Hume on causation: against the quasi-realist interpretation

ABSTRACT

In recent years, a number of philosophers have promoted a quasi-realist (or projectivist) interpretation of Hume's theory of causation. In this paper, we argue against the quasi-realist interpretation of Hume, on the grounds that there is a direct clash between a fundamental element of Hume's system (his empiricist theory of content) and one of the main constraints that governs any form of quasi-realism (and so a fortiori, quasi-realism about causation).

KEYWORDS:

• Hume
• Blackburn
• projectivism
• quasi-realism

1. Introduction

In recent years, a number of philosophers have promoted a quasi-realist (or projectivist) interpretation of Hume’s theory of causation (Blackburn 1984 chapter 6; 1990/2000; 2008 chapter 3), Coventry (2006), Beebee (2006 chapter 6; 2007). In this paper, we argue against the quasi-realist interpretation of Hume, on the grounds that there is a direct clash between a fundamental element of Hume’s system (his empiricist theory of content) and one of the main constraints that governs any form of quasi-realism (and so a fortiori, quasi-realism about causation).¹ We proceed as follows. In §2, we very briefly outline Hume’s empiricist account of concepts and content (his theory of ideas), highlighting in particular an underappreciated feature of it that will prove to be crucial later in the paper. In §3, taking Blackburn’s exposition of quasi-realism in the ethical case as our guiding example, we illustrate a self-imposed ‘no-circularity’ constraint that quasi-realism about any subject matter operates under. In §4, we bring the reflections of §§2-3 together to show that Hume’s empiricist theory of concepts and content generates a problem for quasi-realist interpretations of his views on causation. Finally, §5 contains a short summary and conclusion.

2. Hume’s empiricist theory of concepts and content

Hume divides states of mind – ‘perceptions’ as he calls them – into impressions and ideas. Ideas are effectively concepts or the meanings of words: the determinants of content, either of utterances or mental states.² Impressions are experiences: perceptual experiences (which Hume calls ‘outward sentiments’ or ‘impressions of sensation’), such as the experience one has when one sees a cloud or touches a table; and introspectible experiences (which Hume calls ‘inward sentiments’ or ‘impressions of reflection’), such as experiences of joy or anger.

The fundamental tenet of Hume’s empiricism is his claim that all ideas – all concepts, all determinants of content – are copies of resembling impressions: ‘By ideas I mean the faint images [of impressions] in thinking and reasoning’ (1739, 1). Since we have the concept of a golden mountain even though we have no experience or impression of one, Hume introduces a distinction between simple and complex ideas and impressions. Although ‘many of our complex impressions never are exactly copied in ideas … every simple idea has a simple impression, which resembles it; and every simple impression a correspondent idea’ (1739). In the case of simple ideas and impressions ‘they are exact copies of each other’ (1739.). We have the idea of gold, derived from our impressions of that colour, and we have the idea of a mountain, derived from our impressions of such, and these are the ingredients for our idea of a golden mountain. Hume’s fundamental empiricist principle – the Copy Principle – can then be formulated as follows: all simple ideas are copied from some resembling impression, and all complex ideas are composed of simple ideas. Ultimately, then, all concepts and ideas are in the final resort derived from impressions or experiences.

This much will be very familiar to students of Hume. A point whose significance is perhaps underappreciated is that Hume explicitly regards the relationship between ideas and the impressions they are copied from as causal in nature.³ First, he notes that ‘every simple impression is attended with a correspondent idea, and every simple idea with a correspondent impression’ (1739, 4). On the basis of this ‘constant conjunction’ Hume concludes ‘that there is a great connexion betwixt our correspondent impressions and ideas, and that the existence of one has a considerable influence upon that of the other’ (1739). But do the impressions depend on the ideas or do the ideas depend on the impressions? Hume notes that we find ‘by constant experience, that the simple impressions always take the precedence of their correspondent ideas’ (1739, 5), but not vice-versa. To give a child an idea of a particular colour or taste we give him the relevant experience, but it would be absurd ‘as to endeavour to produce the impressions by exciting the ideas’ (1739). Hume concludes that dependence of ideas on impressions is causal:

The constant conjunction of our resembling perceptions, is a convincing proof, that the one are the causes of the other; and this priority of the impressions is an equal proof, that our impressions are the causes of our ideas, not our ideas of our impressions. (1739, 5)

Hume thus subscribes to a kind of causal theory of concepts and content: the content of a mental state is determined by the impressions that cause its constituent ideas; the sense of a sentence is determined by the impressions that cause the ideas associated with its constituent words; and the same goes for judgement (the expression of a mental state by the utterance of a sentence).⁴

We’ll explain the significance of this fact in §4. First, though, we need to think about the character of quasi-realism in general and in particular the nature of the constraints that govern the quasi-realist project. This is the task of §3.

