nonnormative terms’. The problem isn’t just that this defines naturalism out of the realm of possibility, it’s that it effectively says nothing. For why aren’t the Cornell realists then nonnaturalists? It’s because Parfit doesn’t consider anything ‘natural’ to be normative. What does ‘natural’ mean? Parfit, it seems, defines ‘natural’ merely as ‘not normative’. A small circle indeed! It seems the only evidence of whether something is ‘normative’ or ‘natural’ is the pricking of Parfit’s thumbs!
I’m still not convinced that Parfit is using his definition of “irreducibly normative” as a premise in an argument against non-reductive naturalism. However, what Steve says suggests that Parfit may be using a slippery definition of “natural” in such an argument. That said, I could not find a definition of “natural” in Parfit’s appendix.
Mark writes that on his own view,
normative properties and relations reduce to reasons (a thesis I share with many nonreductivists) and reasons are reducible in non-normative terms (via a ‘desire-based’ theory). … If I understand it correctly, Parfit’s argument against my conjunctive thesis works by simply assuming that its second conjunct is false.
I think that Parfit does have a little more to say, here. Here is the argument that he seems to make on page 11. I would be interested to know what others think of it.
First, consider these two statements, entertained by Parfit in his burning hotel room.
(a) Jumping is the only way to get what I most want,
(b) I must jump (i.e., I have sufficient reason to jump).
The crux of the argument is of course line 2, which is an intuition about meaning.
Hi, Jason.
On my desire-based theory of reasons, Parfit’s (a) and (b) aren’t even equivalent. But more generally, it doesn’t even purport to be an argument against non-analytic naturalism – only analytic naturalists (and only those who understand analyticity in a certain way, at that) will be committed to the claim that if (a) serves in some sense as an ‘analysis’ of (b), then they must express the same proposition. I want to know what the argument is supposed to be against non-analytic naturalism.
Hi Mark,
Agreed: the argument I described is intended as an argument against analytical naturalism. When you said, “reasons are reducible in non-normative terms,” I thought you were proposing an analytical reduction. Sorry about that; you clearly stated earlier that “the correct reductive theses are non-analytic.”
I think Parfit has no argument against non-analytic naturalism. He thinks that reasons turning out to be constituted in some way by desires is like heat turning out to be a medieval king: it’s just impossible. If you understand the concept of reasons, he seems to be saying, you should immediately see that it is impossible, too. He even seems to admit that he has no argument against the view towards the end of section 4.
And, it’s not clear whether that’s a good argument even against analytic naturalism. Consider the following version of it:
(a) that closed shape has eight straight sides,
(b) that shape is an octagon.
1. If an eight-side-based analytic reduction of octagons is true, then when Parfit reasons from (a) to (b), his inference is of the form “p, therefore p.”
2. But it’s not the case that his inference is of the form “p, therefore p”: (a) and (b) express different propositions.
3. Therefore, no eight-side-based analytic reduction of octagons is true.
This argument goes wrong in (2). We know that (a) and (b) here express same propositions. It’s not obvious in any way that they do and thus the inference is not obviously of the form ‘p, therefore p’ to us. This would explain why inference seems different from the the formal version. I’m quite certain that the analytic naturalist would say the same in the moral case.
Thus, if the analytic naturalist thinks that meanings are not transparent, then Parfit’s argument cannot rest on our intuitions about the obviousness of the inference or its apparent form. It must rest of the premise that (a) and (b) express different propositions. But, in an argument against an analytic naturalist that is the most question-begging place to begin from. That’s the conclusion we wanted not the premise.
Hi, Jason.
Parfit seems to think it is obvious but inexplicable that heat could not turn out to be a medieval king – he seems to suggest that though it is not part of the meaning of ‘heat’ that it picks out molecular kinetic energy, it is part of its meaning that it does not pick out a medieval king, and similarly it is part of the meaning of ‘right’ that it does not pick out anything that can be reduced in non-normative terms (which is identical with his stipulation about how to use ‘normative’).
I think we can do better than brute intuitions when it comes to figuring out why heat couldn’t have turned out to be a medieval king. Here’s one reason: heat is still around, and medieval kings aren’t. So by Leibniz’s Law, heat can’t be a medieval king. There are lots of obvious features that heat has that medieval kings don’t, and this is obvious to us. That’s why it’s obvious that heat couldn’t have been a medieval king – not because it is somehow part of the meaning of ‘heat’ that it can’t be a medieval king.
