Cover page

Ratio Book Series

Each book in the series is devoted to a philosophical topic of particular contemporary interest, and features invited contributions from leading authorities in the chosen field.

Volumes published so far:

  1. Irrealism in Ethics, edited by Bart Streumer
  2. Classifying Reality, edited by David S. Oderberg
  3. Developing Deontology: New Essays in Ethical Theory, edited by Brad Hooker
  4. Agents and Their Actions, edited by Maximilian de Gaynesford
  5. Philosophy of Literature, edited by Severin Schroeder
  6. Essays on Derek Parfit's On What Matters, edited by Jussi Suikkanen and John Cottingham
  7. Justice, Equality and Constructivism, edited by Brian Feltham
  8. Wittgenstein and Reason, edited by John Preston
  9. The Meaning of Theism, edited by John Cottingham
  10. Metaphysics in Science, edited by Alice Drewery
  11. The Self?, edited by Galen Strawson
  12. On What We Owe to Each Other, edited by Philip Stratton-Lake
  13. The Philosophy of Body, edited by Mike Proudfoot
  14. Meaning and Representation, edited by Emma Borg
  15. Arguing with Derrida, edited by Simon Glendinning
  16. Normativity, edited by Jonathan Dancy
Title page

Notes on Contributors

Richard Joyce, Department of Philosophy, Victoria University of Wellington, Wellington, New Zealand

Hallvard Lillehammer, Birkbeck, University of London, London

James Lenman, Department of Philosophy, University of Sheffield, Sheffield

Alison Hills, St John's College, Oxford

Sebastian Köhler, c/o Postgraduate Office, School of Philosophy, Psychology and Language Science, University of Edinburgh, Edinburgh

Michael Ridge, School of Philosophy, Psychology and Language Science, University of Edinburgh, Edinburgh

Bart Streumer, Faculty of Philosophy, University of Groningen, Groningen, The Netherlands

1
Irrealism and the Genealogy of Morals

Richard Joyce

Abstract

Facts about the evolutionary origins of morality may have some kind of undermining effect on morality, yet the arguments that advocate this view are varied not only in their strategies but in their conclusions. The most promising such argument is modest: it attempts to shift the burden of proof in the service of an epistemological conclusion. This paper principally focuses on two other debunking arguments. First, I outline the prospects of trying to establish an error theory on genealogical grounds. Second, I discuss how a debunking strategy can work even under the assumption that non-cognitivism is true.

1. Introduction to moral debunking arguments

A genealogical debunking argument of morality takes data about the origin of moral thinking and uses them to undermine morality. The genealogy could be ontogenetic (like Freud’s) or socio-historical (like Nietzsche’s or Marx’s), but the focus of recent attention has been the evolutionary perspective. ‘Debunking’ and ‘undermining’ are intentionally broad terms, designed to accommodate a number of different strategies and conclusions. Sharon Street’s debunking argument, for example, aims to overthrow moral realism, while leaving intact the possibility of non-objective moral facts (e.g. those recognized by a constructivist) (Street 2006). Michael Ruse’s earlier debunking argument often looks like it has the same aim as Street’s, though on occasions he appears to try for a stronger conclusion: that all moral judgements are false (Ruse 1986, 2006, 2009). My own debunking argument has an epistemological conclusion: that all moral judgements are unjustified (Joyce 2006, 2014).

Calling all of these conclusions instances of ‘debunking’ is, in some sense, prejudicial. The rejection of moral realism, for example, counts as a debunking of morality only if one thinks that realism is somehow the natural interpretation of morality – and that is far from obvious. Any act of debunking is at the same time a vindication of something. For example, to show that all moral judgements are false would be to vindicate the error theoretic metaethical view. But I’ll let this pass, and allow ‘debunking’ to remain as a usefully vague intuitive term for these arguments.

