Bart Streumer makes an interesting case for an error theory in ethics—and for an error theory for normativity more generally, but I will focus on the more restricted target. I offer a reply on behalf of naturalists (reductionists, reductive realists) in ethics. My case for resistance will involve identifying a three-fold ambiguity in his use of the term ‘guarantee’. I conclude with some observations about the implications of theories of reference for moral/ethical terms for the debate.
This essay is in no way a critical notice or summary of Bart Streumer’s fine book, which is full of valuable insights.1 My aim is simply to explain how, in my view, naturalists can and should resist his case for an error theory in ethics. As it happens, I agree with a great deal of what he says and I will start by detailing some of this agreement. This will help isolate exactly where I dissent from his case for an error theory. We will not worry about his argument that it is impossible to believe the error theory that he argues for, interesting though it undoubtedly is. Also, we will focus on his case for an error theory in ethics. I think he is right that many of the key issues to do with ethical properties have their counterparts when the topic is normativity more generally, but the key questions that will concern us can be more easily located on the narrower stage. Streumer should not be held responsible for the particular way I phrase what we agree about.
2 Some Points of Agreement
When someone uses a predicate like ‘is dangerous’ to describe something, they are making a claim about its nature. What is that nature? It is the respect in which items to which ‘is dangerous’ applies differ from those to which ‘is dangerous’ does not apply. As it is sometimes put, predication supervenes on nature. That’s a central part of the rationale for the use of many predicates. This is because our concern is typically not with the fact that some predicate applies per se, but with the nature in question. I chose the predicate ‘is dangerous’ to highlight this fact—we care rather a lot about the nature in question in the case of this predicate and not much about the predicate per se. Streumer and I (and many others) are cognitivists in ethics in the following sense: we hold that ethical predication supervenes on nature. A predicate like ‘is morally wrong’ applies to an action just if that action is a certain way, and that certain way we can call the property of being morally wrong. Mutatis mutandis for ‘is morally right’ and the property of being morally right, ‘is morally good’ and the property of being morally good, etc.
Cognitivists face an obvious question. How are these properties—those ascribed by ethical predicates—related to the properties ascribed by non-ethical predicates, descriptive predicates, as Streumer tags them? Or, to put the question directly in terms of properties: how are ethical properties, the properties ascribed by ethical predicates, related to descriptive properties, the properties ascribed by descriptive predicates? Although we are here using a distinction among predicates—the distinction between ethical and descriptive ones—to identify the properties we are concerned with, this in itself says nothing about whether ethical properties and descriptive properties themselves are distinct.
There are a number of arguments that strongly suggest that the properties ascribed by ethical predicates are one and all properties that can be ascribed by descriptive predicates. As we are calling properties ascribed by ethical predicates, ethical properties, and properties ascribed by descriptive predicates, descriptive properties, these arguments can be characterized as telling us that ethical properties are a proper sub-set of descriptive properties (a proper sub-set, because there are descriptive properties that are not ethical properties, being warm and being near Mars, for example). Streumer calls this view reductive realism. We might also call it reductionism, and often “naturalism” is used for views of this kind.
Streumer (p. 9f) details one argument for this conclusion that goes back, in one way or another, to Jaegwon Kim (1993).2 It rests on two very plausible claims. The first is that it is both necessary and a priori that it is impossible to have an ethical property without having one or more descriptive properties. In that sense, ethical properties are based in or dependent on descriptive properties.3 The second is that it is both necessary and a priori that ethical properties supervene on descriptive properties. It is both necessary and a priori, for example, that acts that differ in the moral properties they possess differ in the descriptive properties they possess. Streumer spells out how these two claims tell us that each and every ethical property is necessarily co-extensive with some descriptive property. The final step is to argue that, given the notion of property in play here, this means that each and every ethical property is identical to some descriptive property.