So if you want to argue that reasons could not be reducible in non-normative terms, you need to exhibit some property that reasons have, which nothing constructed out of the non-normative could have. You can’t say that it is irreducibility to the non-normative, because that is what is at stake. You can’t say that it is normativity, and that normativity is reason-involvingness, because since reasons themselves involve reasons, they are trivially normative on this view, whether they are reducible or not.
My point is: here is a test – a formula for developing an argument that reasons are irreducible to the non-normative. You come up with a feature that reasons have that nothing constructed out of the non-normative could have. So it’s not that irreducibility is not the kind of thing that can be argued for. It’s just that Parfit isn’t, in fact, arguing for it.
A small followup point to Jussi: I said above that whether (1) was correct depends on your conception of analyticity. Only on some conceptions of analyticity does it follow that if (a) is the analysis of (b) and is analytically so, then (a) and (b) must express the same proposition. This won’t always be so.
For example, someone who accepts Jeff King’s account of philosophical analysis but has a Fregean conception of the semantic values of predicates might hold that in
(a) murder fails to maximize happiness
(b) murder is wrong
(a) and (b) express different propositions, even though the concept expressed by ‘wrong’ is conceptually analyzed in terms of the concepts expressed by ‘fails to’, ‘maximize’, and ‘happiness’. Such a person could claim that the analysis counts as ‘analytic’, because having the concept expressed by ‘wrong’ means having a complex concept which consists of the other parts, and hence involves being able to grasp this connection, but deny that (a) and (b) express the same proposition. Hence such a view would accept (2) but deny (1). And there are other conceptions of analyticity and of analysis that will have the same result.
At the risk of being redundant, I would like to summarize our discussion of Parfit’s views on naturalism. I trust that others will tell me if I have understood things incorrectly.
Parfit could be making this argument:
As Mark pointed out, this argument commits an informal fallacy (persuasive definition). In addition, as several people pointed out, if “analyze” is taken in the strict and old-fashioned sense, premise 2 is false: non-reductive naturalisms like Cornell Realism (and maybe also Foot’s neo-Aristotelianism) do not say that normative concepts can be analytically defined. They say that normative phenomena just are or are entirely constituted by natural phenomena. That’s different.
Here is another interpretation of Parfit’s discussion.
But, to put it mildly, it is not certain that premise 1 is true. Premise 1 is explicitly denied by both reductive and non-reductive naturalists – as well as by those who are skeptical of conceptual truths, in general. Parfit may think that premise 1 is as obvious as facts about heat are not merely facts about medieval kings. But these two claims are not analogous. We can explain in a straight-forward way why facts about heat could not be facts about medieval kings: as Mark pointed out, there is still heat, but there are no longer medieval kings.
Though this is somewhat speculative, here is a way in which Parfit might be linking these arguments:
But again, premise 1 is uncertain.
I conclude (provisionally) that Parfit has not provided any persuasive arguments against non-reductive naturalism in this piece.
I’m also curious about the second premise. Not all naturalists think that reason-facts are psychological facts. Do they think that they are causal facts? I’m not sure about that. At least someone like Jackson would probably not think that they are. The disjunctive properties he thinks moral properties are do not necessarily play any unique causal role. So, I think there is room for a naturalist to not to accept 2 of the second argument. On the other hand, it does seem plausible to everyone that facts about reasons do play some role in causal explanations. Railton’s example of milk-buying example is a good one. Alternatively, if causal and psychological facts exhaust natural facts, then there is no difference between the first premise and the conclusion – and thus there is no real argument. The premise and the conclusion are just as uncertain or certain.
Hi, Jussi.
I think this is an example of arguing against exemplars, instead of offering a general argument. I don’t think that normative facts are causal (or dispositional) facts, and I don’t think that they are psychological facts. If we limited possible reductive theories to ones which say that having a reason to do something is being caused to do it, or being disposed to be caused to do it, it’s easy to see why Parfit would think those fail to capture the normative dimension. But that’s not a general argument; it’s persuasive only if you forget that such views occupy only a very small portion of the logical space of reductive theories.
Jason says:
Here is another interpretation of Parfit’s discussion.
1. It is a conceptual truth that facts about reasons are not merely causal or psychological facts. (p. 21)
2. Naturalism (reductive or non-reductive) says that facts about reasons are merely causal or psychological facts.
3. Therefore, naturalism is false.
But, to put it mildly, it is not certain that premise 1 is true. Premise 1 is explicitly denied by both reductive and non-reductive naturalists – as well as by those who are skeptical of conceptual truths, in general.
