Basic questions in the philosophy of causation fall into two main areas. First, there are central metaphysical questions concerning the nature of causation, such as the following: What are causal laws? What is it for two states of affairs to be causally related? Which are primary -- causal relations between states of affairs, or causal laws? How are causal facts related to noncausal facts? How can one explain the formal properties of causation -- such as irreflexivity, asymmetry, and transitivity? What is the ground of the direction of causation?

Second, there are issues concerning the epistemology of causation. Can causal relations be directly observed? How can the existence of causal laws be established? What statistical methods can be used to confirm causal hypotheses, and how can those methods be justified?

Such metaphysical and epistemological issues first came sharply into focus as a result of David Hume's penetrating scrutiny of causation, and the theses that he defended (Hume 1739-40, 1748). On the metaphysical side, HUME argued for the view that causal facts are reducible to noncausal facts, while, on the epistemological side, Hume argued that causal relations between events, rather than being directly observable, can only be known by establishing the existence of "constant conjunctions," or general laws.

The major metaphysical choice is between realist and reductionist approaches to causation. According to the latter, all causal facts are logically supervenient upon the totality of noncausal states of affairs. It is logically impossible, then, for two possible worlds to disagree with respect to some causal fact while agreeing completely with respect to all noncausal facts.

Reductionist approaches to causation have dominated the philosophical landscape since the time of Hume, and many different accounts have been advanced. Three types of approaches are, however, especially important. First, there are approaches that start out from the general notion of a law of nature, then define the ideas of necessary and sufficient nomological conditions, and, finally, employ the latter concepts to explain what it is for one state of affairs to cause another (Mackie 1965). Second, there are approaches that employ subjunctive conditionals in an attempt to give a counterfactual analysis of causation (Lewis 1973, 1979, 1986). Third, there are probabilistic approaches, where the central idea is that a cause must, in some way, make its effect more likely (Reichenbach 1956; Good 1961-62; Suppes 1970; Eells 1991; Mellor 1995).

Each of these three types of approaches faces difficulties specific to it. The attempt to analyze causation in terms of nomological conditions, for example, is hard pressed to provide any account of the direction of causation -- a problem that quickly becomes evident when one notices that one state of affairs may be a nomologically sufficient condition for another either because the former is causally sufficient for the latter, or because, on the contrary, the latter is causally necessary for the former.

In the case of counterfactual approaches to causation, a crucial problem is that traditional analyses of subjunctive conditionals employ causal notions. Alternative accounts have been proposed, involving similarity relations over possible worlds. But these alternative accounts are exposed to decisive objections.

Finally, there are also specific problems for probabilistic accounts, two of which are especially important. First, probabilistic accounts have struggled to find an interpretation of their central claim -- that causes must, in some way, make their effects more likely -- that is not open to counterexamples. Second, probabilistic approaches to causation typically involve the very counterintuitive consequence that a completely deterministic world could not contain any causally related events (Tooley 1987).

There are also other objections, however -- of a very serious sort -- that tell against all reductionist approaches. First, one can show that some worlds with probabilistic laws may agree with respect to all causal laws and all noncausal facts, yet differ with respect to causal relations between events. So some causal facts not only are not logically supervenient upon the totality of noncausal states of affairs, they are not even supervenient upon the combination of that totality together with all causal laws (Carroll 1994; Tooley 1987).

Second, there are arguments showing that no reductionist approach to causation can account for the direction of causation. One problem, for example, is that very simple worlds containing causally related events may be devoid of all of the noncausal features upon which reductionist accounts rely to define the direction of causation -- such as increasing entropy and the presence of open causal forks. Another problem is that, given deterministic laws of an appropriate sort -- such as, for example, the laws of Newtonian physics -- one can show that, corresponding to some worlds where a reductionist account assigns the correct direction to causation, there will be inverted worlds where the direction of causation is opposite to that specified by any reductionist account (Tooley 1990b).

Given these difficulties, it is natural to explore realist alternatives, and the most plausible form of realism involves viewing causation as a theoretical relation between states of affairs. The development of this type of approach, however, presupposed solutions to two problems that confronted realist interpretations of theories in general. First, there was the semantical problem of how one could even make sense of a realist interpretation of theoretical terms. Second, there was the epistemological problem of how one could justify any statement containing theoretical terms when those terms were interpreted as referring to unobservable states of affairs.

It is not surprising, then, that until those obstacles were surmounted, reductionist approaches to causation held sway. Now, however, satisfactory answers to the above problems are available. Thus, in the case of the semantical problem, one promising approach involves the use of existential quantification ranging over properties and relations to assign a realist interpretation to theoretical terms (Lewis 1970), while, as regards the epistemological problem, there is now widespread acceptance of a type of inductive reasoning -- variously referred to as the method of hypothesis, abduction, hypothetico-deductive reasoning, and inference to the best explanation -- that will allow one to justify theoretical claims realistically construed.

These philosophical developments have made it possible, then, to take seriously the idea that causation is a theoretical relation. To construct such an account, however, one needs to set out an analytically true theory of causation, and, at present, only one such theory has been worked out in any detail (Tooley 1987, 1990a). It seems likely, however, both that more theories will be proposed in the near future, and that, given the difficulties that reductionism faces, an account of the nature of causation along realist lines will turn out to be correct.