This argument, the last step especially—the one from necessary co-extension to identity—has been much debated. From my admittedly biased perspective (see, e.g., Jackson 2017: §4), I think that Streumer gives very effective replies to the objections (see especially the discussion in Chapter ii). But as this is a point of agreement between us, I will neglect the details. I would only add that even if one thought that it is consistent to hold that ethical and descriptive properties are necessarily co-extensive but distinct, the fact that each and every ethical property is necessarily co-extensive with some descriptive property would seem a powerful reason for holding that in fact each and every ethical property is identical to some descriptive property.
There is, moreover, another, equally powerful argument for the thesis that each and every ethical property is some descriptive property or other. This is the argument to be found in many who badge themselves as metaphysical naturalists. It urges that we know enough about what our world is like to know that all properties to be found in it, and especially all those to be found in it that play the kinds of causal roles that would allow us to know that some act or other possessed them, are descriptive properties.4 Ethical properties had better be descriptive properties is the thought, for otherwise we would have a very quick route to an error theory and not the interesting route Streumer lays out for us.
To avoid possible misunderstandings, it should be noted that the reductionist view that Streumer and I agree about and which we have just been discussing—the thesis that ethical properties are descriptive properties—is distinct from a reductionism about ethical predicates to the effect that they can be analysed in descriptive terms. As it happens, I also hold a reductionist view about ethical language (see, e.g., Jackson 1998: Ch. 5), but the two theses—ontological reductionism and semantic reductionism—are distinct.
3 Streumer’s Master Argument
How does Streumer get from reductive realism to the error theory? By arguing that if ethical properties are descriptive properties, then a certain kind of descriptive guarantee of the correctness of our ethical judgements should be possible, but it is very plausible that no such guarantee is possible. But, as he and I agree, the only viable view about ethical properties is that they are descriptive properties, we have, he urges, no choice but to deny that there are any ethical properties. In sum: ethical properties must be descriptive properties, but if they are, a certain kind of guarantee is possible. It isn’t possible. Ergo, there are no ethical properties.
(G) There are no descriptively specified conditions in which people’s normative judgements are guaranteed to be correct.
(H) There are no descriptively specified conditions in which people’s moral judgements are guaranteed to be correct.
Here we should think of the reference to moral judgements as being to substantive, categorical ones. For our purposes here, what to say about moral judgements like “If killing anybody for no good reason is wrong, then killing somebody for no good reason is wrong” is by the way.
Premise 1: If ethical properties exist, they are descriptive properties.
Premise 2: If ethical properties are descriptive properties, there are descriptively specified conditions in which people’s moral judgements are guaranteed to be correct (i.e. (H) is false).
Premise 3: There are no descriptively specified conditions in which people’s moral judgements are guaranteed to be correct (i.e. (H) is true).
Therefore, ethical properties do not exist.
I have already indicated my agreement with Premise 1. What about Premises 2 and 3? What to say about them depends on how we should read the notion of guaranteeing, and accordingly on how we should understand (H) itself, as I will now detail.
4 The Three Readings of (H)
It is crucial to Streumer’s case that (H) be plausible. When all is said and done, his case for the error theory rests on the claim that reductionism or reductive realism (which, recall, is the only viable view about ethical properties according to both Streumer and me) contradicts (H), and that’s only a problem for believing in ethical properties if (H) is plausible. Now he grants that “Since (H) is a negative existential claim, it is hard to be completely certain that we take it to be true” (p. 53 with “H” for “G”). My objection is that we do not take (H) to be true in the first place. Most of us rightly take (H) to be false. It is, accordingly, a virtue of reductive realism that it entails that (H) is false. Well not quite. (H) is ambiguous, depending essentially on how “guaranteed” in it is to be read, and when we sort out the three ways “guaranteed” might be read, we get three readings of (H). Two of these readings are ones on which we reasonably take (H) to be false, or so I will argue. However, the third possible reading of (H) is one on which we all pretty much agree that (H) is true. But the truth of (H) read in this third way is entirely consistent with reductionism.
(HN) There are no descriptively specified conditions that necessitate the truth of moral judgements.
