Skip to content

What is Wrong with Hume’s Analysis of Causality

Reviewing Hume's argument

 Hume argues that - in speaking of the relationship between cause and effect - this relationship cannot be clarified by using terms like "efficacy, agency, power, force, energy, necessity, connexion, and productive quality," (T 157) because, when we investigate the ideas to which these latter terms refer, we find their provenance as ideas is questionable. So far as Hume can see, our experience provides no impressions which could be said to be the originals of these ideas. (T 161) 

Since these ideas of efficacy (and its synonyms) are essential to our ordinary understanding of causal relations, if they are to be given up, our ordinary understanding of causal relations needs to be revised and Hume is ready to oblige. He proposed that to speak of one thing causing another should be understood as an elliptical way of saying (the impression of) B has always been conjoined to (the impression of A) in my past experience and - due to my psychological constitution - I have come to expect B to be conjoined to A in the future. The necessity of the inference to B that I draw from my encounter with A is no more or less than the "propensity, which custom produces to pass from an object to the idea of its usual attendant." (T l65 ) 

Hume is fully aware that "of all the paradoxes, which I ... have advance[d] in the course of this treatise, the present one is the most violent, and that 'tis only by dint of solid proof and reasoning I can ever hope it will have admission, and overcome the inveterate prejudices of mankind." (T 166) Hume then reminds his reader of the remorseless steps of the argument which lead to his solid proof that our commonly held views concerning the efficacy of causes are mistaken. "Before we are reconciled to this doctrine, how often must we repeat to ourselves, that the simple view of any two objects or actions, however related, can never give us any idea of power, or of a connexion betwixt them: that this idea arises from the repetition of their union: that the repetition neither discovers nor causes any thing in the objects, but has an influence only on the mind by that customary transition it produces: that this customary transition is, therefore, the same with the power or necessity; which are consequently qualities of perceptions, not of objects, and are internally felt by the soul, and not perceived externally in bodies? (T 166) Hume's crucial assumption Hume's confidence in his argument is justified - given that we are dealing with two distinct objects. Thus if our two objects are A and B, and they have exhibited constant conjunction in the past - and have thus come to my attention as two objects which are causally related - my idea of A is provided by my impression of A, and my idea of B is provided by my impression of B. Now ex hypothesis, there are no other ideas involved in the constant conjunction of A and B, and therefore no other impressions are available from which my idea of some further link between A and B might arise. 

Thus any relationship between A and B - apart from the fact that they have been constantly conjoined - would have to reduce to the fact that, as so conjoined, they must be contiguous in space and follow each other in time. In other words, it must be the case that if A and B are two distinct objects, the only relations they can fall into are spatial and temporal. Thus - on the assumption that A and B are distinct - it is not surprising that our experience of their conjunction does not provide us with some additional impression which could account for the idea of a causal link between them, an impression which would satisfy us as accounting for our idea that causes bring about or produce their effects. Now to advance this argument Hume had first to establish that his ''any two objects . . . however related" are, first and foremost, two distinct objects, ie., separable from each other. Clearly, if they were joined in some fashion, we would experience an impression of this link which could provide us with our idea of the necessary connection between the 'two' objects. A great deal depends on this point and Hume does his best to secure his view concerning the distinctness of objects. The two principles grounding his view are: 1) "[E]very thing which is different, is distinguishable, and every thing which is distinguishable, is separable by the imagination." (T 233) and 2) What can be thought of separately, can exist separately. As Hume puts it: "'Tis an established maxim in metaphysics, That whatever the mind clearly conceives includes the idea of possible existence" (T 32) or in another version: "Whatever is clearly conceived may exist; and whatever is clearly conceived, after any manner, may exist after the same manner." (T 233) 

Thus following the first principle, since causes are different from effects, they can be distinguished from effects - we do so distinguish them whenever we make a statement concerning the cause of a fire or a motor accident. From this it follows, by the second principle, that the causes designated - since they could be thought about separately - could have occurred without their effects and vice-versa. I can certainly imagine such a state of affairs so surely it could exist. A direct consequence of the application of these two principles is that spatial and temporal relations between things are to be regarded as being contingent, because different arrangements of things in space and time can be imagined and could, therefore, exist without any contradiction. Thus any conjunction - and a fortiori any constant conjunction - between two things can be regarded as a contingent conjunction. In short, on Hume's view, there is simply no room for the idea of a necessary link between cause and effect, and indeed, no relationship between any two distinct things which is not reducible to the contingent fact that A was contiguous with B - a spatial relationship - and that A followed B, or was co- temporary with it, or preceded it - a temporal relationship (plus the further contingent relation of constant conjunction). 