I think this is the point of Parfit’s remark (pp. 18-19) on why, on his view, the difference between his view and that of naturalists of any stripe is not a matter of differing conceptions of concepts. He maintains that he and naturalists are not speaking of the same thing in deploying normative concepts.
As a non-naturalist myself, and reflecting on this thread, I am inclined to think he is right. It is hard, after all, to conceive what an argument against a proposed reduction across categories (as the non-naturalist conceives it) must look like. By hypothesis, the anti-reductivist thinks something must be left out, and the reductivist thinks not. In this case, perhaps the best that can be done, and in part what Parfit takes himself to be doing, is ostending what that might be that is left out. It is patently left out in cases of reduction to motivation, as those non-reductivists who make normative claims about what one has reason to do, or one ought to do, deny that what they mean can be exhausted by any claim about how anyone is motivated to behave. So, I take the point that Parfit has not given an argument that is likely to move a naturalist. But, two things:
(1) does it follow that he is failing to do something he takes himself to be doing? I’m not sure I see that.
(2) To shift the burden here a bit, what exactly is it that someone who characterizes his view as “naturalist” thinks to be doing? Is it simply denying that, whatever non-naturalists think they are doing, they are wrong? Part of the amorphous nature of any attack on naturalism may be due to the lack of clarity as to exactly what is involved in the naturalist program itself. So is it Parfit’s burden to explain what “naturalism” may mean, so as to deny it? Or is it plausible for him to pick some views (such as Williams’ or Brandt’s) that take themselves to be offering a naturalist program, and show that they are not thinking of ‘reason’ in the way that he is?
Couple of quick things. First, there are many naturalist views that give very detailed accounts of the semantics of normative claims. There’s Jackson’s functionalism, the Boyd-Brink view based on externalist semantics, Railton’s revisionism, and so on. These are highly developed views about what normative claims are about and what makes them true. But, it doesn’t seem that Parfit is interested in discussing such views.
And, it’s not like we cannot pursue arguments against such views. There’s plenty around. There’s more one can do besides saying that this is what I mean. Discussing arguments like the Horgan/Timmons Moral Twin-Earth cases or th ones from judgment-internalism is fruitful in thinking about whether naturalists can capture what we mean. But, again, it does seem like Parfit wants to engage with them.
There’s also a further worry from the naturalists side. Assume that non-reductive, non-naturalist cognitivism is the right story about what our normative claims are about. This opens up the possibility that the world doesn’t do its part and we are left with error theory.
Of course there are specific proposals for the semantics of specific naturalist views. The question I meant to ask was, what is the point of framing such views as naturalist? It’s not clear to me that there is consensus on just what is common to the approaches you indicate (and others) and thus what the non-naturalist can do other than indicate where specific views go wrong (by non-naturalist lights). Parfit does do precisely that. If the claim is that he doesn’t do enough of it, or doesn’t engage the most interesting species of naturalism, that would seem to me to be a fair claim. But that’s not the objection that people have been making., which is that he has no argument at all.
Moreover, arguments against such views assume that there is shared content in speaking of ‘reasons’, ‘normativity’, and so on. I take Parfit to be claiming that this is a mistake — that, given the sort of non-naturalism he is interested in and the shapes that various reductive theories take, it is not plausible to think that they are talking about the same thing. That’s an argument that we should not expect to find arguments such as we would find between competing conceptions of a concept. To the extent, for example, that Jason doubts that Premise 1 of the argument he offers in his mini-recapitulation, I think Parfit will say that that indicates that the two of them are talking past each other, not disagreeing about the extension of a concept with shared meaning. If he is right, I’m not sure what sorts of arguments we should expect.
In correspondence this morning, Parfit emphasized the point Mark LeBar has just made. He thinks that his stipulations to use the word ‘normative’ to mean ‘irreducible’ are not question-begging, because his view is not that reductive views are false, but that they are simply using their words differently. I disagree with Mark’s proposal that this is not the sort of claim that we should expect arguments for; I believe that I showed how one could evaluate such a thesis in Realism and Reduction.