3. The quasi-realist project and the no-circularity constraint

In order to get clear on the character of quasi-realism in general, we’ll concentrate in this section on the version of quasi-realism that has been most fully developed by the leading proponent of the theory, Simon Blackburn’s account of moral judgement.

The first thing to note about quasi-realism is a change in focus. Earlier views in the broadly non-cognitivist tradition (e.g. the emotivist theory developed in chapter 6 of Ayer [1936]) saw themselves as engaged in a fundamentally analytic exercise: Ayer sees himself as attempting to give an account of what moral statements mean that coheres better with the conventions governing our actual moral language than either the naturalist theories of moral language that fall prey to Moore’s open question argument or Moore’s own account of moral statements as unanalysable and sui generis. Blackburn is motivated in part by a desire to avoid the charge of committing the ‘speech-act fallacy’: the fallacy of inferring from the fact that the moral judgement that ø is right performs the speech act of commending ø or expressing an attitude of approval towards it that it does not also perform a speech act of describing ø or expressing a belief about it. (This would be fallacious because we are perfectly familiar with the phenomenon of an utterance performing two different sorts of speech act simultaneously: ‘There’s an angry bull in the next field’ both describes what’s in the next field and issues a warning not to enter it). Blackburn writes:

[T]his fallacy need not be committed. First of all, an expressive theory should not infer that the attitude gives the role of the saying from the fact that it is expressed when the saying is made. So long as the attitude may give the role, the argument for saying that it does is the superior explanation of the commitments which we then arrive at. There is no inference of the form ‘this attitude is expressed, so these remarks have no truth-conditions’, but only ‘this attitude is expressed; if we see the remark as having no truth-conditions the philosophy improves [because e.g. we are relieved of the metaphysical and epistemological debts of cognitivism]; so let us see the remark as expressive rather than descriptive’. There is no fallacy here. (Blackburn 1984, 169–170)

So the idea is that we assume – as a hypothesis – only that moral utterances express non-cognitive attitudes, and on that basis alone give an explanation of the characteristic features of moral judgement. If we are able on that basis to explain why there is nothing problematic about e.g. applying the notions of truth and falsity to moral judgements, or embedding moral statements in the antecedents of conditionals, we will have successfully defended an alternative to full-blown moral realism: a quasi-realist alternative that is realist because of what it earns (the right to the notions of moral truth, moral fact, and the rest) but quasi-realist because of the slender basis on which it does so (e.g. the explanation makes no appeal to moral states of affairs or propositions about such states of affairs).

Before articulating in more detail the quasi-realist view and the constraints which it imposes on its own explanation of moral judgement, it might help to say briefly why the full-blown realist alternative is held by Blackburn to be problematic.

Consider the questions: ‘What are we doing when we make moral judgements?’, ‘Why do we do what we do when we make moral judgements?’, and ‘Are any of the things we do when we make moral judgements justified?’ According to moral realism, the key notion in answering these questions is that of a moral state of affairs: we assume that moral judgements express moral beliefs, and we assume also that the nature of these beliefs is to be explained via an account of the states of affairs whose obtaining is regarded as necessary and sufficient for their truth.

According to Blackburn, the moral realist style of explanation which takes moral states of affairs as explanantia essentially founders on a dilemma. If the putatively moral states of affairs that figure in the explanantia are regarded as natural (i.e. as belonging to the subject matter of the natural and/or social sciences), G.E. Moore’s open-question argument (Moore 1903) suggests that we thereby fail to give an adequate account of the content of moral beliefs: a judgement about a natural state of affairs appears to lack the action-guiding force of a genuine moral judgement. But if the putatively moral states of affairs are held to be non-natural, the kinds of metaphysical and epistemological problems highlighted in J.L. Mackie’s celebrated ‘argument from queerness’ threaten the tenability of the view (Mackie 1977).⁵

In order to avoid the dilemma that disables realism, Blackburn suggests that we attempt to explain the nature of moral judgement via an explanatory route that does not rely on the notion of a moral state of affairs. In place of moral states of affairs, the quasi-realist seeks explanations that give a central role to distinctively ethical non-cognitive sentiments. In constructing the explanation, the projectivist allows appeal to natural properties and states of affairs, beliefs about natural properties and states of affairs, and non-cognitive attitudes and sentiments towards natural properties and states of affairs. Moral states of affairs play no role in the explanation, nor do moral beliefs (beliefs whose truth would be a matter of the obtaining of a moral state of affairs).

