Hume’s account of causation has a good claim to being one of the most influential views in the history of philosophy. It not only set much of the agenda for large swathes of analytic philosophy in the 20th century and beyond, but it also awoke Immanuel Kant from his “dogmatic slumber” – as he put it in his Prolegomena to any Future Metaphysics – and prompted him to write the mighty Critique of Pure Reason, itself a hugely influential work and arguably the starting-point for the continental tradition in philosophy.
So why has Hume’s view on causation proved to be so influential? Well, let’s start with the state of play in philosophy at the time Hume was writing. The dominant view of causation at the time was a part of what Edward Craig (in The Mind of God and the Works of Man calls the “Image of God doctrine”. The idea here is, as the name suggests, that we are made in God’s image: our mental faculties are of course rather feeble compared to God’s, but they are of the same kind as God’s. If you were in the grip of the Image of God doctrine, you might think something like this. Our mental faculties are at their most perfect – their most God-like – when we’re engaged in a priori reasoning, for example when we’re constructing a mathematical proof. And in a mathematical proof, we can (if we’re really good at maths) just “see” or “intuit” that each successive stage of the proof follows from, or is entailed by, the preceding stage. So, if our mental faculties generally are God-like, then the same kind of thing must be going on when we turn our attention to the causal structure of the world. At least in principle, if I look at some event – the cue ball hitting the black ball in snooker, say – I can tell, just by observing that event, what must happen next: I can infer, on the basis of just that experience, what the collision will cause, just as I can in principle tell just by looking at a mathematical theorem what follows from it.
Hume’s fundamental insight when it comes to causation is that that story cannot possibly be right. No matter how hard I look, and no matter how much I know about the size and shape and weight of the balls and their position on the table, nothing whatsoever follows about what the collision is going to cause. Of course, what I expect to happen is that the black ball will move off in a certain direction and (let’s suppose) land in the corner pocket. But that is not something I can deduce just from careful observation of the collision. As Hume puts it: “If we reason a priori, any thing may appear able to produce anything”.
A crucial – and perhaps the most influential – part of Hume’s argument is that we cannot observe the causal relation itself. When I look at the snooker table, all I see is the cue ball hitting the black ball, followed by the black ball moving off: I don’t see any “necessary connection” between the two. Because I have frequently seen this conjunction of events before, I come to expect the black ball to move – but this expectation is generated not by any kind of a priori inference. Rather, it is generated by “habit or custom”: we are endowed with a kind of instinct that prompts us to expect, on the basis of past “constant conjunctions”, that the same thing will happen this time. It is this insight – coupled with the empiricist thought that all of our “ideas” or concepts must be based on experience – that led analytic philosophers to go in one of two directions.
First, many philosophers in the empiricist tradition (itself traceable to the three great “British Empiricists”, of whom Hume was the third, after Locke and Berkeley) came to regard “causation” as an illegitimate pseudo-concept. Since we cannot trace our idea of causation to some observable feature of reality, we really don’t mean anything at all when we say that one thing caused something else. Thus we should shun causal talk all together. This was a dominant view in analytic philosophy until quite late in the 20th century. Russell famously quipped that “the law of causality” was “a relic of a bygone age, surviving, like the monarchy, only because it is erroneously supposed to do no harm”, and Peter van Inwagen said of “cause” that it is a “horrible little word”.
Second, some philosophers took the view that causation just is a matter of constant conjunction: all we mean when we say “the short circuit caused the fire” or “touching the hot iron caused me to feel pain” is just that the former kind of event is always followed by the latter kind. Unfortunately, this view as it stands is clearly hopeless. (Imagine that when I flip a switch, two lights – A and B – go on, one just after the other. The illumination of A is constantly conjoined with the illumination of B, but of course the former is not a cause of the latter; rather, they are both common effects of my flipping the switch.) But most of the major theories of causation that were developed in the latter part of the 20th century retain the guiding thought that causation reduces to regularity – they just make the story much more complicated in order to avoid this kind of counter-example. Statistical relevance theories, John Mackie’s INUS condition account, and David Lewis’s counterfactual theory all belong in this category.