What these disparate arguments often share is a presupposition of cognitivism. Moral judgements can be all false only if moral judgements are the kind of thing that can have truth value. Moral judgements can be all unjustified (in an epistemic sense) only if moral judgements are beliefs. In other words, the non-cognitivist – who holds that moral judgements (as mental states) are not beliefs and (as speech acts) are not assertions – will survey the debate over these debunking arguments with an unperturbed air.1

As already mentioned, the debunking argument that I have advocated (and thus, obviously, think most promising) is one with an epistemological conclusion. It is not my intention to defend or elaborate this argument further on this occasion, though it is probably best if I rehearse it briefly in order to provide a comparison class. Rather, in this paper I want to explore two different debunking avenues. First, I shall investigate what the prospects are for a debunking argument that aims to establish a moral error theory. Second, I shall question whether the non-cognitivist is warranted in his/her complacency; perhaps a debunking argument against non-cognitivism could be developed. My objectives are diagnostic rather than promotional, thus my conclusions regarding both these avenues will be non-committal. Given this, beginning with a brief look at a kind of debunking argument that I think likely to succeed will provide a useful backdrop.

2. Epistemological debunking

Recent years have seen a burgeoning of discussion about the evolutionary origins of the human moral faculty.2 Part of any such nativist explanation must be an account of what it was about moral thinking that served the reproductive purposes of our ancestors. On this point hypotheses diverge, but on most accounts moral thinking was advantageous because it in some manner enhanced their cooperative tendencies. What is striking about these nativist hypotheses is that they seem entirely compatible with the error theoretic stance; they do not appear to imply or presuppose that any of our ancestors’ moral judgements were true.

This is not so of evolutionary explanations of any kind of judgement. For example, humans quite possibly have an adaptive mechanism for distinguishing faces from other visual stimuli. But if one were to be (bizarrely!) an error theorist about faces, then the evolutionary explanation for why it might have useful for our ancestors to have this mechanism would surely fizzle. By contrast, the evolutionary hypothesis that moral thinking emerged because it strengthened social cohesion is no less plausible for the error theorist than anyone else. The best explanation of the face-identifying adaptation classifies it as a truth-tracking mechanism; the best explanation of the moral faculty does not classify it as a truth-tracking mechanism. This, it would seem, has epistemological consequences.

Most epistemological theories (and, I am tempted to add dogmatically, all sensible epistemological theories) hold that a belief’s being justified depends on its standing in one or other specific relationship to the fact that it represents. To discover that a belief does not stand in this relation to the relevant fact is to discover that the belief lacks justification. (Whether it shows that the belief has lost its justification, or shows that it was never justified in the first place, depends on which family of epistemological theories one favours.) If the evidence were to come down in support of moral nativism, then this would seem to be confirmation that our moral beliefs have their origins in a process that is not designed for truth-tracking.3 Note that this would not be a matter of conjuring up a far-fetched unfalsifiable skeptical hypothesis according to which our moral beliefs are bogus (like Descartes’ demon); it would be the confirmation of an empirical hypothesis that appears compatible with the systematic falsehood of moral judgements. Such a confirmation, I claim, undermines the epistemic standing of moral judgements.

Justification, of course, is a relative affair. My belief that p may be justified while your belief that p is not. Perhaps at an earlier time my belief that p was also unjustified; perhaps in the future it will become unjustified again (if, say, I ignore mounting evidence against the belief). Thus the conclusion that all moral beliefs are unjustified should not be interpreted as making a stronger claim than is reasonable. The proposition that a belief is unjustified does not exclude the possibility that justification can be attained or reinstated in the future. The force of the epistemological debunking argument is to issue a challenge, to shift a burden of proof.

It is often claimed that the fact that skepticism (about any object of everyday belief) cannot be refuted does not thrust that skeptical stance upon believers, so long as the non-skeptical position also cannot be refuted. Thus it is claimed that the skeptic shoulders a burden of proof: it is not enough to make skepticism irrefutable, the skeptic needs positive arguments against belief. In the event that neither the skeptical nor non-skeptical position is refutable, the non-skeptic can happily carry on with his or her everyday beliefs.

Moral nativism promises to upset this picture by providing a new hypothesis about the place of moral judgements in the world (one, moreover, potentially with empirical backing). Even those who were confident that their moral beliefs are true cannot ignore the evolutionary debunking argument, inasmuch as it is incumbent upon them either to establish that the nativist hypothesis is false or to demonstrate that moral beliefs are true even according to that hypothesis. Either way, they have some work to do. To maintain confidence in moral beliefs in advance of this work is epistemically negligent; any principle that allows one to do so is gullibility dressed up as a methodology.