Thus it is the fact that objects are distinct and separable which permits Hume to drive home his argument against causal necessity. The fact that we do not discover any impression of necessity - linking a cause to an effect - is not surprising, once it has been established through the above premisses, that the world, or our experience of it, is constituted by a concatenation of objects or impressions which are 'atoms' ie., essentially distinct and thus separable. As such, they can only be contingently related to each other.

 Challenging Hume's assumption 

To challenge Hume we need to challenge the truth of his first principle. Later I will show that his second principle can be challenged if the challenge to the first is successful. Our criticism of the first principle takes the form of asking whether Hume does not beg the question when he says "Everything which is different is distinguishable and hence separable." Certainly everything which is different is distinguishable, but we should, I think, pause to entertain the question: are things distinguishable and hence separable because they are different, or are they different and hence separable because they have been distinguished? In other words, if the nature of our experience is not fundamentally atomistic but must instead be regarded as a continuum - within which we may nominate or distinguish those aspects of the continuum which interest us (and then call them 'things' or 'events') - then we could argue that Hume does beg the question, and that his destructive analysis of causal necessity need not be accepted without some further argument on his part. Hume's argument Time Hume begins his discussion of Time by saying: "The idea of time, being derived from the succession of our perceptions of every kind, ideas as well as impressions. . . . " (T 34-5) But do our impressions occur successively (ie., as distinct perceptions following one another) and is it the case that only if they were like this could I derive the idea of time? If I watch a ripple travelling across the surface of a pond do I not derive the idea of time from this experience even though there is no sense of a succession of perceptions (distinct, separable impressions) the way there intuitively is in Hume's example of five notes played on a flute succeeding each other and - in their successive relationship - giving us the idea of time? (T 36) 

Clearly, whether we think that our intuitive idea of time depends on our experience of successive, distinct impressions (or on our experience of a continuum), is revealed in the examples we offer. In any case the question is open to dispute and Hume can certainly be accused of begging the question. Temporal succession is not a problem when we consider the continuum. The items we nominate as "separate" things within the Continuum are not actually separate so that - on this view - temporal succession is not an issue. However , if we take the atomistic view then the idea of the one thing ( the cause) preceding the other (the effect) becomes a bit tricky: thus: is there any temporal gap between the cause and the effect? And if there is not are they still separate things? Hume was aware of this issue and attempted to deal with it as follows. 

Temporal Succession 

The restriction of temporal succession, ie. that the cause should immediately precede the effect, is essential, because where there is no temporal succession there could be no experience of resembling instances of temporal succession and thus of constant conjunction, which, according to Hume, is the real source of our propensity to link cause and effect. How then does Hume deal with this crucial concept of temporal succession. Hume's first argument runs as follows "Some pretend that 'tis not absolutely necessary a cause shou'd precede its effect; but that any object or action, in the very first moment of its existence, may exert it's productive quality, and give rise to another object or action perfectly co-temporary with itself. But beside that experience in most instances seems to contradict this opinion, we may establish the relation of priority by a kind of inference or reasoning." (T 76, my italics) Hume does not provide instances so let us consider a few. Take an inflated balloon in the shape of a tube: constrict it slightly in the center by making a circle round it with your thumb and finger. Then squeeze one end of the balloon and watch the other end bulge out. Does this constitute an example of "one object or action [giving rise] to another object or action perfectly co- temporary with itself"? I would be inclined to say "Yes". It is certainly not obvious that this is a false or misleading description of what happens. Take an example more favourable to Hume, that of one billiard ball striking another. Here it looks like the one event precedes the other or at least is not perfectly co-temporary with it. But we can well imagine watching a film of the event taken by a superb slow-motion camera. 

Now on the basis of those slow- motion video tapes of 'crucial moments in sport' with which we are all familiar, I would be inclined to believe that such a film of the billiard balls in collision would provide us with an impression in which it would be hard to tell when the first action ended and the second began: it might well look as if "one object or action was giving rise to another object or action perfectly co- temporary with itself." A third example: place some iron filings on a sheet of paper and move a magnet underneath it. It certainly looks as if the moving magnet "in the very first moment of its existence may exert its productive quality, and give rise to another object or action (the movement of the iron filings) perfectly co- temporary with itself." 

