The Humean analysis of causation reads as follows:

"We may define a cause to be an object, followed by another, and where all the objects similar to the first, are followed by objects similar to the second." (Inquiry)

A counterfactual condition is typically added to this, popularized by Lewis, so that the revised Humean causation generally reads as follows:

  1. If A had not occurred, B would not have occurred.
  2. If A had occurred, B would have occurred.
  3. A and B both occurred.

The general objection that is made to the Humean analysis of causation is to say that there are clear examples in which the Humean analysis either fails due to its ignoring of cases that don't favor its conditions, or in its failed presuppositions that causation is primarily observed in distinct temporal events.

The first sort of objections include what are known as 'finks' and 'antidotes'. Finks were popularized by C.B Martin's example of an electro-fink, which prevents the typical effect of 'B' (current running to conductor), and even causes unprecedented effects that the given cause 'A' (touching the wire with a conductor) could not give in itself. The reason this poses a possible threat to the Humean analysis of causation is because the Humean philosophy of regularities states that the occurrence of 'A' would result in the occurrence of 'B'. Lewis responds to this objection with the 'Reformed Conditional Analysis', which essentially stresses the would in his first and second statements. That is to say, Lewis reiterates that the given cause A would have resulted in effect B if its properties were not affected by another counterfactual cause, such as the electro-fink. I think this response is somewhat adequate, so 'finks' aren't to be answered to.

On the other hand, I am curious about the other line of objections, namely 'antidotes'. The objection of 'antidotes' are best explained in the following example: Let's say someone is poisoned. The properties of poison which allow it to kill a person are in no way affected by an antidote, which is unlike the previous example of the electro-fink (which directly affects the wire's properties). Rather, the antidote affects the physiology of the body. So here we find an example of some causation in which the cause A (poison) does not result in the effect B (death), but in no way was the cause A affected so as to alter its properties (which is the condition of the Reformed Conditional Analysis.) If the underlying criticism of Hume is in skepticism over what can truly be known of causality, this would in turn render the Humean analysis descriptively inadequate and contrary to what is truly observed.

The second string of the general objection against Hume includes examples that seem to do away with a 'trigger' event or temporally divided causes and effects, upon which Hume believes that we conceive of causes and effects (this is evidenced in the typical simplistic illustration he gives to understand causation, in the billiard ball). One such example includes General Relativity's spontaneous manifestation of gravitational powers which are in continuous interaction with space-time (Molnar 2003). Another example includes two books causing each other to stay upright by their leaning on each other, and a magnet staying on a fridge. Rather than interpreting such causes as "an object, followed by another", or two separate events that yield some regularity, we are able to simply analyze interactive forces occurring upon each simultaneously so as to produce in one state a certain effect (for books, of staying upright rather than falling down, and for the magnet, of staying on the fridge rather than falling off).

If Hume's attempt is to demonstrate the inability of the mind to conceive, directly or indirectly, of causal powers, it seems that Hume doesn't have much of a backing granted these objections, which compel rather justified inference of belief in causal powers. However, how does the proponent of Humean analyses of causation answer to such objections?

(Forgive the length of this question. The intent involved was to provide an extensive explanation of my understanding of Hume and more modern alternatives to his general theory of causation such as those named 'counterfactual', and to explain thoroughly the objections made, as I see them.)

  • Hume's point is not that mind can not conceive of causation, but that it can never confirm it directly, causation is inferred. It is inferred from past experience, which by its nature can bestow nothing more than regularity. These examples seem to deal with difficulties of defining what cause is satisfactorily, but that is secondary to Hume.
    – Conifold
    Oct 9, 2015 at 1:52
  • 1
    I think the question could still use some editing, but you're basically asking how to do contemporary Humeans defend the denial of metaphysical causation?
    – virmaior
    Oct 9, 2015 at 2:03
  • @virmaior More so how Humeans adequately deny causal powers in light of examples of causation that seem to require a reference to them.
    – Chosen One
    Oct 9, 2015 at 2:19
  • @Conifold I suppose an explanation of why Hume believes causal powers cannot be observed would be in order than, granted that we observe cases where if a thing were not present than a certain outcome would occur, one example being a safety mechanism in a nuclear reactor. That a nuclear reactor has the continuing power to explode is exemplified in the necessity of the safety mechanism (if you want to take the chance of not believing me, do us the favor of not getting a job at a nuclear powerplant). The power of the reactor to explode is still immanent despite the absence of its effect.
    – Chosen One
    Oct 9, 2015 at 2:44
  • @Conifold What is more, an explanation of Hume's standard for what qualifies modes of acquiring knowledge would also be beneficial, since degrading a conclusion on the basis that it is 'inferential' is hardly justified in itself, and requires an explanation as to its meaning (which I assume there is, and which I would be more than delighted to hear).
    – Chosen One
    Oct 9, 2015 at 2:44

3 Answers 3


As for antidotes/electro fink:

Causation in a wider sense could be understood as "in most cases". That could be because there are possible intermediating variables, like antidotes interrupting a poisoning in progress.