Lewis, as perhaps the most significant single contributor to analytic philosophy over the last 40 years or so, constitutes one piece of evidence for the claim I made earlier about Hume’s influence on philosophy. Lewis himself conceives of a good deal of his own philosophy as contributing to Hume’s agenda, and names one of his most significant philosophical theses – Humean supervenience – after the man he calls the “greater denier of necessary connections”. Humean supervenience is, roughly, the thesis that the world is, at bottom, a mosaic of unconnected “matters of particular fact”, and that the connections we talk about – such as causal relations – depend on this pattern (which includes patterns of regularity or “constant conjunctions”). Hence the “Humean” in “Humean supervenience”: for Lewis, as for Hume (at least on some interpretations of Hume), there are no necessary connections in nature. Indeed, one might see the cornerstone of Lewis’s whole philosophical system – “modal realism” – as an aspect of this denial of necessary connections in nature. Lewis holds that modal facts – such as facts about necessity and possibility, as well as counterfactuals (“if A had been the case, then B would have been the case”) – are really facts about what is true across “possible worlds”. For example, to say that something is necessary is to say that it is true in all possible worlds. So this conception of modality itself enshrines a commitment to the thesis of “no necessary connections in nature” – since to say that, for example, something happens of necessity is not to say something about the intrinsic nature of the actual world (that is, the world we inhabit) but rather it is to say something about what happens across all possible worlds. Thus we can happily make claims about necessity without thinking that necessity is a feature of any particular possible world, including our own.
The continental tradition, of course, took a rather different tack – one that can be traced directly back to Kant and, through him, to Hume. Kant charmingly described causation on Hume’s view a “bastard of the imagination, impregnated by experience”. For Hume, causation is a product of the imagination, in the sense that the imagination is the mental faculty which, via the custom or habit that prompts our inference from causes to effect, produces the impression, and hence the idea, of necessary connection; and the imagination is “impregnated by experience” in the sense that is past experience of constant conjunction that sets up this habit. As you might have guessed from Kant’s description of Hume’s view, he wasn’t a big fan. For Kant, Hume gets the order of explanation the wrong way round. Hume, being an empiricist, thought that all of our ideas (or concepts), including the idea of causation, come from experience (or “impressions”) – so it is our experience that explains how we come to have the concepts (such as causation) that we do. (In the case of causation, though, the impression of necessary connection is an “impression of reflection” rather than a sensory impression: its source lies in the operation of our minds and not our senses.) Kant thought this was all back-to-front: we need to have the concept of causation, among others, in order to be capable of having any experience at all. So, according to Kant, that concept cannot itself be the product of experience, but must have its source in the “pure understanding”.
One way to see the broader dispute between Hume and Kant – and the way Kant saw it – is to think about the status of metaphysics as a branch of philosophical enquiry. Hume’s view was that we cannot go beyond experience when it comes to investigating the nature of reality. Take, for example, the principle that every event has a cause. On Hume’s view (arguably at any rate; nearly every claim one makes about Hume’s views has been disputed by at least one interpreter), we have good evidence that this principle is true, since all the events we have ever observed have indeed had causes. But our evidence is strictly empirical, and so the principle is capable of empirical refutation. And what goes for this principle goes across the board: a priori metaphysics – the study of reality via principles that are established solely on the basis of a priori reasoning or pure thought – is nonsense. Or, as Hume put it: “Consign it then to the flames: For it can contain nothing but sophistry and illusion”. A major aim of Kant’s Critique of Pure Reason was to reestablish metaphysics as “the Queen of all the sciences”: since it turns out that concepts such as causation are the product of the pure understanding, metaphysical principles (for example, the principle that every event has a cause) can after all be established a priori. Thus metaphysics can be rescued from Hume’s bonfire.
The downside, however, is that these principles that we can know a priori – and indeed any claim we can meaningfully make about the world at all – will apply only to (as Kant sometimes puts it) “objects of experience” (or “phenomena”) and not to “things in themselves” (noumena). So Kant is what is sometimes known as “transcendental idealist”. We can have no knowledge of – and indeed (at least on some interpretations of Kant) cannot even formulate meaningful claims about – things in themselves, because our concepts can only apply to objects of experience, and not to whatever it is that lies behind those experiences. For example, we can’t even say that our experiences of the world are caused by things in themselves. Causation is one of the a priori concepts or “categories” that determine how we experience and conceptualise the world; we therefore cannot meaningfully apply it to unconceptualised, noumenal reality.
Kant’s transcendental idealism was the starting-point for the “German idealism” of philosophers such as Fichte and Hegel. Kant is thus sometimes thought of the father (or perhaps the grandfather) of continental philosophy. So, given that Kant’s idealism was itself a direct response to Hume’s empiricism, we can perhaps think of Hume as its grandfather (or maybe its great-grandfather), albeit one who would not have entirely approved of his distant descendants.
The same might be said, however, of some of Hume’s distant descendants on the analytic side. Interestingly, despite its deep roots in Hume’s empiricism, quite a lot of contemporary analytic metaphysics would appear to assume that we can do something that neither Hume nor Kant thought was possible: we can investigate the nature of the world in itself just by thinking about it from the philosopher’s armchair. Hume and Kant would have been united in their suspicion of such “speculative metaphysics”.
Helen Beebee is professor of philosophy at the University of Birmingham and author of Hume on Causation (Routledge, 2006)