The Direction of Causation

Abstract

O N A simple regularity theory of causation, a cause is both necessary and sufficient for its effect in such a sense that it follows automatically that the effect is equally sufficient and necessary for its cause, and the only difference between the relation of cause to effect and that of effect to cause is that of temporal order: the cause precedes the effect. Given that two events are causally connected in a direct line with one another, whichever of the two comes first in time is called the cause. It seems, however, that this cannot be the whole of the difference between causing and being the efect of. A cause is commonly taken to explain its effect in a way in which an effect does not explain its cause, even where we can retrodict the cause from the effect as confidently as we can predict the effect from the cause. Consequently, if two events occur close together, this may be explained by their being joint effects of a common cause, but it will not be explained by their being joint causes of a common effect. The same principle applies if two features, or two kinds of events, tend frequently to occur together. A tendency for white hair and pink eyes to go together in various species of animal may be explained by these features being joint effects of a single gene; but we could not explain a tendency for high humidity and falling barometric pressure to go together by their being joint causes of the falling of rain. Indeed, it is questionable whether temporal order is even part of the difference between causing and being the efect of. Admittedly, we believe that causes always, or nearly always, precede their effects, and we often use temporal order to decide which of two causally connected events is the cause. But on the other hand, we do not rule out the possibility of an effect occurring simultaneously with its cause. What is more, it seems at least conceivable that there should, in special cases, be evidence for backward causation, for an effect preceding its cause in time. For example, it seems conceiv-