See PHILOSOPHY OF LAN-. GUAG. causation, the relation between cause and effect, or the act of bringing about an effect, which may be an event, a state, or an object (say, a statue). The concept of causation has long been recognized as one of fundamental philosophical importance. Hume called it ‘the cement of the universe’: causation is the relation that connects events and objects of this world in significant relationships. The concept of causation seems pervasively present in human discourse. It is expressed by not only ’cause’ and its cognates but by many other terms, such as ‘produce’, ‘bring about’, ‘issue’, ‘generate’, ‘result’, ‘effect’, ‘determine’, and countless others. Moreover, many common transitive verbs (‘causatives’), such as ‘kill’, ‘break’, and ‘move’, tacitly contain causal relations (e.g., killing involves causing to die). The concept of action, or doing, involves the idea that the agent (intentionally) causes a change in some object or other; similarly, the concept of perception involves the idea that the object perceived causes in the perceiver an appropriate perceptual experience. The physical concept of force, too, appears to involve causation as an essential ingredient: force is the causal agent of changes in motion. Further, causation is intimately related to explanation: to ask for an explanation of an event is, often, to ask for its cause. It is sometimes thought that our ability to make predictions, and inductive inference in general, depends on our knowledge of causal connections (or the assumption that such connections are present): the knowledge that water quenches thirst warrants the predictive inference from ‘X is swallowing water’ to ‘X’s thirst will be quenched’. More generally, the identification and systematic description of causal relations that hold in the natural world have been claimed to be the preeminent aim of science. Finally, causal concepts play a crucial role in moral and legal reasoning, e.g., in the assessment of responsibilities and liabilities. Event causation is the causation of one event by another. A sequence of causally connected events is called a causal chain. Agent causation refers to the act of an agent (person, object) in bringing about a change; thus, my opening the window (i.e., my causing the window to open) is an instance of agent causation. There is a controversy as to whether agent causation is reducible to event causation. My opening the window seems reducible to event causation since in reality a certain motion of my arms, an event, causes the window to open. Some philosophers, however, have claimed that not all cases of agent causation are so reducible. Substantival causation is the creation of a genuinely new substance, or object, rather than causing changes in preexisting substances, or merely rearranging them. The possibility of substantival causation, at least in the natural world, has been disputed by some philosophers. Event causation, however, has been the primary focus of philosophical discussion in the modern and contemporary period.
The analysis of event causation has been controversial. The following four approaches have been prominent: the regularity analysis, the counterfactual analysis, the manipulation analysis, and the probabilistic analysis. The heart of the regularity (or nomological) analysis, associated with Hume and J. S. Mill, is the idea that causally connected events must instantiate a general regularity between like kinds of events. More precisely: if c is a cause of e, there must be types or kinds of events, F and G, such that c is of kind F, e is of kind G, and events of kind F are regularly followed by events of kind G. Some take the regularity involved to be merely de facto ‘constant conjunction’ of the two event types involved; a more popular view is that the regularity must hold as a matter of ‘nomological necessity’ – i.e., it must be a ‘law.’ An even stronger view is that the regularity must represent a causal law. A law that does this job of subsuming causally connected events is called a ‘covering’ or ‘subsumptive’ law, and versions of the regularity analysis that call for such laws are often referred to as the ‘covering-law’ or ‘nomic-subsumptive’ model of causality.
The regularity analysis appears to give a satisfactory account of some aspects of our causal concepts: for example, causal claims are often tested by re-creating the event or situation claimed to be a cause and then observing whether a similar effect occurs. In other respects, however, the regularity account does not seem to fare so well: e.g., it has difficulty explaining the apparent fact that we can have knowledge of causal relations without knowledge of general laws. It seems possible to know, for instance, that someone’s contraction of the flu was caused by her exposure to a patient with the disease, although we know of no regularity between such exposures and contraction of the disease (it may well be that only a very small fraction of persons who have been exposed to flu patients contract the disease). Do I need to know general regularities about itchings and scratchings to know that the itchy sensation on my left elbow caused me to scratch it? Further, not all regularities seem to represent causal connections (e.g., Reid’s example of the succession of day and night; two successive symptoms of a disease). Distinguishing causal from non-causal regularities is one of the main problems confronting the regularity theorist. According to the counterfactual analysis, what makes an event a cause of another is the fact that if the cause event had not occurred the effect event would not have. This accords with the idea that cause is a condition that is sine qua non for the occurrence of the effect. The view that a cause is a necessary condition for the effect is based on a similar idea. The precise form of the counterfactual account depends on how counterfactuals are understood (e.g., if counterfactuals are explained in terms of laws, the counterfactual analysis may turn into a form of the regularity analysis). The counterfactual approach, too, seems to encounter various difficulties. It is true that on the basis of the fact that if Larry had watered my plants, as he had promised, my plants would not have died, I could claim that Larry’s not watering my plants caused them to die. But it is also true that if George Bush had watered my plants, they would not have died; but does that license the claim that Bush’s not watering my plants caused them to die? Also, there appear to be many cases of dependencies expressed by counterfactuals that, however, are not cases of causal dependence: e.g., if Socrates had not died, Xanthippe would not have become a widow; if I had not raised my hand, I would not have signaled. The question, then, is whether these non-causal counterfactuals can be distinguished from causal counterfactuals without the use of causal concepts. There are also questions about how we could verify counterfactuals – in particular, whether our knowledge of causal counterfactuals is ultimately dependent on knowledge of causal laws and regularities. Some have attempted to explain causation in terms of action, and this is the manipulation analysis: the cause is an event or state that we can produce at will, or otherwise manipulate, to produce a certain other event as an effect. Thus, an event is a cause of another provided that by bringing about the first event we can bring about the second. This account exploits the close connection noted earlier between the concepts of action and cause, and highlights the important role that knowledge of causal connections plays in our control of natural events. However, as an analysis of the concept of cause, it may well have things backward: the concept of action seems to be a richer and more complex concept that presupposes the concept of cause, and an analysis of cause in terms of action could be accused of circularity. The reason we think that someone’s exposure to a flu patient was the cause of her catching the disease, notwithstanding the absence of an appropriate regularity (even one of high probability), may be this: exposure to flu patients increases the probability of contracting the disease. Thus, an event, X, may be said to be a probabilistic cause of an event, Y, provided that the probability of the occurrence of Y, given that X has occurred, is greater than the antecedent probability of Y. To meet certain obvious difficulties, this rough definition must be further elaborated (e.g., to eliminate the possibility that X and Y are collateral effects of a common cause). There is also the question whether probabilistic causation is to be taken as an analysis of the general concept of causation, or as a special kind of causal relation, or perhaps only as evidence indicating the presence of a causal relationship. Probabilistic causation has of late been receiving increasing attention from philosophers. When an effect is brought about by two independent causes either of which alone would have sufficed, one speaks of causal overdetermination. Thus, a house fire might have been caused by both a short circuit and a simultaneous lightning strike; either event alone would have caused the fire, and the fire, therefore, was causally overdetermined. Whether there are actual instances of overdetermination has been questioned; one could argue that the fire that would have been caused by the short circuit alone would not have been the same fire, and similarly for the fire that would have been caused by the lightning alone. The steady buildup of pressure in a boiler would have caused it to explode but for the fact that a bomb was detonated seconds before, leading to a similar effect.