When I presented this argument on an earlier occasion, I made the rash decision to label it an error theoretic conclusion (Joyce 2006, p. 223). I did this via suggesting that the label ‘error theory’ might denote a disjunction of metaethical positions: either the view that all moral judgements are false or the view that all moral judgements are unjustified. I now recant this suggestion for the following reason. Suppose all moral judgements are unjustified. This is consistent with moral judgements being true, and, moreover, objectively true; thus the claim that all moral judgements are unjustified is compatible with moral realism. But the error theory had better not be compatible with moral realism, therefore the view that all moral judgements are unjustified had better not be sufficient for an error theory.4 It is preferable to keep our metaethical theories separate and be clear that the conclusion to this debunking argument is epistemological in nature. The thesis that all moral judgements are unjustified lacks a label, though it is perfectly acceptable to call it a version of moral skepticism.5

I think the epistemological debunking argument outlined in this section has legs. But the benefits of establishing the error theory by stretching the extension of the label in the manner just described (and just renounced) are, to quote Russell, the advantages of theft. I turn now to exploring the prospects of using a debunking argument to establish the moral error theory through honest toil.

3. Error theoretic debunking

Certainly there are circumstances where learning about the origin of a belief can reveal that belief to be false. My belief that hypnosis cannot instil genuine beliefs in people is falsified if I discover that I was caused to have this belief through hypnosis. But clearly nothing so swift and sneaky as this is going to work in the case of moral judgements and moral nativism. The moral judgement that promise-breaking is wrong, say, simply doesn’t imply anything about its own origins in the way that the belief about the limits of hypnosis does. Rather, we shall see, the error theoretic debunking argument depends on a principle of parsimony.

Let us start with Street’s debunking argument, whose conclusion is that moral realism is probably false. She argues that the moral realist, confronted with the truth of moral nativism (we are imagining), faces a dilemma concerning the relation between our moral judgements (products of the distortions and contingencies of our evolutionary ancestry) and the supposed realm of objective moral facts. On the one hand, if there is no relation then it would be an astonishing coincidence if many of our moral judgements were even approximately true – a conclusion supposedly disagreeable to the realist. The problem with the other horn of the dilemma is that it is, according to Street, empirically dubious. I have already noted that the usual nativist hypotheses see the ancestral adaptive pay-off of having a moral faculty in terms of enhancing certain cooperative tendencies, not in terms of tracking moral truths. Street thinks this ‘adaptive link hypothesis’ is superior to any truth-tracking hypothesis for three reasons: it is more parsimonious, it is clearer, and it is more illuminating of the phenomenon it seeks to explain (2006, p. 129). Street’s irrealist conclusion might be put as follows: ‘There are no objective moral facts.’ Yet she doesn’t deny the possibility of moral facts – they will simply be of a constructivist nature.

What good, one might ask, is this to an error theorist? Let me approach this by quickly comparing Ruse’s argument. Ruse maintains that being imbued with a kind of objectivity is the whole point of moral thinking, evolutionarily speaking. Morality serves its adaptive function of strengthening our motivation to cooperate by seeming to be imbued with a kind of inescapable external prescriptivity. ‘It is precisely because we think that morality is more than mere subjective desires that we are led to obey it’ (Ruse 1986, p. 103). But, Ruse argues, this objectivity is an adaptive illusion. He argues for this latter claim via an implicit appeal to parsimony: once we have explained why morality seems to be objective, there is simply no call for any further explaining in terms of positing a realm of objective moral facts. At this point the conclusion to Ruse’s argument looks very similar to that of Street’s, reached by somewhat different means. He writes: ‘[M]orality is a collective illusion foisted upon us by our genes. Note, however, that the illusion lies not in the morality itself, but in its sense of objectivity’ (1986, p. 253).

However, Ruse’s discussion contains elements that aren’t present in Street’s thinking, opening the door to the stronger error theoretic conclusion. For a start, his emphasis on the adaptive importance of the objectivity with which moral prescriptions are infused is not something Street mentions. A strong thread running through his argument is that moral realism is written into the phenomenology of moral experience. But he goes further, apparently moving from phenomenology to semantics: ‘Ethics is subjective, but its meaning is objective’ (Ruse 2006, p. 22); ‘[W]hat I want to suggest is that … the meaning of morality is that it is objective’ (Ruse 2009, p. 507). The move from phenomenology to semantics is not something to which one can help oneself for free, but at the same time it’s not unreasonable to assume that the meaning of a term is going to reflect our experience of the phenomena denoted by that term. If humans are designed by natural selection to experience morality as objective, then this perhaps makes more plausible the already not-ridiculous thesis that objectivity is an essential quality of morality, conceptually speaking. With this thesis operating as a bridging premise, one can get from the sub-conclusion ‘There are no objective moral facts’ to the conclusion ‘There are no moral facts.’ (The two propositions would stand in the same relation as ‘There are no four-sided squares in the box’ and ‘There are no squares in the box’ stand in.)