But what Hume wants to say is that there is always the threat of one case that cannot be ruled under our laws of nature (or inductive knowledge in general). And that basically is the birth of our modern theory of science: every law of nature must be falsifiable. But they are nevertheless, in theory, thought as a general, necessary claim about how the connection between cause and effect is. Therefore it does not object his statements, because you just picked examples where it is not "causation" in a sense applicable to describe events in nature as they in fact are:

Most poisons cause some amino acids to denaturate. That, over time, causes circulatory collapse if the cause "denaturation" is not removed/stopped. This, as long as the cause "denaturation" is not removed and the circulatory collapse made reversable, causes death within a certain period of time. You see the different use of "causing" in the different narratives? The dependence on time and other events?

As for equilibriums:

"Causation" in a two-way-situation (equilibrium) can be described as "causing each other", but since it is pure potential capacity (as a physical and philosophical term) without actualization, in fact nothing would have been caused. You describe a potential capacity. That's it. Someone/something has to have caused the books to stand against each other or the stable gravitional system, though.

Causation traditionally is one form of movement/change since Aristotle's Physics. As an equilibrium is a state, not a change of states, it is simply a strange use of the term "causation" if you describe the situation as if the books would cause each other to stand up. So again, you are using the term in another way.

Explaining the usage of "causation" in philosophy:

I have used some philosophical terms against your objections that I think I should back up. Causation as a term includes a cause and an effect. The effect is traditionally a change of state, as I wrote above. In addition, causation in a strict sense (like Hume is referring to) means what you could name "direct" causation.

Of course, you could say that the murder of Sarajevo caused WW I, but this wouldn't be a proper understanding of causation. But why?

That is because of what Kant states (and proofs thereafter!) in the Second Analogy of the Critique of Pure Reason (B232):

All changes take place according to the law of the connection of Cause and Effect.

Which means that

a) Causation is change.

b) There has to be a law of the connection, which includes necessity (see A216|B263)

Where there is no necessary rule between cause and effect, there is no causation.

And, to come back to Hume, Kant agrees with him saying that there is no necessity in experience, because all knowledge from experience is only inductive and ipso facto arbitrary. Therefore laws of causation (transcendental laws of nature) have to be a priori for him, resulting from the fact that we can only conceive objects in time.

To summarize:

Taking the meaning of "causation" seriously as a philosophical term, all your objections can be answered. You can use "causation" analogously, but you should be careful which characteristics of causation can be retrieved through the analogon.

The late Wittgenstein would perhaps answer that you are playing another language game using the term in the way you do. That's why you have to make the rules of the games explicit first and then decide whether the other usage makes sense or not. But you should not question the other usage without understanding it first. That's why I referred to Kant as a source of philosophical understanding of the term.

Philosophical objections are the strongest if they are objections from within the very same conceptual framework (or language game, like I named it).

  • As an aside, we are talking about Aristotle's "efficient cause" and ignoring his other three types: material, formal, final. The term "causality" isn't so consistent. But is it really accurate to say the Kant "agrees" with Hume? I take Kant to mean there is no "pure experience," so it will always, of necessity, contain causation as an invariant. Hume would l not, I think, go so far. There is nothing "outside of experience" that secures causes against, say, random fluctuations. His "habit of mind" is itself "naturalizable" and not really on a par with the Kantian a priori. Oct 10, 2015 at 20:05
  • Maybe you're right to question if they would have agreed philosophically. I'm pretty sure they wouldn't, because Kant basically criticized all his philosophical predecessors. But the point Kant indeed agreed with was that regarding experience, all knowledge is inductive and therefore arbitrary.
    – Philip Klöcking
    Oct 10, 2015 at 20:09
  • Not to quibble, but I would have said Kant says all proper knowledge entails experiential content, is inductive, is limited, but is precisely not arbitrary. It is categorically founded and as "universal" as things will ever get "for us." Whereas Hume might have said "arbitrary" in the long run, though I'm not sure. Oct 10, 2015 at 21:23
  • 1
    That is precisely where Kant has to introduce "knowledge a priori" as necessary (and therefore philosophical!) knowledge, because the content/materia of experience (not the form) is arbitrary. He in fact says that by asking for the necessary conditions of experience (which, of course, is using experiantial content in some way, but only in abstract) we can escape Humean problems regarding knowledge. And that is why Kant said in 1781 that before there never was philosophy and Hegel in 1806 (after using Kants method in all consequence) that philosophy has ended. 25 years, that is.
    – Philip Klöcking
    Oct 10, 2015 at 21:30
  • 1
    The term "in abstract" actually is from Hegels reception of Kant. Thinking is thinking the concrete object of experience in abstract, its form. It refers to the ontological difference of the things that there are and the things as they are. Thinking is thinking the objects as they are, not that there are. As they are constituted by our cognitive faculty, to be precise in a kantian sense.
    – Philip Klöcking
    Oct 11, 2015 at 10:19