These examples show, I think, that there is certainly no prima facie case for Hume's contention that "experience in most instances seems to contradict this claim" (viz., that an "object or action, in the very first moment of its existence, may exert it's productive quality, and give rise to another object or action perfectly co-temporary with itself.") 

Hume's reductio argument 

However, Hume has another argument for the relation of priority of time in the cause before the effect, not drawn from experience, but one dependent on "a kind of inference or reasoning." an argument which is - in fact - a proposed reductio ad absurdum of the idea of co-temporary cause and effect. Hume begins: "'Tis an established maxim both in natural and moral philosophy, that an object which exists for any time in its full perfection without producing another, is not its sole cause; but is assisted by some other principle, which pushs it from its state of inactivity and makes it exert that energy, of which it was secretely possest." (T 76) He is here thinking of a stone, for example, which could start an avalanche, but not until a passing mountain goat dislodges it. He then asserts: "Now if any cause may be perfectly co-temporary with its effect, 'tis certain, according to this maxim, that they must all of them be so . . . . If one cause were co-temporary with its effect, and this effect with its effect and so on 'tis plain there wou'd be no such thing as succession, and all objects must be coexistent." (T 76) which, of course, is absurd. Hume has argued that if we admit the possibility of co-temporary causal relations then, in effect, everything must happen at once: and "The consequence [of this] would be no less than the destruction of that succession of causes, which we observe in the world, and indeed the utter annihilation of time." [T 76] 

However we should pause to consider whether co-temporary causal relations do lead to this dramatic result. In the three examples given above (the balloon, etc.) there is certainly a prima facie case for the view that the cause and the effect are co-temporary but time is not thereby annihilated. Instead the situation is one in which the idea of time is not thought of as derived from our experience of a succession of discrete events each prior to the next, but rather as the transformation or evolution of one event into another. Under this construal, the separateness of the two events in time is only a function of their arbitrary nomination (by us) as distinct aspects of the evolving world which interests us. Take the case of our balloon being squeezed. Are there two events here, a cause: the balloon being squeezed at one end and an effect: the balloon bulging at the other? Or is there only one thing going on: the balloon being squeezed and [simultaneously] bulging at the other end? Clearly, to say that these two aspects of the one event are co-temporary does not - as Hume suggests - have the implication that this event, as a whole, does not take any time to happen. 

An alternative understanding of causality I have argued that Hume is mistaken here and the mistake rests finally on his atomistic view of the objects or actions which he thinks of as constituting experience. If they are truly atomistic in character, then, of course, there can be no connection between them that is not external and therefore contingent, and it immediately follows that any impression that we may be under that they are connected in some other fashion must vanish when we see the situation clearly, ie., draw the implications of this atomism. 

However, I have tried to show that his arguments for the atomistic character of the objects and events which constitute experience are not convincing and as a consequence I am led to the view that a continuum conception of experience seems to square more closely with our experience of things and actions. What does a continuum view imply so far as our understanding of causality is concerned? It implies that I cannot think, E.G., of a car accident, without also thinking - as a matter of necessity - that the event [which I nominate as] immediately preceding it is in fact continuous with the accident, ie., not distinct or separable from it. Furthermore, it is in virtue of this continuity between the 'two' events that I regard this nominated preceding portion of the accident as the cause of it. 

However if the nominated events are inseparable within the continuum then simply distinguishing them (by nominating one portion of the continuous event as the cause of another portion) does not - contra Hume - involve the possibility of separating them. Here Hume's second principle is open to attack. Where we take the continuum conception as basic, imagining something as separate - on the grounds that it is nominally distinguishable - does not amount to anything more than picturing the portion nominated in the mind's eye. But, on the continuum view, to picture it as existing, is to picture it as only a nominally distinguishable aspect of the continuum, ie., as something which is, in fact, joined inseparably to the rest of the continuum. Hume thought that to say of something that it existed added no more to our idea of it (T 66-7) but again, this assumption is not self-evident: where we accept the continuum conception as basic, to think of something as existing is, necessarily, to embed it in the continuum, to give it inseparable links of continuity within the evolving world. How an acceptance of the continuum view affects our understanding of causal relations If this is true than causality should never be thought of as a process which links up events within the continuum, or a process whereby one distinct event produces or brings into being, another distinct event. 

