counterfactual analysis of causation See CAUSA -. TIO. counterfactuals, also called contrary-to-fact conditionals, subjunctive conditionals that presuppose the falsity of their antecedents, such as ‘If Hitler had invaded England, Germany would have won’ and ‘If I were you, I’d run’.
Conditionals (or hypothetical statements) are compound statements of the form ‘If p, (then) q’, or equivalently ‘q if p’. Component p is described as the antecedent (protasis) and q as the consequent (apodosis). A conditional like ‘If Oswald did not kill Kennedy, then someone else did’ is called indicative, because both the antecedent and consequent are in the indicative mood. One like ‘If Oswald had not killed Kennedy, then someone else would have’ is subjunctive. Many subjunctive and all indicative conditionals are open, presupposing nothing about the antecedent. Unlike ‘If Bob had won, he’d be rich’, neither ‘If Bob should have won, he would be rich’ nor ‘If Bob won, he is rich’ implies that Bob did not win. Counterfactuals presuppose, rather than assert, the falsity of their antecedents. ‘If Reagan had been president, he would have been famous’ seems inappropriate and out of place, but not false, given that Reagan was president. The difference between counterfactual and open subjunctives is less important logically than that between subjunctives and indicatives. Whereas the indicative conditional about Kennedy is true, the subjunctive is probably false. Replace ‘someone’ with ‘no one’ and the truth-values reverse.
The most interesting logical feature of counterfactuals is that they are not truth-functional. A truth-functional compound is one whose truth-value is completely determined in every possible case by the truth-values of its components. For example, the falsity of ‘The President is a grandmother’ and ‘The President is childless’ logically entails the falsity of ‘The President is a grandmother and childless’: all conjunctions with false conjuncts are false. But whereas ‘If the President were a grandmother, the President would be childless’ is false, other counterfactuals with equally false components are true, such as ‘If the President were a grandmother, the President would be a mother’. The truth-value of a counterfactual is determined in part by the specific content of its components. This property is shared by indicative and subjunctive conditionals generally, as can be seen by varying the wording of the example. In marked contrast, the material conditional, p / q, of modern logic, defined as meaning that either p is false or q is true, is completely truth-functional. ‘The President is a grandmother / The President is childless’ is just as true as ‘The President is a grandmother / The President is a mother’. While stronger than the material conditional, the counterfactual is weaker than the strict conditional, p U q, of modern modal logic, which says that p / q is necessarily true. ‘If the switch had been flipped, the light would be on’ may in fact be true even though it is possible for the switch to have been flipped without the light’s being on because the bulb could have burned out. The fact that counterfactuals are neither strict nor material conditionals generated the problem of counterfactual conditionals (raised by Chisholm and Goodman): What are the truth conditions of a counterfactual, and how are they determined by its components? According to the ‘metalinguistic’ approach, which resembles the deductive-nomological model of explanation, a counterfactual is true when its antecedent conjoined with laws of nature and statements of background conditions logically entails its consequent. On this account, ‘If the switch had been flipped the light would be on’ is true because the statement that the switch was flipped, plus the laws of electricity and statements describing the condition and arrangement of the circuitry, entail that the light is on. The main problem is to specify which facts are ‘fixed’ for any given counterfactual and context. The background conditions cannot include the denials of the antecedent or the consequent, even though they are true, nor anything else that would not be true if the antecedent were. Counteridenticals, whose antecedents assert identities, highlight the difficulty: the background for ‘If I were you, I’d run’ must include facts about my character and your situation, but not vice versa. Counterlegals like ‘Newton’s laws would fail if planets had rectangular orbits’, whose antecedents deny laws of nature, show that even the set of laws cannot be all-inclusive. Another leading approach (pioneered by Robert C. Stalnaker and David K. Lewis) extends the possible worlds semantics developed for modal logic, saying that a counterfactual is true when its consequent is true in the nearest possible world in which the antecedent is true. The counterfactual about the switch is true on this account provided a world in which the switch was flipped and the light is on is closer to the actual world than one in which the switch was flipped but the light is not on. The main problem is to specify which world is nearest for any given counterfactual and context. The difference between indicative and subjunctive conditionals can be accounted for in terms of either a different set of background conditions or a different measure of nearness.
Counterfactuals turn up in a variety of philosophical contexts. To distinguish laws like ‘All copper conducts’ from equally true generalizations like ‘Everything in my pocket conducts’, some have observed that while anything would conduct if it were copper, not everything would conduct if it were in my pocket. And to have a disposition like solubility, it does not suffice to be either dissolving or not in water: it must in addition be true that the object would dissolve if it were in water. It has similarly been suggested that one event is the cause of another only if the latter would not have occurred if the former had not; that an action is free only if the agent could or would have done otherwise if he had wanted to; that a person is in a particular mental state only if he would behave in certain ways given certain stimuli; and that an action is right only if a completely rational and fully informed agent would choose it.
See also CAUSATION, POSSIBLE WORLDS. W.A.D.