I suggested to Parfit that this doesn’t make reductive views come out true, because it is part of reductive views that they are about the ORDINARY senses of ‘right’ and ‘reason’ – the ones used in normative moral theorizing – a thesis that Parfit can’t allow them, and still maintain that he is using the terms in the ORDINARY sense. His response was that he doesn’t care if he is using ‘right’ and ‘reason’ in their ORDINARY sense – the sense in which they are used in normative moral theorizing (though I think this conflicts with the remarks about Sidgwick and Ross). What he cares about is whether there is some interesting thing to talk about, which reductive views miss out on.
That’s fair – but in that case, he still needs an argument that there is anything which satisfies his non-ORDINARY uses of ‘right’ and ‘reason’, and he can’t any longer say that to deny this is to deny that anything is right, or a reason, for it will only amount to denying that anything is ‘right’ in Parfit’s made-up sense – not denying that anything is right.
Any way you cut it, Parfit and I are still disagreeing about something. I don’t claim to have any arguments which show that Parfit’s view couldn’t possibly be true. (He’s the one who claims to have arguments that my view couldn’t possibly be true, at least about the interesting senses of ‘right’ and ‘reason’.) I think it’s nonexplanatory, and that reduction is a fruitful explanatory hypothesis, and well worth investigating. I have a higher credence on the normative being reducible than on its not being reducible, because there is a productive current research program which is making steady progress on explaining features of the normative in non-normative terms, but I have seen no similar progress on the non-reductivists’ side in explaining the kinds of things in moral metaphysics, moral epistemology, and moral semantics that I think reductive hypotheses help to explain.
I don’t claim to defend ‘naturalism’ as such, and so I feel no particular urge to define it, but since I think that normative properties and relations – the very ones that are of interest in normative moral theory and which Sidgwick and Ross disagreed about – bear the same relationship to non-normative properties and relations as triangularity bears to the relations denoted by ‘three’, ‘sided’ and ‘plane figure’, I’m pretty confident that my view is one Parfit thinks is false.
The following scenario caricatures the debate between me and Parfit: Suppose that there has been a murder. Parfit thinks the murderer is Jones. I think the murderer is Smith. We’re both detectives, and our job is to figure out who it is, so we begin to make our cases to one another. I say things like:
‘Look, the murder was done with an axe, and Smith owns an axe. He had a motive. And I know there’s that puzzle about how he got into the building, but my interview with the security guard is this afternoon. It’s a work in progress, but I think this hypothesis is promising.’
Parfit says things like:
‘I’m going to use the word ‘murderer’ as a shorthand for ‘murderer who isn’t Smith’. Schroeder says, ‘the murderer is Smith’, and that’s great – he’s surely speaking truly, given what he means by his words, but the only inference to make, is that he must mean something different by his words than I do. The murderer is too different a kind of thing from Smith. I can’t say exactly why, but in my view, the murderer is as different a kind of thing from Smith as heat is from medieval kings. Moreover, no one was the murderer unless they used an axe, and I’m assuming that Smith didn’t use an axe.’
Mark LeBar,
I thought it would have been clear what the *naturalism* of all those views consists of. Here’s one way to think about it. Imagine that God creates a new world. He puts in all the naturalist properties which can be studied by empirical sciences. Then he stops. He creates nothing else to that world – not a single non-natural property or a bridge-law that would get such properties in virtue of the natural properties. The naturalist claims that moral claims would still be true in that world. This is what the non-naturalist will have to deny. Moral claims would only start to be true if God continued to create more stuff.
Sometimes it looks like Parfit wants to say that there has to be nothing more in the world to make the irreducable moral claims (that are not about the natural features of the world) true. He writes that: “For there to be irreducibly normative facts, there does not have to be some part of reality to which these facts correspond”. Moral facts like mathematical facts are not wordly facts.
This makes me want to ask a lot of questions. Aren’t the facts supposed to make our moral claims true? If they are not part of the reality how could they? What is the ontological status of these facts if not part of the reality? Aren’t mathematical facts facts true in virtue of the meanings of the mathematical terms, not facts at all (Field), fictions, or empirical truths like some philosophers claim? If moral facts are facts like these, isn’t the Sidgwickian view just analytic naturalism, error theory, non-cognitivism, or non-reductive naturalism in disquise? Or are there synthetic mathemtatical truths as well? But, what makes them true?