According to Andrea Falcon's SEP article, "Aristotle on Causality", this was first tackled by Aristotle in his four-fold theory of causation. Aristotle recognized that the theory of causation he had inherited typically distinguished two types of cause: material and efficient and thus he needed to justify his introduction of final cause, which he did:

Physics II 8 contains Aristotle's most general defense of final causality. Here Aristotle establishes that explaining nature requires final causality by discussing a difficulty that may be advanced by an opponent who denies that there are final causes in nature. Aristotle shows that an opponent who claims that material and efficient causes alone suffice to explain natural change fails to account for their characteristic regularity.

Since in the world there are many things occurring at once, sometimes in one way and at other times in another, the necessary character can sometimes be difficult to discern. One might speculate this was why astronomy was seen originally as the science par excellence that reveals the hidden order as the stars move to the music of the spheres. In fact we only need look up at the sky to observe this, and see it directly as a theater of permanence, in contrast to the earth where all things are subject to change: coming to be (genesis) and passing away (phthora).

Of course, on the cosmological scale the heavens are as subject to change as much as all on earth are, but this is to our senses magnified by both the instruments of our practical and theoretical sciences. On the human scale - that which we see directly by our own eyes - this distinction remains. Then one can ask: is this regularity, this permanence that we see in the skies - the stars in their courses - is this illusionary or real? Or, more deeply, does this reflect something real?

In Aristotle's language, Hume showed that causality is a form of regularity, whose regularity needs explaining. For Hume this was the mind's habit of regularizing.

Kant replaced this habit by what he called a transcendental argument: arguing from what is to what the conditions must be to make this possible.

But given Aristotle's thesis that such an explanation of the regularity found in nature is a final cause, it's an intriguing question whether Hume or Kant regarded it as such.

Falcon, Andrea, "Aristotle on Causality", The Stanford Encyclopedia of Philosophy (Spring 2015 Edition), Edward N. Zalta (ed.), URL = https://plato.stanford.edu/archives/spr2015/entries/aristotle-causality/.


The proponent of a Humean stance has a very simple, straightforward, and widely accepted response: science is probabilistic.

That was Hume's point. He is not concerned with whether or not we can usefully talk about "causality," the orders of events, or the psychology of attribution. He is doing a reductio ad absurdum of empiricism, pointing out that "forces" are not themselves observables, nor can they be shown to have any metaphysical necessity. No matter how many times we observe something, we must always make the inductive leap to assume it continues that way. There is no final proof that this "law" will always be the case.

At that time, the assumption that "true" scientific laws were absolutely secure, in theory, and that "causation" was essential to the whole scientific project was largely unchallenged, except by the possible interventions of God. Like Berkeley's attack on "matter," Hume's simple billiard-ball model of "causality" was meant to show just how much metaphysical stuff remained unaccountable in the most basic "laws" of science. It is all induction, statistical regularities, local phenomena. Which is not "truth" but "good enough." Until the black swan appears, or we attain speeds where Newtonian mechanics no longer holds, or the sun indeed fails to rise. Kant awakes in alarm, refutations pour forth....

Though the issue is still open, today most scientists and "pragmatic" philosophers would accept Hume's probabilistic account, which entered science proper with Boltzmann's explanation of entropy and became doctrinal with the stunning success of quantum theory. Your counter-examples may raise interesting questions about language or the ontology of "forces," but I can't really make out that they are refutations of Hume.

  • That is if such causality is considered to be derived inductively. The counterarguments are to provide examples that pressure us to provide an explanation that is more necessary than simply an induction in the way Hume thinks (they are indeed meant to challenge the notion that we experience only causally distinct events). What is being challenged is not so much Hume's understanding of induction, but rather his application of such induction to causality.
    – Chosen One
    Oct 23, 2015 at 20:32
  • In that case, sure. Hume is making an argument about efficient cause, which does not take into account "strange attractors" and such. So while there may be valid expansions of causality in science, I would not describe them as "refuting" Hume. The examples do not seem to me to present some deeper, more foundational theory of causality per se. They do not supersede probabilistic accounts. But I will have to read them over again. Oct 23, 2015 at 20:47

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .