19.6: Counterfactual conditionals

The Lewis-Kratzer proposal provides a great deal of help in understanding how the meaning of a conditional sentence is compositionally derived. However, determining the right meanings for certain types of conditionals is still a significant challenge.²⁹ Counterfactuals are an especially challenging case. Consider the contrast between the hypothetical conditional in (34a) and the counterfactual conditional in (34b).³⁰

(34)    a. If Shakespeare did not write Hamlet, someone else did.
          b. If Shakespeare had not written Hamlet, someone else would have.³¹

Most English speakers would probably agree that the hypothetical conditional in (34a) is true, but would probably judge the counterfactual conditional in (34b) to be false. This contrast suggests that some different rule of interpretation must apply to counterfactual conditionals. We have said that a counterfactual conditional presupposes that the antecedent is false; but this by itself is not sufficient to cause sentence (34b) as a whole to be regarded as false. Notice that even a speaker who believes the antecedent in (34a) to be false, i.e., who believes that Shakespeare did write Hamlet, would probably judge the sentence as a whole to be true.

Ideally we would like to apply the same analysis of if to both types of conditionals, but this would make it hard to explain why the two sentences in (34), which are structurally very similar have different truth conditions. What makes the counterfactual conditional in (34b) so odd is that it seems to imply that there is (or was) something about our world which made the writing of Hamlet inevitable. The hypothetical conditional in (34a) carries no such inference. How can we account for this difference?

In the preceding section we sketched out a procedure for interpreting conditionals that do not contain an overt quantifier. In many contexts, an epistemic necessity modal has to be assumed in order to arrive at the intended interpretation. The truth conditions of the sentence are calculated by adding the content of the antecedent to what is known about the actual world in order to derive the appropriate restriction on the set of possible worlds. This procedure yields an interpretation something like (35) for the hypothetical conditional in (34a). Intuitively, this feels like a reasonable interpretation. Part of what we know about the world is that plays do not grow on trees, so if a play such as Hamlet exists (another part of what we know about the actual world), then someone must have written it.

(35)    [all w: (w is consistent with the available evidence) ∧ (the normal course of events is followed as closely as possible in w) ∧ (Shakespeare did not write             Hamlet in w)] someone else wrote Hamlet in w

With the counterfactual conditional in (34b), the process is more complex. We cannot simply add the content of the antecedent to what is known about the actual world, because the antecedent is assumed to be false in the actual world. One approach is to quantify over those possible worlds in which the antecedent is true, but which are otherwise as similar as possible to the actual world. Roughly speaking, (34b) could be paraphrased as follows: “For all worlds w in which Shakespeare did not write Hamlet, but which are otherwise as similar as possible to the actual world in the relevant ways: someone else wrote Hamlet in w.” Of course, the success of such an analysis depends on how one determines the relevant points of similarity that need to be considered.

This general approach can help explain why the counterfactual conditionals in (26a–b), repeated here as (36a–b), have different truth conditions. Sentence (36a) restricts the domain of quantification to worlds which are as similar as possible to the actual world, aside from the stipulation that kangaroos have no tails. In these worlds presumably kangaroos do not use crutches, since that would constitute an extra unforced difference as compared to the actual world. Sentence (36b) however adds the additional stipulation that kangaroos do use crutches in all the relevant worlds. For this reason, kangaroos would be more likely to topple over in the worlds relevant to evaluating (36a) than in those relevant to evaluating (36b).

(36)    a. If kangaroos had no tails, they would topple over.

          b. If kangaroos had no tails and they used crutches, they would topple over.

Now the phrase “as similar as possible” is admittedly vague, and it is reasonable to wonder whether using this criterion to restrict the domain of quantification will be very helpful in determining the meaning of a sentence. However, some authors have argued that the vagueness and context-dependence of the term are in fact good things, because counterfactuals themselves are somewhat vague, and the correct interpretation depends heavily on context.³² Consider the following examples from Quine (1960: 221):

(37)    a. If Caesar were in command, he would use the atom bomb.

          b. If Caesar were in command, he would use catapults

A given feature of the real world may be given more or less priority in determining relative closeness between two worlds depending on various contextual factors, including the purposes of the speaker. In (37a), for example, Caesar’s ruthless nature may outrank his historical setting, but in (37b) the technological resources of his era are given higher priority. The speaker’s purpose plays an important role in determining which ordering source should be applied in each case. Quine (1960: 221) expresses this principle in the following words:

The subjunctive [= counterfactual; PK] conditional depends, like indirect quotation and more so, on a dramatic projection: we feign belief in the antecedent and see how convincing we then find the consequent. What traits of the real world to suppose preserved in the feigned world of the contrary-to-fact antecedent can only be guessed from a sympathetic sense of the fabulist’s likely purpose in spinning his fable.

The pair of sentences in (34) above is quite similar to the famous pair in (38). Once again, the hypothetical conditional in (38a) seems to be true, while most people would probably judge the counterfactual conditional in (38b) to be false. However, the historical facts in this case are still somewhat controversial and poorly understood, which makes it difficult to decide which points of comparison would be relevant for determining the “most similar” possible worlds.

(38)    a. If Oswald didn’t kill Kennedy, someone else did.

          b. If Oswald hadn’t killed Kennedy, someone else would have.³³

Consider instead the counterfactual conditional in (39). While not everyone would consider this sentence to be true, it at least makes a claim that a historian could consider as a serious hypothesis:

(39)    If John Wilkes Booth hadn’t killed Abraham Lincoln, someone else would have.

What claim does (39) make? Based on our discussion above, this sentence could be paraphrased roughly as follows: “For all worlds w in which Booth did not kill Lincoln, but which are otherwise as similar as possible to the actual world in the relevant ways: someone else killed Lincoln in w.” In this context, relevant points of similarity to the real world on April 14, 1865 (the night when Lincoln was shot) might include the following:

• The on-going civil war: Gen. Lee’s army had surrendered in Virginia on April 9, 1865 but fighting continued for a few more months to the south and west;
• The location of the capital city, Washington DC, on the border between a Confederate state (Virginia) and a nominally Union state (Maryland) where many residents (including Booth) were pro-slavery and sympathetic to the Confederacy;
• The lax provisions in place for protecting the President during that era;
• The anger aroused among supporters of slavery by Abraham Lincoln’s speech of April 11, 1865, in which he announced his intention to extend voting rights to at least some African-Americans, including those who had fought for the Union.

By asserting that Lincoln’s assassination would take place in any world which shares these properties (and perhaps others) with the real world, sentence (39) seems to imply that the assassination was inevitable.

There is much more to be said about counterfactuals, but further discussion would be beyond the scope of the present book. We turn now to another use of the conditional sentence pattern, which we will argue contributes use-conditional rather than truth-conditional meaning.

---

²⁹ This section draws heavily on von Fintel (2012).

³⁰ Counterfactual and hypothetical conditionals are often referred to as “subjunctive” and “indicative” conditionals, respectively; but as we noted in §19.3, there is not always a perfect correlation between verb morphology and the degree of hypotheticality.

³¹ These examples come from Morton (2004).

³² D. Lewis (1973b: 91ff); von Fintel (2012).

³³ These examples come from Adams (1970).