Many, and I am one of them, take (HN) to be false. We take its falsity to follow from the widely agreed supervenience of the ethical on the descriptive. Here is how the thought runs in one version. Take any act, A, which is, let us say, morally right. It must have a descriptive nature (the point we noted earlier). Supervenience tells us that this nature determines without remainder A’s being morally right; this is because any act that differs from A in being morally right must differ in descriptive nature. That is to say, an act cannot have A’s descriptive nature without being morally right. Moreover, it is plausible that this descriptive nature can be captured in descriptively specified conditions. But then it follows that the descriptively specified conditions guarantee, in the necessitation sense, the truth of the judgment that A is right. Mutatis mutandis for other moral judgments.
The upshot is that the falsity of (HN) is something plausibly granted by many philosophers working in ethics, provided that they are cognitivists of some stripe or other.5 And they grant this independently of where they stand on the matters canvassed earlier—whether or not the supervenience of ethical properties on descriptive ones, plus the fact that ethical properties are dependent on descriptive properties, entails the necessary co-extension of ethical and descriptive properties, and whether or not necessary co-extension would imply, or provide strong evidence for, identity. The point is, accordingly, relatively non-contentious. The moral is clear. Implying the falsity of (HN) is arguably no bad thing, and so is no reason for holding that reductionism about ethical properties is in serious trouble, and certainly not the kind of trouble that would force one to embrace an error theory.
(HAP) There are no descriptively specified conditions that a priori necessitate the truth of moral judgements.
I suspect many will agree that (HAP) is true, perhaps citing Hume or perhaps the idea that the passage from descriptive information to moral judgment is always substantive (“non-trivial” sometimes gets used in this context) in a way that means that the passage cannot be thought of as a case of a priori necessitation. They will, accordingly, be sympathetic with the claim that implying the falsity of (HAP) would be a big problem for reductive realism (and thereby for believing in ethical properties, for those like Steumer and me who are convinced that one should be a reductionist about ethical properties). And I grant that reductive realism implies the falsity of (HAP). If ethical properties are descriptive properties, then there are descriptively specified conditions that a priori necessitate the truth of moral judgements. However, although many will be sympathetic, many will not be, and for good reason. (HAP) is far from obviously true, as I now detail.
There is a simple and powerful consideration favouring the idea that the passage from descriptive conditions to the truth of ethical judgements is sometimes a priori valid—sometimes, because of course not all passages from the descriptive to the ethical are a priori valid. Nearly everyone agrees that the step from an action’s being the breaking of a promise to its being wrong is not a priori valid or indeed necessary (in the sense of being necessarily truth preserving). As a rule, it is wrong to break promises (or, perhaps, to knowingly break promises or some variant on this, but this kind of issue is by the way here) but that’s the extent of it. The issue before us is rather whether or not some passages from the descriptive to the ethical are a priori valid, and the simple and powerful consideration I have in mind turns on the point that the fundamental issues in ethics are not matters for experiment.
Should utilitarians be average or total utilitarians? Should they, that is, hold that the right action is the one that maximises total expected utility, or that the right action is the action that maximises average expected utility? Derek Parfit (1984: Ch. 16) made this issue salient for many of us, but very few think that the way to answer it, and what to say about the “repugnant conclusion” are matters for experiment. The situation is akin to that which obtains for two-boxing versus one-boxing in Newcomb problems.6 Experiments may tell us how many philosophers and decision theorists fall into one camp and how many fall into the other camp, but not which one they ought to fall into. The same goes for questions like, Is breaking a promise bad in and of itself, or only inasmuch as it has bad expected consequences? and, Does equity have moral value over and above the fact that it tends to promote harmony? It would be misconceived to tackle these important questions by carrying out experiments. This means that, at some fundamental level, the passage from the descriptive to the ethical is not a matter for experiment. But that is to say that it is a priori (in one good sense of that term). And if that’s right, then (HAP) is false—there are descriptively specified conditions that a priori necessitate the truth of moral judgements. This is, of course, consistent with the only too obvious fact that which descriptively specified conditions a priori necessitate the truth of moral judgements is a highly controversial matter. It is, accordingly, not a problem for cognitivism that it implies that (HAP) is false.