This bridging premise is a key part of this error theoretic debunking argument. Street rejects it, hence her conclusion is not error theoretic. And of course it is an extremely controversial thesis, over which much metaethical ink has been spilt. Part of the problem is that the term ‘objectivity’ is not well defined, and it gets used differently in different areas of philosophy. (For discussion see Joyce 2007a, 2009.) The notion that Ruse seems to have in mind is that of moral prescriptions having a kind of external authority: we feel bound to follow them because we experience them as not of our own making (unlike, say, the non-objective prescriptions of fashion).6 Many philosophers will agree with Ruse that we tend to experience moral norms in this manner, though only some of them (a good number, to be sure) will go along with the stronger claim that this kind of objectivity is essential to morality, such that a normative framework stripped of this objectivity wouldn’t even count as a ‘moral’ system. Those that do support the stronger semantic claim include both realists (who think that this objectivity can be satisfied) and irrealists (who think that it cannot be satisfied).

Ignoring, for a moment, the difficulty of establishing this bridging premise, let me try to reconstruct the argument that employs it. Whether this actually reflects Ruse’s reasoning is not my primary concern, but I will continue to attribute it to him if only for the sake of argument. The argument turns on the application of a parsimony principle:

  1. Objective moral facts aren’t required to explain anything.
  2. If some type of fact plays no explanatory role, then this is ground for disbelieving in this type of fact.

There are deep questions to be raised about both these premises, which I shall turn to in a moment, but initially I want to discuss them just sufficiently to motivate the need for a third premise.

In a sense, nothing is required to explain anything. What I mean by this quizzical claim is that one always has choices in how to explain any phenomenon. If the cat knocks over the vase, one can always explain the broken vase without employing the concept cat. Instead of using biological or zoological categories, one could (in principle) make reference to a conglomeration of organic chemicals moving about the room, or a swarm of particles and energy. Thus the concept cat isn’t required in any explanation of anything. But this hardly means that cats are explanatorily impotent. The crucial point is that cats are reducible to entities that are described at other theoretical levels: chemistry or physics, for example. Thus, even if it were true that reference to objective moral facts isn’t needed to explain anything, it wouldn’t follow that objective moral facts are explanatorily impotent. For this conclusion a further premise must be added:

  1. Objective moral facts aren’t reducible to any facts that do have explanatory roles.

These premises yield the sub-conclusion:

  1. Therefore, there is ground for disbelieving in objective moral facts (i.e. there is ground for rejecting moral realism).

We can now add the bridging premise:

  1. Morality is essentially objective.

And the error theoretic conclusion follows:

  1. Therefore, there is ground for disbelieving in moral facts.

Every single one of the premises is problematic. Let us start by considering premises 1 and 3 together. Ruse’s argument for premise 1 is often presented via an analogy (Ruse 1986, pp. 256–257, 2006, pp. 22–23, 2009, pp. 504–505). He refers to the spike of interest in séances in Europe in the aftermath of World War 1. Imagine a grief-stricken mother attending such a séance, during which time she comes to believe that her dead son has spoken to her from beyond the grave. We can explain everything that needs explaining about this belief by reference to psychological and sociological factors; there is no need to suppose that the belief might be true. Similarly (Ruse thinks), moral nativism explains everything that needs explaining about why humans judge certain actions to have objective moral status; there is no need to suppose that these judgements might be true.

The weakness of the analogy is brought out when we attend to premise 3. In order to suppose that the mother’s belief is true, we would have to presume that the world contains supernatural forces, post-mortem consciousness, ghosts, etc. – that is, some pretty spooky ontology. It is far from obvious that this is what is required to suppose that judgements about objective morality are true. Moral naturalists (of an objectivist stripe) will often identify moral properties with naturalistic properties that we already accept in our ontological scheme. A utilitarian, for example, may identify moral goodness with happiness.7 By contrast, any attempt to identify, say, ghosts with some cluster of naturalistic properties looks hopeless. In other words, the analogue of premise 3 for ghosts looks obviously true. But premise 3 as it stands for objective moral properties will be doubted by many, and therefore cannot stand without argumentative support.