Jussi, I don’t want to turn this thread into a referendum on naturalism, but what you suggest (limitation to what properties studied in the empirical sciences), is only one conception of naturalism, and perhaps not the best. (Does your characterization embrace the property, “is created by God,” a la Plantinga? Does it embrace the properties necessary for the practical of empirical science itself, a la Hampton?) Williams would reject that characterization; so far as I can tell, so would Jackson. Maybe there are workarounds for this one, maybe some other conception is better. But what do all these have in common with the intent you suggest, of focusing on the empirical sciences? Maybe what they have in common is just what they are opposed to (e.g. normativity and other stuff which we all know what it is but can’t quite say, precisely). That seems to leave a considerable amount of heterogeneity for Parfit to confront.
Having said that, I accept Mark S’s point (made here and in correspondence), which I take to be: there are reductivisms, and then there are reductivisms. Some take the phenomena to be reduced in a pretty crude way, wave their hands over what they don’t capture and insist they’ve got it all. All the anti-reductivist can do here — what I see Parfit as doing, at least in his discussions of Williams and Brandt — is pointing out what has been left out. If one insists that in fact nothing has been left out, then it’s hard to see what more there is to say other than, “We’re just not talking about the same thing.” That’s a sad point, because it is pretty much a conversation-stopper.
Mark’s claim, on the other hand, is that there are reductivist strategists who accept that complaint as legitimate, but don’t think all reductive accounts have to share this fate. Maybe he’s right, and if he is, then it is fair to complain that Parfit hasn’t dealt with the most competent of the reductivist strategies. But I think here there is a burden on such a “new wave” of reductivists to make clear the distance from the old approaches — to make clear how there is something to be learned from those failed reductions that can point the way to better reductions, rather than to anti-reductivism.
One last comment in defense of what Parfit does do: Mark S speaks (in his paper) of the “propagation” of problems in failed reductions. I take that to be something like the point that we should expect failed reductions to break down someplace in trying to fill the conceptual functions that the reduced concept serves. And I see Parfit doing this, not in his section on naturalism, but in his discussion of Gibbard. Here he makes a number of claims about Gibbard’s pragmatics, that for instance in the course of arguing, Gibbard wants to make use of ordinary (non-reduced) normative notions in thinking about conversational demands, understanding, answers to questions, and the like. If Parfit is right (and I am not insisting that he is), then if Gibbard is right, his own argumentative practice would be undermined. Some, anyway, of the conceptual roles that the antireductivist (and, apparently, his interlocutor) thinks normative concepts play cannot adequately be played by what Gibbard proposes as an analysis. That would, it seems to me, be the right sort of argument to give — to show that the reduction breaks down when one considers the conceptual work that the reduced concept is supposed to do. Setting aside whether Parfit is right about Gibbard or not, that seems to me at least a plausible form of argument, does it not?
I’m not sure I was less than convinced by the argument. I mean many expressivists I know are happy to use other normative notions in construing their accounts of a given normative term. They are also happy to talk about irreducible normative truths, facts, reasons, oughts, properties and so on. It’s just that they later on bring out expressivist accounts about also these other areas of normative discourse they used previously. This sort of piecemeal approach seems to make sense.
If there is a problem with this, I’m not sure that it’s been given to us so far. There are interesting debates about the normative talk we might need in psychology and meaning discourses. These are the areas where expressivism is probably most problematic and might have some self-undermining consequences. But, it’s not like we’ve there yet in Parfit’s discussion of Gibbard.
Well, I agree with Mark LeBar that in his discussion of Gibbard, Parfit is doing the right thing, to assess what Gibbard’s view predicts about lots of related phenomena. If it predicts the wrong things, then it’s wrong.
Unfortunately, I think that the only way to assess whether expressivism predicts the right things, is to spell out the sentences which state your predictions, formulate a complete expressivist semantics, and then check to see what the expressivist semantics says about the conditions under which those sentences are assertible, or whether that semantics is empirically inadequate for some other reason. I don’t see Parfit doing this; he seems to be relying on some kind of intuitive judgments about what Gibbard’s expressivism is committed to. I think the literature on expressivism over the past 2+ decades has proven that there’s no reason to think we’re able to make such intuitive judgments about the commitments of expressivism in a non-question-begging way, when part of the expressivist’s theory is about the semantics for such sentences.
Nevertheless, Mark is right that I think that the same problem about ‘propagating implications’ will come up for expressivist views; I do think that you have to work awfully hard in order to show this, though. Parfit is going in the right direction; it’s just obvious what Gibbard will say to each of the things he’s raised so far.