Some will dispute what I say in the above paragraph. Although it is, I hazard, pretty much standard doctrine that fundamental disputes in ethics are not to be settled by carrying out experiments, not everyone will agree. But what matters for Streumer’s master argument is whether or not (HAP) has the kind of status that would make it the case that cognitivism’s leading via reductionism to its falsity counts as a reductio that justifies us in moving to an error theory. Streumer’s claim is that cognitivism leads via reductionism to big trouble; that’s the key to his case for an error theory. My point is that leading to the falsity of (HAP) is not big trouble; for on a plausible and widely accepted view (and, as it happens, one I accept), (HAP) is false.
(HNC) There are no descriptively specified conditions that non-controversially and definitively settle the truth of moral judgements.
If cognitivism via reductionism leads to the falsity of (HNC), that would count as leading to big trouble. If cognitivism via reductionism meant that there are descriptively specified conditions that non-controversially and definitively settle the truth of moral judgements, Streumer would have a powerful argument for an error theory.
The next section is concerned with why someone might think that cognitivism via reductionism leads to the falsity of (HNC)—or something akin to (HNC)—and why this would be a mistake.
5 Reference Theory for Ethical Terms
We can see why someone might think that cognitivism leads to the falsity of (HNC). The thought would run as follows: the only viable version of cognitivism is a version of reductionism or reductive realism. Reductive realism is the view that ethical properties are identical to descriptive properties. Any version of reductive realism worth taking seriously will have an account of which descriptive properties are which ethical properties. As descriptive properties and ethical properties are identified in the first place via the kinds of predicates that ascribe them, this account will advert to a view about how ethical terms come to ascribe the very same properties as descriptive terms. When we examine this account, we will see that it implies that (HNC) is false. For, runs the thought, the account will give us a relatively straightforward way of settling the reference of ethical terms, an account which will allow us to establish pretty definitively which descriptive property gets ascribed by, say, ‘is morally right’. Armed with this discovery, we can then settle the truth of moral judgements about the moral rightness of some action by investigating whether or not the action in question has the identified descriptive property.
An example helps spell the thought out. Suppose a simple causal theory of reference is true for ethical terms. All that’s needed to find the property ascribed by, e.g., ‘is morally right’ is to find the property that causally regulates our use of ‘is morally right’.7 Finding the property in question—the regulating property—will likely be a relatively straightforward empirical investigation of word usage. Anyone sympathetic to reductive realism will hold that the regulating property is a descriptive property—a view which is in any case independently plausible. Armed with the result of this investigation and supposing that it tells us that D is the property in question, we can then settle a whole range of contentious ethical questions concerning what is morally right simply by examining which actions have D. Life was meant to be easy, after all.
Or take a simplified version of the kind of view that Philip Pettit and I like, the view we call moral functionalism to highlight its affinities with analytical functionalism in the philosophy of mind.8 Here’s a rough sketch. We folk have opinions about which actions described in descriptive terms fall under which ethical descriptions; we hold, for example, that killing for fun is wrong. We have opinions about the inter-connections between matters described in ethical terms; we hold, for example, that having a moral duty has implications for what ought to be done. We have opinions about the connections between what’s morally right and what is done; we hold, for example, that a conviction that one ought to do something is connected with a motivation to do it.
We can think of these claims, or suitable emendations and elaborations of them, as a set of simultaneous equations to be solved by finding the identifications of ethical properties with descriptive properties that makes the equations come out true, or near enough true. In terms of this kind of theory, the thought we are scrutinizing might run as follows: solving the equations may not be quite as easy as finding the properties that causally regulate our use of moral terms, but in time surely the job can be done (if it cannot, so much the worse for moral functionalism). When the job is done, we will have an easy way to settle contentious moral questions. Investigate the descriptive properties of the actions whose moral status is being scrutinized, and then refer to the arrived at solutions!