In order to understand the character of quasi-realism, it is important to appreciate that not assigning moral states of affairs or moral beliefs an explanatory role in the story about the nature of moral judgement does not involve denying that there are such things as moral states of affairs or moral beliefs: if, on the basis of the more austere set of explanatory materials we are able to tell a story which fully legitimates ordinary practice’s commitment to such notions as moral truth and moral fact we will not have contradicted ourselves by simultaneously asserting and denying their existence. Not assuming that there are moral facts at the outset of an explanatory project does not commit one to denying their existence if that explanatory project is successful.⁶

Quasi-realism, then, is the project of recovering, from an austere projectivist starting point, the features of moral thought and talk – including moral facts and moral truth – that the realist simply takes for granted as a legitimate starting point for his explanation of moral judgement. The word ‘realism’ appears in the name of the view because at the level of ordinary moral practice it enables us – with complete justification – to say that some moral judgements are true, that moral judgements express beliefs, that it is a fact that torturing animals is wrong, and that this fact is in a philosophically innocent sense mind-independent. But ‘realism’ in the name is qualified as ‘quasi’ because the philosophical explanation that earns us the right to notions of moral truth and moral fact proceeds, not in terms of distinctively moral states of affairs, but in terms of the more austere set of materials that doesn’t include them. As Blackburn puts it:

Projectivism is the philosophy of evaluation which says that evaluative properties are projections of our own sentiments (emotions, reactions, attitudes, commendations). Quasi-realism is the enterprise of explaining why our discourse has the shape it does, in particular by way of treating evaluative predicates like others, if projectivism is true. It thus seeks to explain, and justify, the realistic-seeming nature of our talk of evaluations – the way we think we can be wrong about them, that there is a truth to be found, and so on (Blackburn 1984, 180).

A good way to highlight the constraints governing the quasi-realist enterprise is to focus on the descriptions of moral agents that are permitted to figure among the explanantia. If we think of such agents schematically, as input-output devices, the admissible inputs can include beliefs about natural states of affairs (non-moral beliefs), desires directed towards natural states of affairs, and the admissible outputs can include attitudes or sentiments of approval and disapproval. Blackburn thinks that this restricted set of materials provides a sufficient basis for e.g. a solution to the Frege-Geach Problem and the task of constructing a philosophically respectable notion of moral truth.⁷

For our purposes, the important point to note is that the quasi-realist explanantia must not include moral facts or states of affairs, or beliefs with moral content. Entitlement to these is supposed to be the upshot of the quasi-realist explanatory project: any view which included them among the explanantia would be a form of full-blown (i.e. non-quasi) realism. This illustrates what we might call the no-circularity constraint that governs the ethical quasi-realist’s explanatory project. The attempt to ‘place’ ethics ventured by the projectivist quasi-realist ‘must not draw upon the existence of values, obligations and the rest’ (1993, 372). The distinction between the point occupied at the beginning of the explanatory task and the position occupied at the end is again crucial. Although we won’t end up denying that there are values, obligations, and so on, ‘in the context of the placing operation, use of ethical predicates is banned’ (1993, 383n.17).⁸ This ban is essentially what we’ve called the ‘no circularity constraint’.⁹

The no-circularity constraint generalises and applies to any form of quasi-realism, no matter what the subject matter: in a quasi-realist explanation of colour judgement, for example, no colour facts, colour states of affairs or propositions about the colours of things must be invoked or presupposed in the explanantia, on pain of the account lapsing into a form of full-blown (non-quasi) realism about colour judgement. To put it in its most general form, in a quasi-realist explanation of the judgement that Fa and its cognates, the explanantia must not include propositions, facts, or states of affairs of the form that Fa.¹⁰