Events, although distinguishable - we can, after all, nominate and hence distinguish as many events in a given portion of the continuum as we like - are not in fact separated within the continuum and thus they never stand in any actual external relation to each other. Nor are they internally related. They are nominal entities - ontological abstractions which owe their being as distinct events to their having been arbitrarily distinguished as extending from just this point to that point along the continuum. As such, these nominal entities are not related to each other within the continuum, except contingently. Thus event A is associated contingently with event B insofar as they share an arbitrary dividing line. Event A runs from x to y and event B from y to z. The continuum as a whole is not subject to any relationships, external or internal between its parts because it has no separable parts. Causality thus is not a conception which has an ontological application within the continuum. It takes two to do the causal tango and the continuum is on its own. The continuum could be referred to as the cause of itself but this is just a way of saying that the idea of a causal relationship linking two events is not applicable to the continuum. Where then does the notion of a causal relationship properly apply? The idea of a causal relationship between two objects or events can be properly applied only within isolated or effectively isolated portions of the continuum. This isolation can be set up within the laboratory proper, or can be attributed to those quasi-laboratories which nature provides. Within these isolable continua, various factors may be introduced: E.G., we can turn on the stove and see what happens to the water in the pot on the element as it begins to heat. In this way proper external relations between initially distinct events may be set up by introducing new factors into our isolated portion of the continuum. And by doing so, we can then observe what the introduction of event x causes, or brings about, within this isolated continuum. Under such circumstances we can repeat our experiments and confirm various relationships as causal. 

Our confidence in the causal laws we extract from such experiments is in direct proportion to our confidence that the isolated continuum is indeed isolated, and that, therefore, the introduced factor alone must be responsible for what then happens within this isolated portion of the continuum. Experimental situations are, then, the proper arena for the use of causal language. Talk about causal relations translates badly when we try to use it to describe events which are not isolated from the continuum . Thus we feel uneasy when we talk about the causes of the First World War, or the Depression. In such cases, where the nominated portion of the continuum cannot be isolated from the rest of the continuum, any event or events that we nominate as causes seem to be insufficient as causes. In these cases we seem to feel that everything which preceded the nominated event is relevant to its turning out the way it did. 

This is what makes the writing of history such a difficult business when it is regarded as an exercise in nominating the causes of the selected portion of the continuum under discussion. We appreciate intuitively that the concept of causality is not applicable to a situation in which we cannot isolate the process under examination and introduce external factors at will, ie., subject the process to experiment, to determine why things happened the way they did. 


To sum up: if objects or actions are, in fact, not distinct, but continuous with each other; if the world is a continuum, then causal language can only be appropriate under artificial circumstances. These will be circumstances in which a portion of the continuum is isolated in such a way that objects or actions can be introduced into these 'laboratories' to see what happens when such external factors enter into or become continuous with the isolated portion of the continuum. Under such conditions, when we assume that x (the heat introduced) brings about or produces, or is the agent responsible for, or supplies the force or energy, which makes y (the water in the pot) boil, we can regard the introduced factor x as the cause of the change within the isolated continuum. And in doing so we are not in the least tempted to imagine that the relation between the cause and the effect is contingent. Why? Nothing that happens within a continuum is contingent because, its being continuous with itself means that what happens to it reflects the nature of the continuum as a whole, not one isolated part of it being 'affected' [as if by magic] by another externally related part.

 Finally, I would argue that our use of the notion of necessity in our causal language - the usage which so exercised Hume - is a function of our confidence that the isolated portion of the continuum which we are investigating in our laboratory - is indeed isolated. If we are confident that this is the case, when we do introduce some new factor into it, then whatever happens, must - as a matter of logical necessity - be the effect of that introduced factor, since, ex hypothesi, we are confident there is no other factor which could be responsible for the change we observe. Attributions of causal necessity are thus a function of the logical necessity which comes into play when we are sure that the introduced factor is the only one which could be responsible for the changes observed.1 Insofar as we are not confident that the introduced factor is the only one involved we hesitate to say that the introduced factor necessarily caused the observed change. 1 Compare this to Richard Rorty's discussion of the relationship between confidence and certainty in Philosophy and the Mirror of Nature, Princeton University Press, Princeton, l979, p. 157.