I think that Streumer’s conviction that reductive realism is in trouble comes from the line of thought we have just been discussing. This is what is going on in Sections 18 through 21, it seems to me. He thinks that reflection on what reductionists say about how ethical terms get to ascribe descriptive properties—a topic he rightly observes they must address, for, as he says on p. 43, “If a certain normative predicate ascribes a certain descriptive property, this cannot be a brute fact.”—tells that, in one way or another, reductive realists are saddled with highly implausible views about how easy it is to settle ethical disputes. For he thinks that once they have found the needed account of how ethical terms get to ascribe descriptive properties, they will have a relatively easy way to settle on which moral judgments are correct along the lines indicated in the preceding paragraphs. They cannot acknowledge the evident truth of (HNC) and its ilk.
I think that this is a mistake. What is true is that embracing the kind of reductionism or naturalism or reductive realism Streumer and I agree about has a profound effect on how to think about debates in ethics. The impact is that once one has a full inventory of all the properties we can pick out in descriptive terms of some proposed course of action (including of course its relational properties, for example to other actions and plausibly to the attitudes of sentient creatures), the question of whether or not the action is morally right is settled; there is nothing more to find out about the action.9 In that sense, ethics becomes easier. The perennial puzzle of how to find the right way to match up, on the one hand, ethical properties thought of as being outside the picture the sciences give of what our world is like and as having no causal impact on us with, on the other hand, descriptive properties disappears. But it cannot be an objection to a theory that it makes things easier. Moreover, it remains true that there is plenty of scope for controversy within the reductionist picture of what’s going on when we debate ethical questions. Things will not be that easy. The reason is that any account of how it comes to be that some ethical property is ascribed by some descriptive term or other will itself be contentious. True, if some particular theory of reference for ethical terms of a kind consistent with reductive realism—with, that is, the view that ethical terms pick out descriptive properties—was a given, was something we nearly all agree on, ethical debates would be easy in somewhat the way outlined a few paragraphs back. We would, in that case, have a violation of (HNC) or something akin to it. But there is, of course, no such thing as the generally agreed upon theory of reference for ethical terms. Pettit and I might like it to be the case that just about everyone will soon come to agree that some version of the view we like is correct. A supporter of a version of a causal theory of reference for ethical terms which makes it easy to find out which descriptive properties are ascribed by, say, ‘is morally right’ might like it to be the case that just about everyone will soon come to agree that the view they like is correct. But none of us is holding their breath, or at least I’m not.
Blackburn S. 1985. “Supervenience Revisited.” In Hacking I. (ed.), Exercises in Analysis: Essays by Students of Casimir Lewy, 47–67. Cambridge: Cambridge University Press.
Boyd R. 1988. “How to Be a Moral Realist.” In Sayre-McCord G. (ed.), Essays on Moral Realism, 181–228. NY: Cornell University Press.
Broad C.D. 1968. “Certain Features in Moore’s Ethical Doctrines.” In Schilpp P.A. (ed.), The Philosophy of G. E. Moore, 43–67. La Salle, Ill.: Open Court.
Campbell R. and Sowden L. (eds.). 1985. Paradoxes of Rationality and Cooperation: Prisoners’ Dilemma and Newcomb’s Problem. Vancouver: University of British Columbia Press.
Jackson F. 2017. “In Defence of Reductionism in Ethics.” In Singer P. (ed.), Does Anything Really Matter? Essays on Parfit on Objectivity, 195–211. Oxford: Oxford University Press.
Jackson F. Forthcoming . “The Theory-Theory Approach to Ethics.” In Cappelen H. , Burgess A. , and Plunkett D. (eds.), Conceptual Ethics and Conceptual Engineering. Oxford: Oxford University Press.
Streumer (2017). Unless otherwise indicated, references in what follows are to this work.
In the past, I have sometimes talked of ethical properties as being grounded in descriptive ones; in using “dependent on” above, I am following Broad (1968: 60).
By “to be found in it,” I mean to be found instantiated in it. Streumer characterizes the error theory as the view that normative/ethical properties do not exist. He might have characterized it as the view that normative/ethical properties are never instantiated. On some views of properties, there is no difference between the two ways of saying it, and on those where there is a difference, the difference is by the way for our purposes, as far as I can see.
Streumer (p. 46) mentions Boyd (1988) in connection with theories of this kind. I emphasize that the view in the text above is a simplification of extant causal theories.