4. Humean ideas and causal quasi-realism

Although he refers to Hume’s theory of causation as ‘the classic projective theory in philosophy’ (1984: 210), Blackburn does not develop the quasi-realist account of causation in anything approaching the level of detail of his quasi-realist account of ethical judgement. However, he does say enough to give us a relatively clear sense of the general shape of the account. Whereas in the moral case, in the explanantia agents are viewed as input-output devices with admissible inputs allowed to include or presuppose non-moral states of affairs, facts, and beliefs and desires with non-moral content, in the explanantia in the causal case thinkers will be viewed as input-output devices with admissible inputs allowed to include facts about observed regularities and regular successions of similar events, and non-causal beliefs i.e. beliefs about observed regularities and regular successions of similar events. And whereas in the moral case, the outputs will include attitudes or sentiments of approval and disapproval, in the causal case the key outputs will be habits of expectation: inclinations to expect the occurrence of future events similar to those in the regular successions of events we have hitherto experienced (Blackburn 1984, 211). After exposure to ‘a regular succession of events, the mind ‘changes functionally’: it becomes organised so that the impression of the antecedent event gives rise to the idea of the subsequent event’ (1990/2000, 107).¹¹

We are now in a position to state our objection to the interpretation of Hume as a causal quasi-realist. According to the causal quasi-realist, admissible inputs include non-causal beliefs, centrally beliefs about observed regularities in nature and regular successions of similar events. Given the Copy Principle and Hume’s explicit commitment to the causal nature of copying, the content of these beliefs, like all others, is determined by facts about the causal relations that obtain between their constituent concepts (ideas) and the impressions from which they are copied. Given Hume’s theory of concepts, therefore, the causal quasi-realist project is doomed to violate the no-circularity constraint. Invoking even non-causal beliefs as admissible inputs in the austere descriptions of causal thinkers in the explanantia will involve presupposing causal facts: the facts about the causal relations between impressions and ideas that determine their content.^12,13

5. Conclusion

Among the facts which the quasi-realist about causation allows as explanantia in his explanations are (non-causal) beliefs about observed regularities of similar events. Like beliefs of any sort, the contents of these beliefs will involve concepts (or ideas, in Hume’s terminology). According to the Copy Principle, these ideas are ultimately copied from impressions. So, given that for Hume copying is causal, the quasi-realist’s appeal to the beliefs about observed regularities involves an appeal to causal facts. This means that at the foundation of Hume’s system lies an appeal to causal relations that makes it impossible for any quasi-realist interpretation of his account of causation to satisfy the no-circularity constraint. In other words, if we view Hume as a quasi-realist about causation there is a direct clash between the central tenet of his empiricist theory of ideas (the Copy Principle) and the no-circularity constraint that governs quasi-realism. That a philosopher of Hume’s standing would have to be regarded as failing to notice a clash of this magnitude in order to be viewed as subscribing to a projectivist approach to causal judgement constitutes a strong reason for thinking that he did not hold a quasi-realist view of causation.

Acknowledgements

For comments and discussion, we are grateful to John Shand, Olivia Sultanescu, Zach Weber and Alan Weir. Thanks, too, to seminar audiences at the University of Otago, and to an anonymous referee for Inquiry.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 To be clear, it’s no part of our brief to suggest that quasi-realism about causation is itself problematic (we remain neutral on that here): the paper has the more limited aim of arguing that there are good reasons for not interpreting Hume as a causal quasi-realist.

2 So ‘concept’ here is used in a non-Fregean manner i.e. as concerning items at the level of Sinn and not the level of Bedeutung. To borrow from a formulation of John McDowell’s, Hume’s ideas are determinants of the thoughts expressible by sentences containing the associated words (McDowell 1987a, 87). Hume himself identifies ideas and thoughts at (1748, 18). It may be that thinking of Hume as putting forward a theory of meaning and content in this way is not uncontroversial, but we don’t enter into the matter further here since our protagonists have a similar view of Hume in this regard. See e.g. Beebee (2006, 111).

3 Some (but not all) textbooks and commentaries explicitly acknowledge the causal nature of the Copy Principle (e.g. Stroud (1977, 31); Rosenberg (1993, 65–66); Noonan (1999, 62); Baillie (2000, 21); Bennett (2001, 203); Broughton (2006, 52); Garrett (2015, 72); Morris and Brown (2023 §4.1)) but as far as we know we are the first to suggest that it undermines quasi-realist interpretations of Hume on causation. Landy (2006, 122–126) emphasizes the causal nature of copying and argues that it doesn’t undermine Hume’s view when he’s construed as an error theorist: our argument below concerns a different target (the interpretation of Hume as a non-error theoretic projectivist).