Some philosophers have maintained that causation is a basic and unanalyzable relation that is directly observable (Anscombe 1971; Fales 1990). That view, however, is exposed to some very serious objections (Tooley 1990a). Of these, one of the most important concerns the fact that causal beliefs are often established on the basis of statistical information -- using methods that, especially within the social sciences, are very sophisticated. But if causation is a basic and unanalyzable relation, how can non-causal, statistical information possibly serve to establish causal hypotheses?

Recently, this question of how causal hypotheses can be established using statistical information has been the subject of intense investigation. The basic approach has been, first, to identify fundamental principles relating causation to probability. Two such principles that have been suggested as very important, for example, derive from the work of Hans Reichenbach (1949, 1956):

The Screening Off Principle: If A causes C only via B, then, given B, A and C are statistically independent.

The Common Cause Principle: If A and B are statistically dependent, and neither causes the other, then there is a common cause of A and B.

Then, second, one attempts to show that those principles can be used to justify algorithms that will enable one to move from information about statistical relationships to conclusions about causal relations (Glymour et al. 1987; Glymour, Spirtes, and Scheines 1991; Spirtes, Glymour, and Scheines 1993)

This is an interesting and important research program. In its present form, however, it suffers from certain defects. First, some of the principles employed are unsound. Reichenbach's common cause principle, for example, is shown to be false by the inverted worlds objection to causal reductionism mentioned above. Second, if reductionism is false, then it is a mistake to look for algorithms that will specify, for some set of statistical relationships, what the relevant causal relations must be, as given a realist view of causation, different causal relations may underlie a given set of statistical relationships. A sound algorithm, accordingly, will generate only a probability distribution over possible causal relations.

The basic conclusion, in short, is that an investigation into the epistemology of causation cannot proceed in isolation from consideration of the metaphysics of causation, and if it turns out that a reductionist view of causation is untenable, then one needs to employ a realist account that connects causation to probability, and then isolate algorithms that can be justified on the basis of such an account.

See also

-- Michael Tooley


Anscombe, G. E. M. (1971). Causality and Determination. Cambridge: Cambridge University Press.

Carroll, J. W. (1994). Laws of Nature. Cambridge: Cambridge University Press.

Eells, E. (1991). Probabilistic Causality. Cambridge: Cambridge University Press.

Ehring, D. (1997). Causation and Persistence. New York: Oxford University Press.

Fales, E. (1990). Causation and Universals. London: Routledge.

Glymour, C., R. Scheines, P. Spirtes, and K. Kelly. (1987). Discovering Causal Structure. Orlando, FL: Academic Press.

Glymour, C., P. Spirtes, and R. Scheines. (1991). Causal inference. Erkenntnis 35:151-189.

Good, I. J. (1961-62). A causal calculus I-II. British Journal for the Philosophy of Science 11: 305 - 318, 12: 43 - 51.

Hume, D. (1739-40). A Treatise of Human Nature. London: Millar.

Hume, D. (1748). An Enquiry Concerning Human Understanding. London: Millar.

Lewis, D. (1970). How to define theoretical terms. Journal of Philosophy 67:427-446.

Lewis, D. (1973). Causation. Journal of Philosophy 70:556-567. Reprinted, with postscripts, in Philosophical Papers, vol. 2. Oxford: Oxford University Press, 1986.

Lewis, D. (1979). Counterfactual dependence and time's arrow. Noûs 13:455-476. Reprinted, with postscripts, in Philosophical Papers, vol. 2. Oxford: Oxford University Press, 1986.

Lewis, D. (1986). Philosophical Papers, vol. 2. Oxford: Oxford University Press.

Mackie, J. L. (1965). Causes and conditions. American Philosophical Quarterly 2:245-264.

Mackie, J. L. (1974). The Cement of the Universe. Oxford: Oxford University Press.

Mellor, D. H. (1995). The Facts of Causation. London: Routledge.

Menzies, P. (1989). Probabilistic causation and causal processes: a critique of Lewis. Philosophy of Science 56:642-663.

Reichenbach, H. (1949). The Theory of Probability. Berkeley: University of California Press.

Reichenbach, H. (1956). The Direction of Time. Berkeley: University of California Press.

Salmon, W. C. (1980). Probabilistic causality. Pacific Philosophical Quarterly 61:50-74.

Salmon, Wesley C. (1984). Scientific Explanation and the Causal Structure of the World. Princeton, NJ: Princeton University Press.

Sosa, E., and M. Tooley, Eds. (1993). Causation. Oxford: Oxford University Press.

Spirtes, P., C. Glymour, and R. Scheines. (1993). Causation, Prediction, and Search. New York: Springer-Verlag.

Strawson, G. (1989). The Secret Connexion: Causation, Realism, and David Hume. Oxford: Oxford University Press.

Suppes, P. (1970). A Probabilistic Theory of Causality. Amsterdam: North-Holland Publishing Company.

Tooley, M. (1987). Causation: A Realist Approach. Oxford: Oxford University Press.

Tooley, M. (1990a). The nature of causation: a singularist account. In David Copp, Ed., Canadian Philosophers.Canadian Journal of Philosophy Supplement 16:271-322.

Tooley, M. (1990b). Causation: reductionism versus realism. Philosophy and Phenomenological Research 50:215-236.

Tooley, M. (1996). Causation. In Donald M. Borchert, Ed., The Encyclopedia of Philosophy, Supplement. New York: Simon and Schuster Macmillan, pp. 72-75.

von Wright, G. H. (1971). Explanation and Understanding. Ithaca, NY: Cornell University Press.

Woodward, J. (1992). Realism about laws. Erkenntnis 36:181-218 .