Rather than return attention to the bridging premise 5, let us consider dropping all mention of objectivity, which would allow premises 5 and 6 to evaporate. The revised argument is as follows:

  1. 1*.  Moral facts aren’t required to explain anything.
  2. 2.  If some type of fact plays no explanatory role, then this is ground for disbelieving in this type of fact.
  3. 3*.  Moral facts aren’t reducible to any facts that do have explanatory roles.
  4. 4*.  Therefore, there is ground for disbelieving in moral facts.

The stripped down argument looks a lot like one that Gilbert Harman famously uses to frame his discussion (1977). Harman doesn’t endorse the argument, though; he rejects premise 3*, arguing that moral facts are reducible to facts about what reasons we have for acting, which (he thinks), properly understood, are empirical phenomena. Nor does Harman place any emphasis on moral nativism, which for Ruse is the main consideration lying behind the first premise. Harman, rather, appeals to developmental factors to explain how moral judgements might arise from non-truth-tracking mechanisms. This difference doesn’t matter to our current concerns; what is significant is that moral judgements can be genealogically explained in a way that makes no reference to their being true. This supports the first premise presumably in the following manner. If moral judgements can be fully explained without reference to moral facts, then this casts immediate doubt on whether moral facts are needed to explain anything. (Likewise for Ruse, mutatis mutandis, concerning objective moral facts.) It seems to me that this move is reasonable, for what possible instance would we recognize of a moral fact playing a role in explaining phenomenon X, where this act of recognition did not involve the use of a moral judgement? Moral facts appear to have what Crispin Wright calls ‘narrow cosmological role’ (1992): their causal impact always involves someone’s having made a judgement concerning their presence. (Cats, by contrast, have wide cosmological role, affecting the world in a myriad of judgement-independent ways: meowing, casting shadows, producing kittens, knocking over vases.) If moral explanations (e.g. ‘Fred broke the promise because he’s wicked’) always depend on someone’s having made a moral judgement, but moral judgements can always be fully explained without reference to moral facts, then the explanatory potency of moral explanantia (e.g. Fred’s wickedness) is an illusion.

practical authorityreasons

Harman’s presentation of the argument does not explicitly endorse premise 2. Summing up his argument (before embarking on his rejection of 3*) he writes that ‘it remains problematic whether we have any reason to suppose that there are any moral facts’ (1977, p. 23). Imagine it turns out that we do not have any reason to suppose that there are any moral facts. This wouldn’t automatically amount to our having a reason to suppose that there are not any moral facts. The crucial difference is between premise 2 and the weaker 2B:

  1. 2.  If some type of fact plays no explanatory role, then this is ground for disbelieving in this type of fact.
  2. 2B.  If some type of fact plays no explanatory role, then we have no ground for believing in this type of fact.

But the adage is not to be taken as gospel, for there are certainly circumstances where absence of evidence is evidence of absence: most obviously, conditions in which one could reasonably expect to have evidence (see Sober 2009a, p. 64). For example, if there were a leopard hiding in this room somewhere, it would be reasonable for me to expect to encounter some evidence of the fact; the absence of any such evidence provides evidence of a leopard’s absence.

Similarly, if we had some independent information about the probability of there being moral facts, then we might be able to support the stronger conclusion. Suppose we knew that moral facts were improbable, but took our moral judgements nevertheless to provide some support for their obtaining. The discovery that these moral judgements stem from a non-truth-tracking source would undermine this support, thus putting us back in the position of judging moral facts improbable. (This is not exactly disbelief, of course, yet framing the issue in Bayesian terms of degrees of belief is probably how the more nuanced presentation should proceed.8) Yet assessing the prior probability of moral facts obtaining is also a very difficult question regarding which there will be nothing remotely like a consensus among philosophers (see Brosnan 2011, p. 55). So this route seems even less propitious for the error theorist than that sketched in the previous paragraph.

banishment

Even if premise 2 is defensible, however, we have seen that there are many other ‘if’s in an error theoretic debunking argument of this sort, and the argument strays a long way from the genealogy of morals with which it began. Ultimately, moral nativism may find a place as a premise in an error theoretic debunking argument, but it will be a supporting role; the main actors will be propositions of a metaethical nature.9