4 The claim that the copying relation is causal in nature appears only in Hume (1739) and is not repeated in Hume (1748). However, there’s nothing to suggest that this reflected a change of mind on Hume’s part as opposed to being a consequence of his seeking a presentation that would be more attractive stylistically. And the features of the case that lead Hume to assert the causal nature of the relation in the Treatise would do so also by the lights of the views on causation presented in the first Enquiry.

5 Mackie himself holds that any form of moral realism faces these metaphysical and epistemological challenges, not just non-naturalist variants.

6 For an especially clear statement of this, see Blackburn (1993, 372).

7 See Miller (2013) for an account of these. In this paper, we stay neutral on whether the quasi-realist enterprise is successful in regards to these.

8 The no-circularity constraint should be distinguished from the kind of methodological naturalist constraint that Blackburn subscribes to in passages such as:

[T]he problem is one of finding room for ethics, or placing ethics within the disenchanted, non-ethical order which we inhabit, and of which we are a part. 'Finding room' means understanding how we think ethically, and why it offends against nothing in the rest of our world-view for us to do so. It does not necessarily mean 'reducing' ethics to something else (1998, 49).

The constraints are complementary but nonethless distinct. For example, an explanation (in the moral case) that helped itself to facts about Cartesian souls would presumably violate the methodological naturalist constraint but not the no-circularity constraint (so long as it restricted itself to non-moral facts about Cartesian souls); and an explanation (in the case of quasi-realism about causation) that helped itself to causal facts would violate the no-circularity constraint but not the methodological naturalist constraint (on the assumption that causal facts are natural facts).

9 Note that although Blackburn imposes the ban – the no circularity constraint – on the quasi-realist explanatory project, he does not think that it extends to the justificatory project we engage in when we participate in first order normative theorising. In that justificatory project ‘we are back working from within. We are no longer playing the fake externalist game of trying to certify values without using values’ (1996, 89). The idea that the explanatory and justificatory projects are subject to different constraints could of course be questioned, but we will simply grant it to Blackburn here. Our argument is that given the no-circularity constraint Blackburn imposes on the explanatory account, there is a problem for the interpretation of Hume as engaging in it. Note also that the ban on ethical materials in the ‘naturalistic toolkit’ (1993, 378) the projectivist employs doesn’t make him a reductionist naturalist of the sort attacked by Moore: the projectivist, recall, is engaged in explanation rather than analysis.

10 John McDowell (in his 1987b) in effect argues that ethical quasi-realism violates the no circularity constraint because of an inability to properly characterise moral sentiments without using moral concepts in a way that is prohibited by the constraint. We are content to stay neutral on that issue here: our objection applies only to causal quasi-realism in the context of the Copy Principle and the Humean view of ideas outlined above. The key thing to note here is that Blackburn accepts the prohibition and suggests that it can be respected by engaging in ‘naturalized psychology’(1993, 374).

11 Beebee puts it in terms of the projection of ‘habits of inference’(2006, 143).

12 The only routes open to the Humean causal quasi-realist would be to jettison the Copy Principle, give up the idea that copying is causal, or give up the idea that there are any admissible doxastic inputs to the input-output function that represents the causal reasoner in the relevant explanantia. None of these look attractive, though. The first would involve giving up the fundamental empiricist tenet of Hume’s system, the second would involve denying a patently cogent piece of first-order causal reasoning, and the latter would involve emptying the toolkit deployed in causal-quasi realist explanations of any belief whatsoever.

13 Our argument requires the assumption that for Hume, content is determined by the causal relations that obtain between ideas and the impressions from which they are copied. It might be objected that Hume could avoid the problem this generates by adopting a position on which (a) impressions cause ideas, but (b) for impressions to give content to ideas it is enough if there is just a pattern of regular occurrence. This would allow the causal quasi-realist to avoid making assumptions that violate the no-circularity constraint.

In reply, for the sake of argument we concede that this might be a consistent position for a causal quasi-realist to adopt in order to avoid circularity. Recall, though, that our aim in the paper is not to argue against causal quasi-realism per se, but merely to argue that there are good interpretative reasons against the ascription of such a view to Hume. And there does appear to be a problem with the ascription to Hume of a position incorporating (a) and (b). Given that copying is causal, the suggestion effectively credits Hume with the view that ideas are copies of impressions while denying that copying as such plays a role in determining content. Interpreting Hume along these lines sees him as introducing the notion of copying while assigning it no role in his theory. Why would he do this? Unless this question is answered (or some other role for copying identified in Hume’s theory), this interpretation of Hume seems somewhat strained, to say the least. (We are grateful to an anonymous referee for raising this objection).